mirror/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2025-01-08 20:42:41 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity
b70332d806
opencv/modules/highgui/src/cap_msmf.cpp

4179 lines
127 KiB
C++
Raw Normal View History

Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
/*M///////////////////////////////////////////////////////////////////////////////////////
//
// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
//
// By downloading, copying, installing or using the software you agree to this license.
// If you do not agree to this license, do not download, install,
// copy or use the software.
//
//
// Intel License Agreement
// For Open Source Computer Vision Library
//
// Copyright (C) 2000, Intel Corporation, all rights reserved.
// Third party copyrights are property of their respective owners.
//
// Redistribution and use in source and binary forms, with or without modification,
// are permitted provided that the following conditions are met:
//
// * Redistribution's of source code must retain the above copyright notice,
// this list of conditions and the following disclaimer.
//
// * Redistribution's in binary form must reproduce the above copyright notice,
// this list of conditions and the following disclaimer in the documentation
// and/or other materials provided with the distribution.
//
// * The name of Intel Corporation may not be used to endorse or promote products
// derived from this software without specific prior written permission.
//
// This software is provided by the copyright holders and contributors "as is" and
// any express or implied warranties, including, but not limited to, the implied
// warranties of merchantability and fitness for a particular purpose are disclaimed.
// In no event shall the Intel Corporation or contributors be liable for any direct,
// indirect, incidental, special, exemplary, or consequential damages
// (including, but not limited to, procurement of substitute goods or services;
// loss of use, data, or profits; or business interruption) however caused
// and on any theory of liability, whether in contract, strict liability,
// or tort (including negligence or otherwise) arising in any way out of
// the use of this software, even if advised of the possibility of such damage.
//
//M*/
#include "precomp.hpp"
#if (defined WIN32 || defined _WIN32) && defined HAVE_MSMF
/*
Media Foundation-based Video Capturing module is based on
videoInput library by Evgeny Pereguda:
http://www.codeproject.com/Articles/559437/Capturing-of-video-from-web-camera-on-Windows-7-an
Originaly licensed under The Code Project Open License (CPOL) 1.02:
http://www.codeproject.com/info/cpol10.aspx
*/
#include <windows.h>
#include <guiddef.h>
#include <mfidl.h>
#include <Mfapi.h>
#include <mfplay.h>
#include <mfobjects.h>
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
#include <strsafe.h>
Media Foundation-based VideoWriter added
2013-05-06 18:36:51 +08:00
#include <Mfreadwrite.h>
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
#include <new>
#include <map>
#include <vector>
#include <string>
#include <stdio.h>
#include <stdarg.h>
#include <string.h>
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
#pragma warning(disable:4503)
#pragma comment(lib, "mfplat")
#pragma comment(lib, "mf")
#pragma comment(lib, "mfuuid")
#pragma comment(lib, "Strmiids")
Media Foundation-based VideoWriter added
2013-05-06 18:36:51 +08:00
#pragma comment(lib, "Mfreadwrite")
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
#pragma comment(lib, "MinCore_Downlevel")
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
// for ComPtr usage
#include <wrl/client.h>
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
using namespace Microsoft::WRL;
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#include <mferror.h>
#ifdef HAVE_WINRT
#ifdef __cplusplus_winrt
#include <agile.h>
#include <vccorlib.h>
#endif
#include <wrl\async.h>
#include <wrl\implements.h>
#include <wrl\module.h>
#include <wrl\wrappers\corewrappers.h>
#include <windows.media.capture.h>
#include <windows.devices.enumeration.h>
#include <concrt.h>
#include <ppltasks.h>
using namespace Microsoft::WRL::Wrappers;
#endif
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
struct IMFMediaType;
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#ifndef HAVE_WINRT
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
struct IMFActivate;
struct IMFMediaSource;
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#endif
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
struct IMFAttributes;
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
namespace
{
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
template <class T> void SafeRelease(T **ppT)
{
if (*ppT)
{
(*ppT)->Release();
*ppT = NULL;
}
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
/// Class for printing info into consol
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
class DebugPrintOut
{
public:
~DebugPrintOut(void);
static DebugPrintOut& getInstance();
void printOut(const wchar_t *format, ...);
void setVerbose(bool state);
bool verbose;
Code review notes fixed.
2013-04-03 18:48:23 +08:00
private:
DebugPrintOut(void);
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
};
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#include "cap_msmf.hpp"
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
// Structure for collecting info about types of video, which are supported by current video device
struct MediaType
{
unsigned int MF_MT_FRAME_SIZE;
unsigned int height;
unsigned int width;
unsigned int MF_MT_YUV_MATRIX;
unsigned int MF_MT_VIDEO_LIGHTING;
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
int MF_MT_DEFAULT_STRIDE; // stride is negative if image is bottom-up
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
unsigned int MF_MT_VIDEO_CHROMA_SITING;
GUID MF_MT_AM_FORMAT_TYPE;
wchar_t *pMF_MT_AM_FORMAT_TYPEName;
unsigned int MF_MT_FIXED_SIZE_SAMPLES;
unsigned int MF_MT_VIDEO_NOMINAL_RANGE;
GetProperty method for MSMF VideoCapture implemented.
2013-05-23 00:50:54 +08:00
unsigned int MF_MT_FRAME_RATE_NUMERATOR;
unsigned int MF_MT_FRAME_RATE_DENOMINATOR;
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
unsigned int MF_MT_PIXEL_ASPECT_RATIO;
unsigned int MF_MT_PIXEL_ASPECT_RATIO_low;
unsigned int MF_MT_ALL_SAMPLES_INDEPENDENT;
unsigned int MF_MT_FRAME_RATE_RANGE_MIN;
unsigned int MF_MT_FRAME_RATE_RANGE_MIN_low;
unsigned int MF_MT_SAMPLE_SIZE;
unsigned int MF_MT_VIDEO_PRIMARIES;
unsigned int MF_MT_INTERLACE_MODE;
unsigned int MF_MT_FRAME_RATE_RANGE_MAX;
unsigned int MF_MT_FRAME_RATE_RANGE_MAX_low;
GUID MF_MT_MAJOR_TYPE;
GUID MF_MT_SUBTYPE;
wchar_t *pMF_MT_MAJOR_TYPEName;
wchar_t *pMF_MT_SUBTYPEName;
MediaType();
~MediaType();
void Clear();
};
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
/// Class for parsing info from IMFMediaType into the local MediaType
class FormatReader
{
public:
static MediaType Read(IMFMediaType *pType);
~FormatReader(void);
private:
FormatReader(void);
};
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
DWORD WINAPI MainThreadFunction( LPVOID lpParam );
typedef void(*emergensyStopEventCallback)(int, void *);
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
class RawImage
{
public:
~RawImage(void);
// Function of creation of the instance of the class
static long CreateInstance(RawImage **ppRImage,unsigned int size);
void setCopy(const BYTE * pSampleBuffer);
void fastCopy(const BYTE * pSampleBuffer);
unsigned char * getpPixels();
bool isNew();
unsigned int getSize();
private:
bool ri_new;
unsigned int ri_size;
unsigned char *ri_pixels;
RawImage(unsigned int size);
};
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
class ImageGrabberCallback : public IMFSampleGrabberSinkCallback
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
{
public:
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
void pauseGrabbing();
void resumeGrabbing();
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
RawImage *getRawImage();
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
// IMFClockStateSink methods
STDMETHODIMP OnClockStart(MFTIME hnsSystemTime, LONGLONG llClockStartOffset);
STDMETHODIMP OnClockStop(MFTIME hnsSystemTime);
STDMETHODIMP OnClockPause(MFTIME hnsSystemTime);
STDMETHODIMP OnClockRestart(MFTIME hnsSystemTime);
STDMETHODIMP OnClockSetRate(MFTIME hnsSystemTime, float flRate);
// IMFSampleGrabberSinkCallback methods
STDMETHODIMP OnSetPresentationClock(IMFPresentationClock* pClock);
STDMETHODIMP OnProcessSample(REFGUID guidMajorMediaType, DWORD dwSampleFlags,
LONGLONG llSampleTime, LONGLONG llSampleDuration, const BYTE * pSampleBuffer,
DWORD dwSampleSize);
STDMETHODIMP OnShutdown();
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
Media Foundation camera capture fixed. Camera-based VideoCapture updated to fit changes in ImageGrabber from prev commit
2013-05-15 20:12:25 +08:00
const HANDLE ig_hFrameReady;
const HANDLE ig_hFrameGrabbed;
const HANDLE ig_hFinish;
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
protected:
ImageGrabberCallback(bool synchronous);
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
bool ig_RIE;
bool ig_Close;
Media Foundation camera capture fixed. Camera-based VideoCapture updated to fit changes in ImageGrabber from prev commit
2013-05-15 20:12:25 +08:00
bool ig_Synchronous;
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
long m_cRef;
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
RawImage *ig_RIFirst;
RawImage *ig_RISecond;
RawImage *ig_RIOut;
};
#ifdef HAVE_WINRT
extern const __declspec(selectany) WCHAR RuntimeClass_CV_ImageGrabberWinRT[] = L"cv.ImageGrabberWinRT";
class ImageGrabberWinRT :
public Microsoft::WRL::RuntimeClass<
Microsoft::WRL::RuntimeClassFlags< Microsoft::WRL::RuntimeClassType::WinRtClassicComMix>,
IMFSampleGrabberSinkCallback>, public ImageGrabberCallback
{
InspectableClass(RuntimeClass_CV_ImageGrabberWinRT, BaseTrust)
public:
ImageGrabberWinRT(bool synchronous);
~ImageGrabberWinRT(void);
HRESULT initImageGrabber(MAKE_WRL_REF(_MediaCapture) pSource,
GUID VideoFormat);
HRESULT startGrabbing(MAKE_WRL_REF(_AsyncAction)* action);
HRESULT stopGrabbing(MAKE_WRL_REF(_AsyncAction)* action);
// IMFClockStateSink methods
STDMETHODIMP OnClockStart(MFTIME hnsSystemTime, LONGLONG llClockStartOffset) { return ImageGrabberCallback::OnClockStart(hnsSystemTime, llClockStartOffset); }
STDMETHODIMP OnClockStop(MFTIME hnsSystemTime) { return ImageGrabberCallback::OnClockStop(hnsSystemTime); }
STDMETHODIMP OnClockPause(MFTIME hnsSystemTime) { return ImageGrabberCallback::OnClockPause(hnsSystemTime); }
STDMETHODIMP OnClockRestart(MFTIME hnsSystemTime) { return ImageGrabberCallback::OnClockRestart(hnsSystemTime); }
STDMETHODIMP OnClockSetRate(MFTIME hnsSystemTime, float flRate) { return ImageGrabberCallback::OnClockSetRate(hnsSystemTime, flRate); }
// IMFSampleGrabberSinkCallback methods
STDMETHODIMP OnSetPresentationClock(IMFPresentationClock* pClock) { return ImageGrabberCallback::OnSetPresentationClock(pClock); }
STDMETHODIMP OnProcessSample(REFGUID guidMajorMediaType, DWORD dwSampleFlags,
LONGLONG llSampleTime, LONGLONG llSampleDuration, const BYTE * pSampleBuffer,
DWORD dwSampleSize) { return ImageGrabberCallback::OnProcessSample(guidMajorMediaType, dwSampleFlags, llSampleTime, llSampleDuration, pSampleBuffer, dwSampleSize); }
STDMETHODIMP OnShutdown() { return ImageGrabberCallback::OnShutdown(); }
// Function of creation of the instance of the class
static HRESULT CreateInstance(ImageGrabberWinRT **ppIG, bool synchronous = false);
private:
MAKE_WRL_AGILE_REF(_MediaCapture) ig_pMedCapSource;
MediaSink* ig_pMediaSink;
};
#endif
// Class for grabbing image from video stream
class ImageGrabber : public ImageGrabberCallback
{
public:
~ImageGrabber(void);
HRESULT initImageGrabber(IMFMediaSource *pSource, GUID VideoFormat);
HRESULT startGrabbing(void);
void stopGrabbing();
// IUnknown methods
STDMETHODIMP QueryInterface(REFIID iid, void** ppv);
STDMETHODIMP_(ULONG) AddRef();
STDMETHODIMP_(ULONG) Release();
// Function of creation of the instance of the class
static HRESULT CreateInstance(ImageGrabber **ppIG, unsigned int deviceID, bool synchronous = false);
private:
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
unsigned int ig_DeviceID;
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
IMFMediaSource *ig_pSource;
IMFMediaSession *ig_pSession;
IMFTopology *ig_pTopology;
Media Foundation camera capture fixed. Camera-based VideoCapture updated to fit changes in ImageGrabber from prev commit
2013-05-15 20:12:25 +08:00
ImageGrabber(unsigned int deviceID, bool synchronous);
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
HRESULT CreateTopology(IMFMediaSource *pSource, IMFActivate *pSinkActivate, IMFTopology **ppTopo);
Windows RT related code cleanup. Windows RT sample cleanup. Video I/O code cleanup.
2013-06-24 16:10:18 +08:00
HRESULT AddSourceNode(IMFTopology *pTopology, IMFMediaSource *pSource,
IMFPresentationDescriptor *pPD, IMFStreamDescriptor *pSD, IMFTopologyNode **ppNode);
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
HRESULT AddOutputNode(IMFTopology *pTopology, IMFActivate *pActivate, DWORD dwId, IMFTopologyNode **ppNode);
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
};
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
/// Class for controlling of thread of the grabbing raw data from video device
class ImageGrabberThread
{
friend DWORD WINAPI MainThreadFunction( LPVOID lpParam );
public:
~ImageGrabberThread(void);
Media Foundation camera capture fixed. Camera-based VideoCapture updated to fit changes in ImageGrabber from prev commit
2013-05-15 20:12:25 +08:00
static HRESULT CreateInstance(ImageGrabberThread **ppIGT, IMFMediaSource *pSource, unsigned int deviceID, bool synchronious = false);
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
void start();
void stop();
void setEmergencyStopEvent(void *userData, void(*func)(int, void *));
ImageGrabber *getImageGrabber();
protected:
virtual void run();
private:
Media Foundation camera capture fixed. Camera-based VideoCapture updated to fit changes in ImageGrabber from prev commit
2013-05-15 20:12:25 +08:00
ImageGrabberThread(IMFMediaSource *pSource, unsigned int deviceID, bool synchronious);
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
HANDLE igt_Handle;
DWORD igt_ThreadIdArray;
ImageGrabber *igt_pImageGrabber;
emergensyStopEventCallback igt_func;
void *igt_userData;
bool igt_stop;
unsigned int igt_DeviceID;
};
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
// Structure for collecting info about one parametr of current video device
struct Parametr
{
long CurrentValue;
long Min;
long Max;
long Step;
long Default;
long Flag;
Parametr();
};
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
// Structure for collecting info about 17 parametrs of current video device
struct CamParametrs
{
Parametr Brightness;
Parametr Contrast;
Parametr Hue;
Parametr Saturation;
Parametr Sharpness;
Parametr Gamma;
Parametr ColorEnable;
Parametr WhiteBalance;
Parametr BacklightCompensation;
Parametr Gain;
Parametr Pan;
Parametr Tilt;
Parametr Roll;
Parametr Zoom;
Parametr Exposure;
Parametr Iris;
Parametr Focus;
};
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
typedef std::wstring String;
typedef std::vector<int> vectorNum;
typedef std::map<String, vectorNum> SUBTYPEMap;
typedef std::map<UINT64, SUBTYPEMap> FrameRateMap;
typedef void(*emergensyStopEventCallback)(int, void *);
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
/// Class for controlling of video device
class videoDevice
{
public:
videoDevice(void);
~videoDevice(void);
void closeDevice();
CamParametrs getParametrs();
void setParametrs(CamParametrs parametrs);
void setEmergencyStopEvent(void *userData, void(*func)(int, void *));
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#ifdef HAVE_WINRT
long readInfoOfDevice(MAKE_WRL_REF(_IDeviceInformation) pDevice, unsigned int Num);
void waitForDevice()
{
if (vd_pAction) {
HRESULT hr;
DO_ACTION_SYNCHRONOUSLY(hr, vd_pAction, GET_CURRENT_CONTEXT);
vd_pAction = nullptr;
}
}
#else
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
long readInfoOfDevice(IMFActivate *pActivate, unsigned int Num);
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#endif
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
wchar_t *getName();
int getCountFormats();
unsigned int getWidth();
unsigned int getHeight();
MediaType getFormat(unsigned int id);
bool setupDevice(unsigned int w, unsigned int h, unsigned int idealFramerate = 0);
bool setupDevice(unsigned int id);
bool isDeviceSetup();
bool isDeviceMediaSource();
bool isDeviceRawDataSource();
bool isFrameNew();
IMFMediaSource *getMediaSource();
RawImage *getRawImageOut();
private:
enum typeLock
{
MediaSourceLock,
RawDataLock,
OpenLock
} vd_LockOut;
wchar_t *vd_pFriendlyName;
ImageGrabberThread *vd_pImGrTh;
CamParametrs vd_PrevParametrs;
unsigned int vd_Width;
unsigned int vd_Height;
unsigned int vd_CurrentNumber;
bool vd_IsSetuped;
std::map<UINT64, FrameRateMap> vd_CaptureFormats;
std::vector<MediaType> vd_CurrentFormats;
IMFMediaSource *vd_pSource;
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#ifdef HAVE_WINRT
MAKE_WRL_AGILE_REF(_MediaCapture) vd_pMedCap;
IMedCapFailHandler* vd_pMedCapFail;
ImageGrabberWinRT *vd_pImGr;
MAKE_WRL_REF(_AsyncAction) vd_pAction;
Concurrency::critical_section vd_lock;
#endif
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
emergensyStopEventCallback vd_func;
void *vd_userData;
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
HRESULT enumerateCaptureFormats(IMFMediaSource *pSource);
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
long setDeviceFormat(IMFMediaSource *pSource, unsigned long dwFormatIndex);
void buildLibraryofTypes();
int findType(unsigned int size, unsigned int frameRate = 0);
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#ifdef HAVE_WINRT
HRESULT enumerateCaptureFormats(MAKE_WRL_REF(_MediaCapture) pSource);
long setDeviceFormat(MAKE_WRL_REF(_MediaCapture) pSource, unsigned long dwFormatIndex, MAKE_WRL_REF(_AsyncAction)* pAction);
long resetDevice(MAKE_WRL_REF(_IDeviceInformation) pDevice);
long checkDevice(_DeviceClass devClass, DEFINE_TASK<HRESULT>* pTask, MAKE_WRL_REF(_IDeviceInformation)* ppDevice);
#else
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
long resetDevice(IMFActivate *pActivate);
long checkDevice(IMFAttributes *pAttributes, IMFActivate **pDevice);
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#endif
long initDevice();
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
};
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
/// Class for managing of list of video devices
class videoDevices
{
public:
~videoDevices(void);
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#ifdef HAVE_WINRT
long initDevices(_DeviceClass devClass);
void waitInit() {
if (vds_enumTask) {
HRESULT hr;
DO_ACTION_SYNCHRONOUSLY(hr, vds_enumTask, GET_CURRENT_CONTEXT);
vds_enumTask = nullptr;
}
}
#else
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
long initDevices(IMFAttributes *pAttributes);
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#endif
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
static videoDevices& getInstance();
videoDevice *getDevice(unsigned int i);
unsigned int getCount();
void clearDevices();
private:
UINT32 count;
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#ifdef HAVE_WINRT
MAKE_WRL_REF(_AsyncAction) vds_enumTask;
#endif
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
std::vector<videoDevice *> vds_Devices;
videoDevices(void);
};
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
// Class for creating of Media Foundation context
class Media_Foundation
{
public:
virtual ~Media_Foundation(void);
static Media_Foundation& getInstance();
bool buildListOfDevices();
private:
Media_Foundation(void);
};
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
/// The only visiable class for controlling of video devices in format singelton
class videoInput
{
public:
virtual ~videoInput(void);
// Getting of static instance of videoInput class
static videoInput& getInstance();
// Closing video device with deviceID
void closeDevice(int deviceID);
// Setting callback function for emergency events(for example: removing video device with deviceID) with userData
void setEmergencyStopEvent(int deviceID, void *userData, void(*func)(int, void *));
// Closing all devices
void closeAllDevices();
// Getting of parametrs of video device with deviceID
CamParametrs getParametrs(int deviceID);
// Setting of parametrs of video device with deviceID
void setParametrs(int deviceID, CamParametrs parametrs);
// Getting numbers of existence videodevices with listing in consol
unsigned int listDevices(bool silent = false);
// Getting numbers of formats, which are supported by videodevice with deviceID
unsigned int getCountFormats(int deviceID);
// Getting width of image, which is getting from videodevice with deviceID
unsigned int getWidth(int deviceID);
// Getting height of image, which is getting from videodevice with deviceID
unsigned int getHeight(int deviceID);
// Getting name of videodevice with deviceID
wchar_t *getNameVideoDevice(int deviceID);
// Getting interface MediaSource for Media Foundation from videodevice with deviceID
IMFMediaSource *getMediaSource(int deviceID);
Code review notes fixed.
2013-04-03 18:48:23 +08:00
// Getting format with id, which is supported by videodevice with deviceID
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
MediaType getFormat(int deviceID, int unsigned id);
// Checking of existence of the suitable video devices
bool isDevicesAcceable();
// Checking of using the videodevice with deviceID
bool isDeviceSetup(int deviceID);
// Checking of using MediaSource from videodevice with deviceID
bool isDeviceMediaSource(int deviceID);
// Checking of using Raw Data of pixels from videodevice with deviceID
bool isDeviceRawDataSource(int deviceID);
// Setting of the state of outprinting info in console
static void setVerbose(bool state);
// Initialization of video device with deviceID by media type with id
bool setupDevice(int deviceID, unsigned int id = 0);
// Initialization of video device with deviceID by wisth w, height h and fps idealFramerate
bool setupDevice(int deviceID, unsigned int w, unsigned int h, unsigned int idealFramerate = 30);
Code review notes fixed.
2013-04-03 18:48:23 +08:00
// Checking of recivig of new frame from video device with deviceID
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
bool isFrameNew(int deviceID);
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#ifdef HAVE_WINRT
void waitForDevice(int deviceID);
#endif
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
// Writing of Raw Data pixels from video device with deviceID with correction of RedAndBlue flipping flipRedAndBlue and vertical flipping flipImage
bool getPixels(int deviceID, unsigned char * pixels, bool flipRedAndBlue = false, bool flipImage = false);
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
static void processPixels(unsigned char * src, unsigned char * dst, unsigned int width, unsigned int height, unsigned int bpp, bool bRGB, bool bFlip);
Code review notes fixed.
2013-04-03 18:48:23 +08:00
private:
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
bool accessToDevices;
videoInput(void);
void updateListOfDevices();
};
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
DebugPrintOut::DebugPrintOut(void):verbose(true)
{
}
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
DebugPrintOut::~DebugPrintOut(void)
{
}
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
DebugPrintOut& DebugPrintOut::getInstance()
{
static DebugPrintOut instance;
return instance;
}
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
void DebugPrintOut::printOut(const wchar_t *format, ...)
{
if(verbose)
{
int i = 0;
wchar_t *p = NULL;
va_list args;
va_start(args, format);
if(wcscmp(format, L"%i"))
{
i = va_arg (args, int);
}
if(wcscmp(format, L"%s"))
{
p = va_arg (args, wchar_t *);
}
wprintf(format, i,p);
va_end (args);
}
}
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
void DebugPrintOut::setVerbose(bool state)
{
verbose = state;
}
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
LPCWSTR GetGUIDNameConstNew(const GUID& guid);
HRESULT GetGUIDNameNew(const GUID& guid, WCHAR **ppwsz);
HRESULT LogAttributeValueByIndexNew(IMFAttributes *pAttr, DWORD index);
HRESULT SpecialCaseAttributeValueNew(GUID guid, const PROPVARIANT& var, MediaType &out);
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
unsigned int *GetParametr(GUID guid, MediaType &out)
{
Code review notes fixed.
2013-04-03 18:48:23 +08:00
if(guid == MF_MT_YUV_MATRIX)
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
return &(out.MF_MT_YUV_MATRIX);
Code review notes fixed.
2013-04-03 18:48:23 +08:00
if(guid == MF_MT_VIDEO_LIGHTING)
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
return &(out.MF_MT_VIDEO_LIGHTING);
Code review notes fixed.
2013-04-03 18:48:23 +08:00
if(guid == MF_MT_DEFAULT_STRIDE)
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
return (unsigned int*)&(out.MF_MT_DEFAULT_STRIDE);
Code review notes fixed.
2013-04-03 18:48:23 +08:00
if(guid == MF_MT_VIDEO_CHROMA_SITING)
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
return &(out.MF_MT_VIDEO_CHROMA_SITING);
Code review notes fixed.
2013-04-03 18:48:23 +08:00
if(guid == MF_MT_VIDEO_NOMINAL_RANGE)
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
return &(out.MF_MT_VIDEO_NOMINAL_RANGE);
Code review notes fixed.
2013-04-03 18:48:23 +08:00
if(guid == MF_MT_ALL_SAMPLES_INDEPENDENT)
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
return &(out.MF_MT_ALL_SAMPLES_INDEPENDENT);
Code review notes fixed.
2013-04-03 18:48:23 +08:00
if(guid == MF_MT_FIXED_SIZE_SAMPLES)
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
return &(out.MF_MT_FIXED_SIZE_SAMPLES);
Code review notes fixed.
2013-04-03 18:48:23 +08:00
if(guid == MF_MT_SAMPLE_SIZE)
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
return &(out.MF_MT_SAMPLE_SIZE);
Code review notes fixed.
2013-04-03 18:48:23 +08:00
if(guid == MF_MT_VIDEO_PRIMARIES)
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
return &(out.MF_MT_VIDEO_PRIMARIES);
Code review notes fixed.
2013-04-03 18:48:23 +08:00
if(guid == MF_MT_INTERLACE_MODE)
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
return &(out.MF_MT_INTERLACE_MODE);
return NULL;
}
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
HRESULT LogAttributeValueByIndexNew(IMFAttributes *pAttr, DWORD index, MediaType &out)
{
WCHAR *pGuidName = NULL;
WCHAR *pGuidValName = NULL;
GUID guid = { 0 };
PROPVARIANT var;
PropVariantInit(&var);
HRESULT hr = pAttr->GetItemByIndex(index, &guid, &var);
if (FAILED(hr))
{
goto done;
}
hr = GetGUIDNameNew(guid, &pGuidName);
if (FAILED(hr))
{
goto done;
}
hr = SpecialCaseAttributeValueNew(guid, var, out);
unsigned int *p;
if (FAILED(hr))
{
goto done;
}
if (hr == S_FALSE)
{
switch (var.vt)
{
case VT_UI4:
p = GetParametr(guid, out);
Code review notes fixed.
2013-04-03 18:48:23 +08:00
if(p)
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
{
*p = var.ulVal;
}
break;
case VT_UI8:
break;
case VT_R8:
break;
case VT_CLSID:
if(guid == MF_MT_AM_FORMAT_TYPE)
{
hr = GetGUIDNameNew(*var.puuid, &pGuidValName);
if (SUCCEEDED(hr))
{
out.MF_MT_AM_FORMAT_TYPE = MF_MT_AM_FORMAT_TYPE;
out.pMF_MT_AM_FORMAT_TYPEName = pGuidValName;
pGuidValName = NULL;
}
}
if(guid == MF_MT_MAJOR_TYPE)
{
hr = GetGUIDNameNew(*var.puuid, &pGuidValName);
if (SUCCEEDED(hr))
{
out.MF_MT_MAJOR_TYPE = MF_MT_MAJOR_TYPE;
out.pMF_MT_MAJOR_TYPEName = pGuidValName;
pGuidValName = NULL;
}
}
if(guid == MF_MT_SUBTYPE)
{
hr = GetGUIDNameNew(*var.puuid, &pGuidValName);
if (SUCCEEDED(hr))
{
out.MF_MT_SUBTYPE = MF_MT_SUBTYPE;
out.pMF_MT_SUBTYPEName = pGuidValName;
pGuidValName = NULL;
}
}
break;
case VT_LPWSTR:
break;
case VT_VECTOR | VT_UI1:
break;
case VT_UNKNOWN:
break;
default:
break;
}
}
done:
CoTaskMemFree(pGuidName);
CoTaskMemFree(pGuidValName);
PropVariantClear(&var);
return hr;
}
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
HRESULT GetGUIDNameNew(const GUID& guid, WCHAR **ppwsz)
{
HRESULT hr = S_OK;
WCHAR *pName = NULL;
LPCWSTR pcwsz = GetGUIDNameConstNew(guid);
if (pcwsz)
{
size_t cchLength = 0;
hr = StringCchLengthW(pcwsz, STRSAFE_MAX_CCH, &cchLength);
if (FAILED(hr))
{
goto done;
}
pName = (WCHAR*)CoTaskMemAlloc((cchLength + 1) * sizeof(WCHAR));
if (pName == NULL)
{
hr = E_OUTOFMEMORY;
goto done;
}
hr = StringCchCopyW(pName, cchLength + 1, pcwsz);
if (FAILED(hr))
{
goto done;
}
}
else
{
hr = StringFromCLSID(guid, &pName);
}
done:
if (FAILED(hr))
{
*ppwsz = NULL;
CoTaskMemFree(pName);
}
else
{
*ppwsz = pName;
}
return hr;
}
GetProperty method for MSMF VideoCapture implemented.
2013-05-23 00:50:54 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
void LogUINT32AsUINT64New(const PROPVARIANT& var, UINT32 &uHigh, UINT32 &uLow)
{
Unpack2UINT32AsUINT64(var.uhVal.QuadPart, &uHigh, &uLow);
}
GetProperty method for MSMF VideoCapture implemented.
2013-05-23 00:50:54 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
float OffsetToFloatNew(const MFOffset& offset)
{
return offset.value + (static_cast<float>(offset.fract) / 65536.0f);
}
GetProperty method for MSMF VideoCapture implemented.
2013-05-23 00:50:54 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
HRESULT LogVideoAreaNew(const PROPVARIANT& var)
{
if (var.caub.cElems < sizeof(MFVideoArea))
{
return S_OK;
}
return S_OK;
}
GetProperty method for MSMF VideoCapture implemented.
2013-05-23 00:50:54 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
HRESULT SpecialCaseAttributeValueNew(GUID guid, const PROPVARIANT& var, MediaType &out)
{
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
if (guid == MF_MT_DEFAULT_STRIDE)
{
out.MF_MT_DEFAULT_STRIDE = var.intVal;
} else
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
if (guid == MF_MT_FRAME_SIZE)
{
UINT32 uHigh = 0, uLow = 0;
LogUINT32AsUINT64New(var, uHigh, uLow);
out.width = uHigh;
out.height = uLow;
out.MF_MT_FRAME_SIZE = out.width * out.height;
}
else
if (guid == MF_MT_FRAME_RATE)
{
UINT32 uHigh = 0, uLow = 0;
LogUINT32AsUINT64New(var, uHigh, uLow);
GetProperty method for MSMF VideoCapture implemented.
2013-05-23 00:50:54 +08:00
out.MF_MT_FRAME_RATE_NUMERATOR = uHigh;
out.MF_MT_FRAME_RATE_DENOMINATOR = uLow;
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
}
else
if (guid == MF_MT_FRAME_RATE_RANGE_MAX)
{
UINT32 uHigh = 0, uLow = 0;
LogUINT32AsUINT64New(var, uHigh, uLow);
out.MF_MT_FRAME_RATE_RANGE_MAX = uHigh;
out.MF_MT_FRAME_RATE_RANGE_MAX_low = uLow;
}
else
if (guid == MF_MT_FRAME_RATE_RANGE_MIN)
{
UINT32 uHigh = 0, uLow = 0;
LogUINT32AsUINT64New(var, uHigh, uLow);
out.MF_MT_FRAME_RATE_RANGE_MIN = uHigh;
out.MF_MT_FRAME_RATE_RANGE_MIN_low = uLow;
}
else
if (guid == MF_MT_PIXEL_ASPECT_RATIO)
{
UINT32 uHigh = 0, uLow = 0;
LogUINT32AsUINT64New(var, uHigh, uLow);
out.MF_MT_PIXEL_ASPECT_RATIO = uHigh;
out.MF_MT_PIXEL_ASPECT_RATIO_low = uLow;
Code review notes fixed.
2013-04-03 18:48:23 +08:00
}
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
else
{
return S_FALSE;
}
return S_OK;
}
GetProperty method for MSMF VideoCapture implemented.
2013-05-23 00:50:54 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
#ifndef IF_EQUAL_RETURN
#define IF_EQUAL_RETURN(param, val) if(val == param) return L#val
#endif
GetProperty method for MSMF VideoCapture implemented.
2013-05-23 00:50:54 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
LPCWSTR GetGUIDNameConstNew(const GUID& guid)
{
IF_EQUAL_RETURN(guid, MF_MT_MAJOR_TYPE);
IF_EQUAL_RETURN(guid, MF_MT_MAJOR_TYPE);
IF_EQUAL_RETURN(guid, MF_MT_SUBTYPE);
IF_EQUAL_RETURN(guid, MF_MT_ALL_SAMPLES_INDEPENDENT);
IF_EQUAL_RETURN(guid, MF_MT_FIXED_SIZE_SAMPLES);
IF_EQUAL_RETURN(guid, MF_MT_COMPRESSED);
IF_EQUAL_RETURN(guid, MF_MT_SAMPLE_SIZE);
IF_EQUAL_RETURN(guid, MF_MT_WRAPPED_TYPE);
IF_EQUAL_RETURN(guid, MF_MT_AUDIO_NUM_CHANNELS);
IF_EQUAL_RETURN(guid, MF_MT_AUDIO_SAMPLES_PER_SECOND);
IF_EQUAL_RETURN(guid, MF_MT_AUDIO_FLOAT_SAMPLES_PER_SECOND);
IF_EQUAL_RETURN(guid, MF_MT_AUDIO_AVG_BYTES_PER_SECOND);
IF_EQUAL_RETURN(guid, MF_MT_AUDIO_BLOCK_ALIGNMENT);
IF_EQUAL_RETURN(guid, MF_MT_AUDIO_BITS_PER_SAMPLE);
IF_EQUAL_RETURN(guid, MF_MT_AUDIO_VALID_BITS_PER_SAMPLE);
IF_EQUAL_RETURN(guid, MF_MT_AUDIO_SAMPLES_PER_BLOCK);
IF_EQUAL_RETURN(guid, MF_MT_AUDIO_CHANNEL_MASK);
IF_EQUAL_RETURN(guid, MF_MT_AUDIO_FOLDDOWN_MATRIX);
IF_EQUAL_RETURN(guid, MF_MT_AUDIO_WMADRC_PEAKREF);
IF_EQUAL_RETURN(guid, MF_MT_AUDIO_WMADRC_PEAKTARGET);
IF_EQUAL_RETURN(guid, MF_MT_AUDIO_WMADRC_AVGREF);
IF_EQUAL_RETURN(guid, MF_MT_AUDIO_WMADRC_AVGTARGET);
IF_EQUAL_RETURN(guid, MF_MT_AUDIO_PREFER_WAVEFORMATEX);
IF_EQUAL_RETURN(guid, MF_MT_AAC_PAYLOAD_TYPE);
IF_EQUAL_RETURN(guid, MF_MT_AAC_AUDIO_PROFILE_LEVEL_INDICATION);
IF_EQUAL_RETURN(guid, MF_MT_FRAME_SIZE);
IF_EQUAL_RETURN(guid, MF_MT_FRAME_RATE);
IF_EQUAL_RETURN(guid, MF_MT_FRAME_RATE_RANGE_MAX);
IF_EQUAL_RETURN(guid, MF_MT_FRAME_RATE_RANGE_MIN);
IF_EQUAL_RETURN(guid, MF_MT_PIXEL_ASPECT_RATIO);
IF_EQUAL_RETURN(guid, MF_MT_DRM_FLAGS);
IF_EQUAL_RETURN(guid, MF_MT_PAD_CONTROL_FLAGS);
IF_EQUAL_RETURN(guid, MF_MT_SOURCE_CONTENT_HINT);
IF_EQUAL_RETURN(guid, MF_MT_VIDEO_CHROMA_SITING);
IF_EQUAL_RETURN(guid, MF_MT_INTERLACE_MODE);
IF_EQUAL_RETURN(guid, MF_MT_TRANSFER_FUNCTION);
IF_EQUAL_RETURN(guid, MF_MT_VIDEO_PRIMARIES);
IF_EQUAL_RETURN(guid, MF_MT_CUSTOM_VIDEO_PRIMARIES);
IF_EQUAL_RETURN(guid, MF_MT_YUV_MATRIX);
IF_EQUAL_RETURN(guid, MF_MT_VIDEO_LIGHTING);
IF_EQUAL_RETURN(guid, MF_MT_VIDEO_NOMINAL_RANGE);
IF_EQUAL_RETURN(guid, MF_MT_GEOMETRIC_APERTURE);
IF_EQUAL_RETURN(guid, MF_MT_MINIMUM_DISPLAY_APERTURE);
IF_EQUAL_RETURN(guid, MF_MT_PAN_SCAN_APERTURE);
IF_EQUAL_RETURN(guid, MF_MT_PAN_SCAN_ENABLED);
IF_EQUAL_RETURN(guid, MF_MT_AVG_BITRATE);
IF_EQUAL_RETURN(guid, MF_MT_AVG_BIT_ERROR_RATE);
IF_EQUAL_RETURN(guid, MF_MT_MAX_KEYFRAME_SPACING);
IF_EQUAL_RETURN(guid, MF_MT_DEFAULT_STRIDE);
IF_EQUAL_RETURN(guid, MF_MT_PALETTE);
IF_EQUAL_RETURN(guid, MF_MT_USER_DATA);
IF_EQUAL_RETURN(guid, MF_MT_AM_FORMAT_TYPE);
IF_EQUAL_RETURN(guid, MF_MT_MPEG_START_TIME_CODE);
IF_EQUAL_RETURN(guid, MF_MT_MPEG2_PROFILE);
IF_EQUAL_RETURN(guid, MF_MT_MPEG2_LEVEL);
IF_EQUAL_RETURN(guid, MF_MT_MPEG2_FLAGS);
IF_EQUAL_RETURN(guid, MF_MT_MPEG_SEQUENCE_HEADER);
IF_EQUAL_RETURN(guid, MF_MT_DV_AAUX_SRC_PACK_0);
IF_EQUAL_RETURN(guid, MF_MT_DV_AAUX_CTRL_PACK_0);
IF_EQUAL_RETURN(guid, MF_MT_DV_AAUX_SRC_PACK_1);
IF_EQUAL_RETURN(guid, MF_MT_DV_AAUX_CTRL_PACK_1);
IF_EQUAL_RETURN(guid, MF_MT_DV_VAUX_SRC_PACK);
IF_EQUAL_RETURN(guid, MF_MT_DV_VAUX_CTRL_PACK);
IF_EQUAL_RETURN(guid, MF_MT_ARBITRARY_HEADER);
IF_EQUAL_RETURN(guid, MF_MT_ARBITRARY_FORMAT);
Code review notes fixed.
2013-04-03 18:48:23 +08:00
IF_EQUAL_RETURN(guid, MF_MT_IMAGE_LOSS_TOLERANT);
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
IF_EQUAL_RETURN(guid, MF_MT_MPEG4_SAMPLE_DESCRIPTION);
IF_EQUAL_RETURN(guid, MF_MT_MPEG4_CURRENT_SAMPLE_ENTRY);
Code review notes fixed.
2013-04-03 18:48:23 +08:00
IF_EQUAL_RETURN(guid, MF_MT_ORIGINAL_4CC);
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
IF_EQUAL_RETURN(guid, MF_MT_ORIGINAL_WAVE_FORMAT_TAG);
// Media types
IF_EQUAL_RETURN(guid, MFMediaType_Audio);
IF_EQUAL_RETURN(guid, MFMediaType_Video);
IF_EQUAL_RETURN(guid, MFMediaType_Protected);
IF_EQUAL_RETURN(guid, MFMediaType_SAMI);
IF_EQUAL_RETURN(guid, MFMediaType_Script);
IF_EQUAL_RETURN(guid, MFMediaType_Image);
IF_EQUAL_RETURN(guid, MFMediaType_HTML);
IF_EQUAL_RETURN(guid, MFMediaType_Binary);
IF_EQUAL_RETURN(guid, MFMediaType_FileTransfer);
IF_EQUAL_RETURN(guid, MFVideoFormat_AI44); // FCC('AI44')
Code review notes fixed.
2013-04-03 18:48:23 +08:00
IF_EQUAL_RETURN(guid, MFVideoFormat_ARGB32); // D3DFMT_A8R8G8B8
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
IF_EQUAL_RETURN(guid, MFVideoFormat_AYUV); // FCC('AYUV')
IF_EQUAL_RETURN(guid, MFVideoFormat_DV25); // FCC('dv25')
IF_EQUAL_RETURN(guid, MFVideoFormat_DV50); // FCC('dv50')
IF_EQUAL_RETURN(guid, MFVideoFormat_DVH1); // FCC('dvh1')
IF_EQUAL_RETURN(guid, MFVideoFormat_DVSD); // FCC('dvsd')
IF_EQUAL_RETURN(guid, MFVideoFormat_DVSL); // FCC('dvsl')
IF_EQUAL_RETURN(guid, MFVideoFormat_H264); // FCC('H264')
IF_EQUAL_RETURN(guid, MFVideoFormat_I420); // FCC('I420')
IF_EQUAL_RETURN(guid, MFVideoFormat_IYUV); // FCC('IYUV')
IF_EQUAL_RETURN(guid, MFVideoFormat_M4S2); // FCC('M4S2')
IF_EQUAL_RETURN(guid, MFVideoFormat_MJPG);
IF_EQUAL_RETURN(guid, MFVideoFormat_MP43); // FCC('MP43')
IF_EQUAL_RETURN(guid, MFVideoFormat_MP4S); // FCC('MP4S')
IF_EQUAL_RETURN(guid, MFVideoFormat_MP4V); // FCC('MP4V')
IF_EQUAL_RETURN(guid, MFVideoFormat_MPG1); // FCC('MPG1')
IF_EQUAL_RETURN(guid, MFVideoFormat_MSS1); // FCC('MSS1')
IF_EQUAL_RETURN(guid, MFVideoFormat_MSS2); // FCC('MSS2')
IF_EQUAL_RETURN(guid, MFVideoFormat_NV11); // FCC('NV11')
IF_EQUAL_RETURN(guid, MFVideoFormat_NV12); // FCC('NV12')
IF_EQUAL_RETURN(guid, MFVideoFormat_P010); // FCC('P010')
IF_EQUAL_RETURN(guid, MFVideoFormat_P016); // FCC('P016')
IF_EQUAL_RETURN(guid, MFVideoFormat_P210); // FCC('P210')
IF_EQUAL_RETURN(guid, MFVideoFormat_P216); // FCC('P216')
Code review notes fixed.
2013-04-03 18:48:23 +08:00
IF_EQUAL_RETURN(guid, MFVideoFormat_RGB24); // D3DFMT_R8G8B8
IF_EQUAL_RETURN(guid, MFVideoFormat_RGB32); // D3DFMT_X8R8G8B8
IF_EQUAL_RETURN(guid, MFVideoFormat_RGB555); // D3DFMT_X1R5G5B5
IF_EQUAL_RETURN(guid, MFVideoFormat_RGB565); // D3DFMT_R5G6B5
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
IF_EQUAL_RETURN(guid, MFVideoFormat_RGB8);
IF_EQUAL_RETURN(guid, MFVideoFormat_UYVY); // FCC('UYVY')
IF_EQUAL_RETURN(guid, MFVideoFormat_v210); // FCC('v210')
IF_EQUAL_RETURN(guid, MFVideoFormat_v410); // FCC('v410')
IF_EQUAL_RETURN(guid, MFVideoFormat_WMV1); // FCC('WMV1')
IF_EQUAL_RETURN(guid, MFVideoFormat_WMV2); // FCC('WMV2')
IF_EQUAL_RETURN(guid, MFVideoFormat_WMV3); // FCC('WMV3')
IF_EQUAL_RETURN(guid, MFVideoFormat_WVC1); // FCC('WVC1')
IF_EQUAL_RETURN(guid, MFVideoFormat_Y210); // FCC('Y210')
IF_EQUAL_RETURN(guid, MFVideoFormat_Y216); // FCC('Y216')
IF_EQUAL_RETURN(guid, MFVideoFormat_Y410); // FCC('Y410')
IF_EQUAL_RETURN(guid, MFVideoFormat_Y416); // FCC('Y416')
IF_EQUAL_RETURN(guid, MFVideoFormat_Y41P);
IF_EQUAL_RETURN(guid, MFVideoFormat_Y41T);
IF_EQUAL_RETURN(guid, MFVideoFormat_YUY2); // FCC('YUY2')
IF_EQUAL_RETURN(guid, MFVideoFormat_YV12); // FCC('YV12')
IF_EQUAL_RETURN(guid, MFVideoFormat_YVYU);
Code review notes fixed.
2013-04-03 18:48:23 +08:00
IF_EQUAL_RETURN(guid, MFAudioFormat_PCM); // WAVE_FORMAT_PCM
IF_EQUAL_RETURN(guid, MFAudioFormat_Float); // WAVE_FORMAT_IEEE_FLOAT
IF_EQUAL_RETURN(guid, MFAudioFormat_DTS); // WAVE_FORMAT_DTS
IF_EQUAL_RETURN(guid, MFAudioFormat_Dolby_AC3_SPDIF); // WAVE_FORMAT_DOLBY_AC3_SPDIF
IF_EQUAL_RETURN(guid, MFAudioFormat_DRM); // WAVE_FORMAT_DRM
IF_EQUAL_RETURN(guid, MFAudioFormat_WMAudioV8); // WAVE_FORMAT_WMAUDIO2
IF_EQUAL_RETURN(guid, MFAudioFormat_WMAudioV9); // WAVE_FORMAT_WMAUDIO3
IF_EQUAL_RETURN(guid, MFAudioFormat_WMAudio_Lossless); // WAVE_FORMAT_WMAUDIO_LOSSLESS
IF_EQUAL_RETURN(guid, MFAudioFormat_WMASPDIF); // WAVE_FORMAT_WMASPDIF
IF_EQUAL_RETURN(guid, MFAudioFormat_MSP1); // WAVE_FORMAT_WMAVOICE9
IF_EQUAL_RETURN(guid, MFAudioFormat_MP3); // WAVE_FORMAT_MPEGLAYER3
IF_EQUAL_RETURN(guid, MFAudioFormat_MPEG); // WAVE_FORMAT_MPEG
IF_EQUAL_RETURN(guid, MFAudioFormat_AAC); // WAVE_FORMAT_MPEG_HEAAC
IF_EQUAL_RETURN(guid, MFAudioFormat_ADTS); // WAVE_FORMAT_MPEG_ADTS_AAC
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
return NULL;
}
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
FormatReader::FormatReader(void)
{
}
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
MediaType FormatReader::Read(IMFMediaType *pType)
{
UINT32 count = 0;
HRESULT hr = S_OK;
MediaType out;
hr = pType->LockStore();
if (FAILED(hr))
{
return out;
}
hr = pType->GetCount(&count);
if (FAILED(hr))
{
return out;
}
for (UINT32 i = 0; i < count; i++)
{
hr = LogAttributeValueByIndexNew(pType, i, out);
if (FAILED(hr))
{
break;
}
}
hr = pType->UnlockStore();
if (FAILED(hr))
{
return out;
}
return out;
}
GetProperty method for MSMF VideoCapture implemented.
2013-05-23 00:50:54 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
FormatReader::~FormatReader(void)
{
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Windows RT related code cleanup. Windows RT sample cleanup. Video I/O code cleanup.
2013-06-24 16:10:18 +08:00
#define CHECK_HR(x) if (FAILED(x)) { goto done; }
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
ImageGrabberCallback::ImageGrabberCallback(bool synchronous):
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
m_cRef(1),
ig_RIE(true),
ig_Close(false),
Media Foundation camera capture fixed. Camera-based VideoCapture updated to fit changes in ImageGrabber from prev commit
2013-05-15 20:12:25 +08:00
ig_Synchronous(synchronous),
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
ig_hFrameReady(synchronous ? CreateEvent(NULL, FALSE, FALSE, NULL): 0),
ig_hFrameGrabbed(synchronous ? CreateEvent(NULL, FALSE, TRUE, NULL): 0),
ig_hFinish(CreateEvent(NULL, TRUE, FALSE, NULL))
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
{}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
ImageGrabber::ImageGrabber(unsigned int deviceID, bool synchronous):
ImageGrabberCallback(synchronous),
ig_DeviceID(deviceID),
ig_pSource(NULL),
ig_pSession(NULL),
ig_pTopology(NULL)
{}
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
ImageGrabber::~ImageGrabber(void)
{
if (ig_pSession)
{
ig_pSession->Shutdown();
}
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
CloseHandle(ig_hFinish);
Media Foundation camera capture fixed. Camera-based VideoCapture updated to fit changes in ImageGrabber from prev commit
2013-05-15 20:12:25 +08:00
if (ig_Synchronous)
{
CloseHandle(ig_hFrameReady);
CloseHandle(ig_hFrameGrabbed);
}
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
SafeRelease(&ig_pSession);
SafeRelease(&ig_pTopology);
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
DPO->printOut(L"IMAGEGRABBER VIDEODEVICE %i: Destroying instance of the ImageGrabber class\n", ig_DeviceID);
}
#ifdef HAVE_WINRT
ImageGrabberWinRT::ImageGrabberWinRT(bool synchronous):
ImageGrabberCallback(synchronous),
ig_pMediaSink(NULL)
{
ig_pMedCapSource = nullptr;
}
ImageGrabberWinRT::~ImageGrabberWinRT(void)
{
//stop must already be performed and complete by object owner
if (ig_pMediaSink != NULL) {
((IMFMediaSink*)ig_pMediaSink)->Shutdown();
}
SafeRelease(&ig_pMediaSink);
RELEASE_AGILE_WRL(ig_pMedCapSource)
CloseHandle(ig_hFinish);
if (ig_Synchronous)
{
CloseHandle(ig_hFrameReady);
CloseHandle(ig_hFrameGrabbed);
}
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
DPO->printOut(L"IMAGEGRABBER VIDEODEVICE: Destroying instance of the ImageGrabberWinRT class\n");
}
HRESULT ImageGrabberWinRT::initImageGrabber(MAKE_WRL_REF(_MediaCapture) pSource,
GUID VideoFormat)
{
HRESULT hr;
MAKE_WRL_OBJ(_VideoDeviceController) pDevCont;
WRL_PROP_GET(pSource, VideoDeviceController, pDevCont, hr)
if (FAILED(hr)) return hr;
GET_WRL_MEDIA_DEVICE_CONTROLLER(pDevCont, pMedDevCont, hr)
if (FAILED(hr)) return hr;
MAKE_WRL_OBJ(_MediaEncodingProperties) pMedEncProps;
WRL_METHOD(pMedDevCont, GetMediaStreamProperties, pMedEncProps, hr, _VideoPreview)
if (FAILED(hr)) return hr;
GET_WRL_VIDEO_ENCODING_PROPERTIES(pMedEncProps, pVidProps, hr);
if (FAILED(hr)) return hr;
ComPtr<IMFMediaType> pType = NULL;
hr = MediaSink::ConvertPropertiesToMediaType(DEREF_AS_NATIVE_WRL_OBJ(ABI::Windows::Media::MediaProperties::IMediaEncodingProperties, pMedEncProps), &pType);
if (FAILED(hr)) return hr;
MediaType MT = FormatReader::Read(pType.Get());
unsigned int sizeRawImage = 0;
if(VideoFormat == MFVideoFormat_RGB24)
{
sizeRawImage = MT.MF_MT_FRAME_SIZE * 3;
}
else if(VideoFormat == MFVideoFormat_RGB32)
{
sizeRawImage = MT.MF_MT_FRAME_SIZE * 4;
}
sizeRawImage = MT.MF_MT_SAMPLE_SIZE;
CHECK_HR(hr = RawImage::CreateInstance(&ig_RIFirst, sizeRawImage));
CHECK_HR(hr = RawImage::CreateInstance(&ig_RISecond, sizeRawImage));
ig_RIOut = ig_RISecond;
ig_pMedCapSource = pSource;
done:
return hr;
}
HRESULT ImageGrabberWinRT::stopGrabbing(MAKE_WRL_REF(_AsyncAction)* action)
{
HRESULT hr = S_OK;
if (ig_pMedCapSource != nullptr) {
GET_WRL_VIDEO_PREVIEW(DEREF_AGILE_WRL_OBJ(ig_pMedCapSource), imedPrevCap, hr);
if (FAILED(hr)) return hr;
MAKE_WRL_REF(_AsyncAction) pAction;
WRL_METHOD_BASE(imedPrevCap, StopPreviewAsync, pAction, hr)
if (SUCCEEDED(hr)) {
SAVE_CURRENT_CONTEXT(context);
*action = reinterpret_cast<MAKE_WRL_REF(_AsyncAction)>(BEGIN_CREATE_ASYNC(pAction, context, this)
HRESULT hr;
DO_ACTION_SYNCHRONOUSLY(hr, pAction, context);
SafeRelease(&ig_pMediaSink);
SetEvent(ig_hFinish);
END_CREATE_ASYNC(hr));
}
}
return hr;
}
HRESULT ImageGrabberWinRT::startGrabbing(MAKE_WRL_REF(_AsyncAction)* action)
{
HRESULT hr = S_OK;
GET_WRL_VIDEO_PREVIEW(DEREF_AGILE_WRL_OBJ(ig_pMedCapSource), imedPrevCap, hr);
if (FAILED(hr)) return hr;
ACTIVATE_OBJ(RuntimeClass_Windows_Foundation_Collections_PropertySet, MAKE_WRL_OBJ(_PropertySet), pSet, hr)
if (FAILED(hr)) return hr;
GET_WRL_MAP(pSet, spSetting, hr)
if (FAILED(hr)) return hr;
ACTIVATE_STATIC_OBJ(RuntimeClass_Windows_Foundation_PropertyValue, MAKE_WRL_OBJ(_PropertyValueStatics), spPropVal, hr)
if (FAILED(hr)) return hr;
_ObjectObj pVal;
boolean bReplaced;
WRL_METHOD(spPropVal, CreateUInt32, pVal, hr, (unsigned int)_VideoPreview)
if (FAILED(hr)) return hr;
WRL_METHOD(spSetting, Insert, bReplaced, hr, DEREF_WRL_OBJ(_StringReference(MF_PROP_VIDTYPE)), DEREF_WRL_OBJ(pVal))
if (FAILED(hr)) return hr;
WRL_METHOD(spSetting, Insert, bReplaced, hr, DEREF_WRL_OBJ(_StringReference(MF_PROP_SAMPLEGRABBERCALLBACK)), reinterpret_cast<_Object>(this))
if (FAILED(hr)) return hr;
MAKE_WRL_OBJ(_VideoDeviceController) pDevCont;
WRL_PROP_GET(ig_pMedCapSource, VideoDeviceController, pDevCont, hr)
if (FAILED(hr)) return hr;
GET_WRL_MEDIA_DEVICE_CONTROLLER(pDevCont, pMedDevCont, hr)
if (FAILED(hr)) return hr;
MAKE_WRL_OBJ(_MediaEncodingProperties) pMedEncProps;
WRL_METHOD(pMedDevCont, GetMediaStreamProperties, pMedEncProps, hr, _VideoPreview)
if (FAILED(hr)) return hr;
GET_WRL_VIDEO_ENCODING_PROPERTIES(pMedEncProps, pVidProps, hr);
if (FAILED(hr)) return hr;
ACTIVATE_OBJ(RuntimeClass_Windows_Media_MediaProperties_MediaEncodingProfile, MAKE_WRL_OBJ(_MediaEncodingProfile), pEncProps, hr)
if (FAILED(hr)) return hr;
WRL_PROP_PUT(pEncProps, Video, DEREF_WRL_OBJ(pVidProps), hr)
if (FAILED(hr)) return hr;
WRL_METHOD(spSetting, Insert, bReplaced, hr, DEREF_WRL_OBJ(_StringReference(MF_PROP_VIDENCPROPS)), DEREF_WRL_OBJ(pVidProps))
if (SUCCEEDED(hr)) {
//can start/stop multiple times with same MediaCapture object if using activatable class
WRL_METHOD(imedPrevCap, _StartPreviewToCustomSinkIdAsync, *action, hr, DEREF_WRL_OBJ(pEncProps), DEREF_WRL_OBJ(_StringReference(RuntimeClass_CV_MediaSink)), DEREF_WRL_OBJ(pSet))
if (FAILED(hr) && hr == REGDB_E_CLASSNOTREG) {
hr = Microsoft::WRL::Make<MediaSink>().CopyTo(&ig_pMediaSink);
if (FAILED(hr)) return hr;
hr = ((ABI::Windows::Media::IMediaExtension*)ig_pMediaSink)->SetProperties(DEREF_AS_NATIVE_WRL_OBJ(ABI::Windows::Foundation::Collections::IPropertySet, pSet));
if (FAILED(hr)) return hr;
WRL_METHOD(imedPrevCap, StartPreviewToCustomSinkAsync, *action, hr, DEREF_WRL_OBJ(pEncProps), reinterpret_cast<MAKE_WRL_REF(_MediaExtension)>(ig_pMediaSink))
}
}
return hr;
}
HRESULT ImageGrabberWinRT::CreateInstance(ImageGrabberWinRT **ppIG, bool synchronous)
{
*ppIG = Microsoft::WRL::Make<ImageGrabberWinRT>(synchronous).Detach();
if (ppIG == NULL)
{
return E_OUTOFMEMORY;
}
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
DPO->printOut(L"IMAGEGRABBER VIDEODEVICE: Creating instance of ImageGrabberWinRT\n");
return S_OK;
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
}
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#endif
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
HRESULT ImageGrabber::initImageGrabber(IMFMediaSource *pSource, GUID VideoFormat)
{
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
ComPtr<IMFActivate> pSinkActivate = NULL;
ComPtr<IMFMediaType> pType = NULL;
ComPtr<IMFPresentationDescriptor> pPD = NULL;
ComPtr<IMFStreamDescriptor> pSD = NULL;
ComPtr<IMFMediaTypeHandler> pHandler = NULL;
ComPtr<IMFMediaType> pCurrentType = NULL;
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
HRESULT hr = S_OK;
MediaType MT;
// Clean up.
if (ig_pSession)
{
ig_pSession->Shutdown();
}
SafeRelease(&ig_pSession);
SafeRelease(&ig_pTopology);
ig_pSource = pSource;
hr = pSource->CreatePresentationDescriptor(&pPD);
if (FAILED(hr))
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
{
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
goto err;
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
}
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
BOOL fSelected;
hr = pPD->GetStreamDescriptorByIndex(0, &fSelected, &pSD);
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
if (FAILED(hr)) {
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
goto err;
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
}
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
hr = pSD->GetMediaTypeHandler(&pHandler);
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
if (FAILED(hr)) {
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
goto err;
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
}
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
DWORD cTypes = 0;
hr = pHandler->GetMediaTypeCount(&cTypes);
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
if (FAILED(hr)) {
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
goto err;
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
}
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
if(cTypes > 0)
{
hr = pHandler->GetCurrentMediaType(&pCurrentType);
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
if (FAILED(hr)) {
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
goto err;
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
}
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
MT = FormatReader::Read(pCurrentType.Get());
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
}
err:
unsigned int sizeRawImage = 0;
if(VideoFormat == MFVideoFormat_RGB24)
{
sizeRawImage = MT.MF_MT_FRAME_SIZE * 3;
}
else if(VideoFormat == MFVideoFormat_RGB32)
{
sizeRawImage = MT.MF_MT_FRAME_SIZE * 4;
}
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
//sizeRawImage = MT.MF_MT_SAMPLE_SIZE;
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
CHECK_HR(hr = RawImage::CreateInstance(&ig_RIFirst, sizeRawImage));
CHECK_HR(hr = RawImage::CreateInstance(&ig_RISecond, sizeRawImage));
ig_RIOut = ig_RISecond;
// Configure the media type that the Sample Grabber will receive.
// Setting the major and subtype is usually enough for the topology loader
// to resolve the topology.
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
CHECK_HR(hr = MFCreateMediaType(pType.GetAddressOf()));
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
CHECK_HR(hr = pType->SetGUID(MF_MT_MAJOR_TYPE, MFMediaType_Video));
CHECK_HR(hr = pType->SetGUID(MF_MT_SUBTYPE, VideoFormat));
// Create the sample grabber sink.
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
CHECK_HR(hr = MFCreateSampleGrabberSinkActivate(pType.Get(), this, pSinkActivate.GetAddressOf()));
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
// To run as fast as possible, set this attribute (requires Windows 7):
CHECK_HR(hr = pSinkActivate->SetUINT32(MF_SAMPLEGRABBERSINK_IGNORE_CLOCK, TRUE));
// Create the Media Session.
CHECK_HR(hr = MFCreateMediaSession(NULL, &ig_pSession));
// Create the topology.
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
CHECK_HR(hr = CreateTopology(pSource, pSinkActivate.Get(), &ig_pTopology));
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
done:
// Clean up.
if (FAILED(hr))
Code review notes fixed.
2013-04-03 18:48:23 +08:00
{
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
if (ig_pSession)
{
ig_pSession->Shutdown();
}
SafeRelease(&ig_pSession);
SafeRelease(&ig_pTopology);
}
return hr;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
void ImageGrabber::stopGrabbing()
{
if(ig_pSession)
ig_pSession->Stop();
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
DPO->printOut(L"IMAGEGRABBER VIDEODEVICE %i: Stopping of of grabbing of images\n", ig_DeviceID);
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
HRESULT ImageGrabber::startGrabbing(void)
{
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
HRESULT hr = S_OK;
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
ComPtr<IMFMediaEvent> pEvent = NULL;
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
PROPVARIANT var;
PropVariantInit(&var);
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
hr = ig_pSession->SetTopology(0, ig_pTopology);
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
DPO->printOut(L"IMAGEGRABBER VIDEODEVICE %i: Start Grabbing of the images\n", ig_DeviceID);
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
hr = ig_pSession->Start(&GUID_NULL, &var);
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
for(;;)
{
HRESULT hrStatus = S_OK;
MediaEventType met;
if(!ig_pSession) break;
hr = ig_pSession->GetEvent(0, &pEvent);
if(!SUCCEEDED(hr))
{
hr = S_OK;
goto done;
}
hr = pEvent->GetStatus(&hrStatus);
if(!SUCCEEDED(hr))
{
hr = S_OK;
goto done;
}
hr = pEvent->GetType(&met);
if(!SUCCEEDED(hr))
{
hr = S_OK;
goto done;
}
if (met == MESessionEnded)
Code review notes fixed.
2013-04-03 18:48:23 +08:00
{
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
DPO->printOut(L"IMAGEGRABBER VIDEODEVICE %i: MESessionEnded \n", ig_DeviceID);
ig_pSession->Stop();
break;
}
if (met == MESessionStopped)
{
DPO->printOut(L"IMAGEGRABBER VIDEODEVICE %i: MESessionStopped \n", ig_DeviceID);
break;
}
if (met == MEVideoCaptureDeviceRemoved)
{
DPO->printOut(L"IMAGEGRABBER VIDEODEVICE %i: MEVideoCaptureDeviceRemoved \n", ig_DeviceID);
Code review notes fixed.
2013-04-03 18:48:23 +08:00
break;
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
}
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
if ((met == MEError) || (met == MENonFatalError))
{
pEvent->GetStatus(&hrStatus);
DPO->printOut(L"IMAGEGRABBER VIDEODEVICE %i: MEError | MENonFatalError: %u\n", ig_DeviceID, hrStatus);
break;
}
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
}
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
DPO->printOut(L"IMAGEGRABBER VIDEODEVICE %i: Finish startGrabbing \n", ig_DeviceID);
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
done:
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
SetEvent(ig_hFinish);
Media Foundation-based VideoWriter added
2013-05-06 18:36:51 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
return hr;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
void ImageGrabberCallback::pauseGrabbing()
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
{
}
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
void ImageGrabberCallback::resumeGrabbing()
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
{
}
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
HRESULT ImageGrabber::CreateTopology(IMFMediaSource *pSource, IMFActivate *pSinkActivate, IMFTopology **ppTopo)
{
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
IMFTopology* pTopology = NULL;
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
ComPtr<IMFPresentationDescriptor> pPD = NULL;
ComPtr<IMFStreamDescriptor> pSD = NULL;
ComPtr<IMFMediaTypeHandler> pHandler = NULL;
ComPtr<IMFTopologyNode> pNode1 = NULL;
ComPtr<IMFTopologyNode> pNode2 = NULL;
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
HRESULT hr = S_OK;
DWORD cStreams = 0;
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
CHECK_HR(hr = MFCreateTopology(&pTopology));
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
CHECK_HR(hr = pSource->CreatePresentationDescriptor(pPD.GetAddressOf()));
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
CHECK_HR(hr = pPD->GetStreamDescriptorCount(&cStreams));
for (DWORD i = 0; i < cStreams; i++)
{
// In this example, we look for audio streams and connect them to the sink.
BOOL fSelected = FALSE;
GUID majorType;
CHECK_HR(hr = pPD->GetStreamDescriptorByIndex(i, &fSelected, &pSD));
CHECK_HR(hr = pSD->GetMediaTypeHandler(&pHandler));
CHECK_HR(hr = pHandler->GetMajorType(&majorType));
if (majorType == MFMediaType_Video && fSelected)
{
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
CHECK_HR(hr = AddSourceNode(pTopology, pSource, pPD.Get(), pSD.Get(), pNode1.GetAddressOf()));
CHECK_HR(hr = AddOutputNode(pTopology, pSinkActivate, 0, pNode2.GetAddressOf()));
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
CHECK_HR(hr = pNode1->ConnectOutput(0, pNode2.Get(), 0));
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
break;
}
else
{
CHECK_HR(hr = pPD->DeselectStream(i));
}
}
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
*ppTopo = pTopology;
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
(*ppTopo)->AddRef();
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
done:
return hr;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
HRESULT ImageGrabber::AddSourceNode(
IMFTopology *pTopology, // Topology.
IMFMediaSource *pSource, // Media source.
IMFPresentationDescriptor *pPD, // Presentation descriptor.
IMFStreamDescriptor *pSD, // Stream descriptor.
IMFTopologyNode **ppNode) // Receives the node pointer.
{
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
ComPtr<IMFTopologyNode> pNode = NULL;
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
HRESULT hr = S_OK;
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
CHECK_HR(hr = MFCreateTopologyNode(MF_TOPOLOGY_SOURCESTREAM_NODE, pNode.GetAddressOf()));
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
CHECK_HR(hr = pNode->SetUnknown(MF_TOPONODE_SOURCE, pSource));
CHECK_HR(hr = pNode->SetUnknown(MF_TOPONODE_PRESENTATION_DESCRIPTOR, pPD));
CHECK_HR(hr = pNode->SetUnknown(MF_TOPONODE_STREAM_DESCRIPTOR, pSD));
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
CHECK_HR(hr = pTopology->AddNode(pNode.Get()));
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
// Return the pointer to the caller.
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
*ppNode = pNode.Get();
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
(*ppNode)->AddRef();
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
done:
return hr;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
HRESULT ImageGrabber::AddOutputNode(
IMFTopology *pTopology, // Topology.
IMFActivate *pActivate, // Media sink activation object.
DWORD dwId, // Identifier of the stream sink.
IMFTopologyNode **ppNode) // Receives the node pointer.
{
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
ComPtr<IMFTopologyNode> pNode = NULL;
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
HRESULT hr = S_OK;
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
CHECK_HR(hr = MFCreateTopologyNode(MF_TOPOLOGY_OUTPUT_NODE, pNode.GetAddressOf()));
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
CHECK_HR(hr = pNode->SetObject(pActivate));
CHECK_HR(hr = pNode->SetUINT32(MF_TOPONODE_STREAMID, dwId));
CHECK_HR(hr = pNode->SetUINT32(MF_TOPONODE_NOSHUTDOWN_ON_REMOVE, FALSE));
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
CHECK_HR(hr = pTopology->AddNode(pNode.Get()));
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
// Return the pointer to the caller.
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
*ppNode = pNode.Get();
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
(*ppNode)->AddRef();
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
done:
return hr;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Media Foundation camera capture fixed. Camera-based VideoCapture updated to fit changes in ImageGrabber from prev commit
2013-05-15 20:12:25 +08:00
HRESULT ImageGrabber::CreateInstance(ImageGrabber **ppIG, unsigned int deviceID, bool synchronious)
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
{
Media Foundation camera capture fixed. Camera-based VideoCapture updated to fit changes in ImageGrabber from prev commit
2013-05-15 20:12:25 +08:00
*ppIG = new (std::nothrow) ImageGrabber(deviceID, synchronious);
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
if (ppIG == NULL)
{
return E_OUTOFMEMORY;
}
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
DPO->printOut(L"IMAGEGRABBER VIDEODEVICE %i: Creating instance of ImageGrabber\n", deviceID);
return S_OK;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
STDMETHODIMP ImageGrabber::QueryInterface(REFIID riid, void** ppv)
{
HRESULT hr = E_NOINTERFACE;
*ppv = NULL;
if(riid == IID_IUnknown || riid == IID_IMFSampleGrabberSinkCallback)
{
*ppv = static_cast<IMFSampleGrabberSinkCallback *>(this);
hr = S_OK;
}
if(riid == IID_IMFClockStateSink)
{
*ppv = static_cast<IMFClockStateSink *>(this);
hr = S_OK;
}
if(SUCCEEDED(hr))
{
reinterpret_cast<IUnknown *>(*ppv)->AddRef();
}
return hr;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
STDMETHODIMP_(ULONG) ImageGrabber::AddRef()
{
return InterlockedIncrement(&m_cRef);
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
STDMETHODIMP_(ULONG) ImageGrabber::Release()
{
ULONG cRef = InterlockedDecrement(&m_cRef);
if (cRef == 0)
{
delete this;
}
return cRef;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
STDMETHODIMP ImageGrabberCallback::OnClockStart(MFTIME hnsSystemTime, LONGLONG llClockStartOffset)
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
{
(void)hnsSystemTime;
(void)llClockStartOffset;
return S_OK;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
STDMETHODIMP ImageGrabberCallback::OnClockStop(MFTIME hnsSystemTime)
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
{
(void)hnsSystemTime;
return S_OK;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
STDMETHODIMP ImageGrabberCallback::OnClockPause(MFTIME hnsSystemTime)
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
{
(void)hnsSystemTime;
return S_OK;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
STDMETHODIMP ImageGrabberCallback::OnClockRestart(MFTIME hnsSystemTime)
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
{
(void)hnsSystemTime;
return S_OK;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
STDMETHODIMP ImageGrabberCallback::OnClockSetRate(MFTIME hnsSystemTime, float flRate)
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
{
(void)flRate;
(void)hnsSystemTime;
return S_OK;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
STDMETHODIMP ImageGrabberCallback::OnSetPresentationClock(IMFPresentationClock* pClock)
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
{
(void)pClock;
return S_OK;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
STDMETHODIMP ImageGrabberCallback::OnProcessSample(REFGUID guidMajorMediaType, DWORD dwSampleFlags,
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
LONGLONG llSampleTime, LONGLONG llSampleDuration, const BYTE * pSampleBuffer,
DWORD dwSampleSize)
{
(void)guidMajorMediaType;
(void)llSampleTime;
(void)dwSampleFlags;
(void)llSampleDuration;
(void)dwSampleSize;
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
HANDLE tmp[] = {ig_hFinish, ig_hFrameGrabbed, NULL};
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
DWORD status = WaitForMultipleObjects(2, tmp, FALSE, INFINITE);
if (status == WAIT_OBJECT_0)
{
printf("OnProcessFrame called after ig_hFinish event\n");
return S_OK;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
if(ig_RIE)
{
ig_RIFirst->fastCopy(pSampleBuffer);
ig_RIOut = ig_RIFirst;
}
else
{
ig_RISecond->fastCopy(pSampleBuffer);
ig_RIOut = ig_RISecond;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Media Foundation camera capture fixed. Camera-based VideoCapture updated to fit changes in ImageGrabber from prev commit
2013-05-15 20:12:25 +08:00
if (ig_Synchronous)
{
SetEvent(ig_hFrameReady);
}
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
else
{
ig_RIE = !ig_RIE;
}
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
return S_OK;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
STDMETHODIMP ImageGrabberCallback::OnShutdown()
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
{
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
SetEvent(ig_hFinish);
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
return S_OK;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
RawImage *ImageGrabberCallback::getRawImage()
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
{
return ig_RIOut;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
DWORD WINAPI MainThreadFunction( LPVOID lpParam )
{
ImageGrabberThread *pIGT = (ImageGrabberThread *)lpParam;
pIGT->run();
Code review notes fixed.
2013-04-03 18:48:23 +08:00
return 0;
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Media Foundation camera capture fixed. Camera-based VideoCapture updated to fit changes in ImageGrabber from prev commit
2013-05-15 20:12:25 +08:00
HRESULT ImageGrabberThread::CreateInstance(ImageGrabberThread **ppIGT, IMFMediaSource *pSource, unsigned int deviceID, bool synchronious)
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
{
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
Media Foundation camera capture fixed. Camera-based VideoCapture updated to fit changes in ImageGrabber from prev commit
2013-05-15 20:12:25 +08:00
*ppIGT = new (std::nothrow) ImageGrabberThread(pSource, deviceID, synchronious);
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
if (ppIGT == NULL)
{
DPO->printOut(L"IMAGEGRABBERTHREAD VIDEODEVICE %i: Memory cannot be allocated\n", deviceID);
return E_OUTOFMEMORY;
}
else
DPO->printOut(L"IMAGEGRABBERTHREAD VIDEODEVICE %i: Creating of the instance of ImageGrabberThread\n", deviceID);
return S_OK;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
ImageGrabberThread::ImageGrabberThread(IMFMediaSource *pSource, unsigned int deviceID, bool synchronious) :
Media Foundation camera capture fixed. Camera-based VideoCapture updated to fit changes in ImageGrabber from prev commit
2013-05-15 20:12:25 +08:00
igt_func(NULL),
igt_Handle(NULL),
igt_stop(false)
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
{
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
Media Foundation camera capture fixed. Camera-based VideoCapture updated to fit changes in ImageGrabber from prev commit
2013-05-15 20:12:25 +08:00
HRESULT hr = ImageGrabber::CreateInstance(&igt_pImageGrabber, deviceID, synchronious);
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
igt_DeviceID = deviceID;
if(SUCCEEDED(hr))
{
hr = igt_pImageGrabber->initImageGrabber(pSource, MFVideoFormat_RGB24);
if(!SUCCEEDED(hr))
{
DPO->printOut(L"IMAGEGRABBERTHREAD VIDEODEVICE %i: There is a problem with initialization of the instance of the ImageGrabber class\n", deviceID);
}
else
{
DPO->printOut(L"IMAGEGRABBERTHREAD VIDEODEVICE %i: Initialization of instance of the ImageGrabber class\n", deviceID);
}
}
else
{
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
DPO->printOut(L"IMAGEGRABBERTHREAD VIDEODEVICE %i: There is a problem with creation of the instance of the ImageGrabber class\n", deviceID);
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
}
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
void ImageGrabberThread::setEmergencyStopEvent(void *userData, void(*func)(int, void *))
{
if(func)
{
igt_func = func;
igt_userData = userData;
}
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
ImageGrabberThread::~ImageGrabberThread(void)
{
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
DPO->printOut(L"IMAGEGRABBERTHREAD VIDEODEVICE %i: Destroing ImageGrabberThread\n", igt_DeviceID);
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
if (igt_Handle)
WaitForSingleObject(igt_Handle, INFINITE);
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
delete igt_pImageGrabber;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
void ImageGrabberThread::stop()
{
igt_stop = true;
if(igt_pImageGrabber)
{
igt_pImageGrabber->stopGrabbing();
}
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
void ImageGrabberThread::start()
{
Code review notes fixed.
2013-04-03 18:48:23 +08:00
igt_Handle = CreateThread(
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
NULL, // default security attributes
0, // use default stack size
MainThreadFunction, // thread function name
this, // argument to thread function
0, // use default creation flags
Code review notes fixed.
2013-04-03 18:48:23 +08:00
&igt_ThreadIdArray); // returns the thread identifier
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
void ImageGrabberThread::run()
{
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
if(igt_pImageGrabber)
{
DPO->printOut(L"IMAGEGRABBERTHREAD VIDEODEVICE %i: Thread for grabbing images is started\n", igt_DeviceID);
HRESULT hr = igt_pImageGrabber->startGrabbing();
Code review notes fixed.
2013-04-03 18:48:23 +08:00
if(!SUCCEEDED(hr))
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
{
DPO->printOut(L"IMAGEGRABBERTHREAD VIDEODEVICE %i: There is a problem with starting the process of grabbing\n", igt_DeviceID);
}
}
else
{
DPO->printOut(L"IMAGEGRABBERTHREAD VIDEODEVICE %i The thread is finished without execution of grabbing\n", igt_DeviceID);
}
if(!igt_stop)
{
DPO->printOut(L"IMAGEGRABBERTHREAD VIDEODEVICE %i: Emergency Stop thread\n", igt_DeviceID);
if(igt_func)
{
igt_func(igt_DeviceID, igt_userData);
}
}
else
DPO->printOut(L"IMAGEGRABBERTHREAD VIDEODEVICE %i: Finish thread\n", igt_DeviceID);
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
ImageGrabber *ImageGrabberThread::getImageGrabber()
{
return igt_pImageGrabber;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
Media_Foundation::Media_Foundation(void)
{
HRESULT hr = MFStartup(MF_VERSION);
if(!SUCCEEDED(hr))
{
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
DPO->printOut(L"MEDIA FOUNDATION: It cannot be created!!!\n");
}
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
Media_Foundation::~Media_Foundation(void)
{
HRESULT hr = MFShutdown();
if(!SUCCEEDED(hr))
{
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
DPO->printOut(L"MEDIA FOUNDATION: Resources cannot be released\n");
}
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
bool Media_Foundation::buildListOfDevices()
{
HRESULT hr = S_OK;
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#ifdef HAVE_WINRT
videoDevices *vDs = &videoDevices::getInstance();
hr = vDs->initDevices(_VideoCapture);
#else
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
ComPtr<IMFAttributes> pAttributes = NULL;
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
CoInitialize(NULL);
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
hr = MFCreateAttributes(pAttributes.GetAddressOf(), 1);
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
if (SUCCEEDED(hr))
{
hr = pAttributes->SetGUID(
MF_DEVSOURCE_ATTRIBUTE_SOURCE_TYPE,
MF_DEVSOURCE_ATTRIBUTE_SOURCE_TYPE_VIDCAP_GUID
);
}
if (SUCCEEDED(hr))
{
videoDevices *vDs = &videoDevices::getInstance();
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
hr = vDs->initDevices(pAttributes.Get());
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
}
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#endif
if (FAILED(hr))
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
{
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
DPO->printOut(L"MEDIA FOUNDATION: The access to the video cameras denied\n");
}
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
return (SUCCEEDED(hr));
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
Media_Foundation& Media_Foundation::getInstance()
{
static Media_Foundation instance;
return instance;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
RawImage::RawImage(unsigned int size): ri_new(false), ri_pixels(NULL)
{
ri_size = size;
ri_pixels = new unsigned char[size];
memset((void *)ri_pixels,0,ri_size);
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
bool RawImage::isNew()
{
return ri_new;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
unsigned int RawImage::getSize()
{
return ri_size;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
RawImage::~RawImage(void)
{
delete []ri_pixels;
ri_pixels = NULL;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
long RawImage::CreateInstance(RawImage **ppRImage,unsigned int size)
{
*ppRImage = new (std::nothrow) RawImage(size);
if (ppRImage == NULL)
{
return E_OUTOFMEMORY;
}
return S_OK;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
void RawImage::setCopy(const BYTE * pSampleBuffer)
{
memcpy(ri_pixels, pSampleBuffer, ri_size);
ri_new = true;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
void RawImage::fastCopy(const BYTE * pSampleBuffer)
{
memcpy(ri_pixels, pSampleBuffer, ri_size);
ri_new = true;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
unsigned char * RawImage::getpPixels()
{
ri_new = false;
return ri_pixels;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
videoDevice::videoDevice(void): vd_IsSetuped(false), vd_LockOut(OpenLock), vd_pFriendlyName(NULL),
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
vd_Width(0), vd_Height(0), vd_pSource(NULL), vd_pImGrTh(NULL), vd_func(NULL), vd_userData(NULL)
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
{
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#ifdef HAVE_WINRT
vd_pMedCap = nullptr;
vd_pMedCapFail = NULL;
vd_pImGr = NULL;
vd_pAction = nullptr;
#endif
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
void videoDevice::setParametrs(CamParametrs parametrs)
{
if(vd_IsSetuped)
{
if(vd_pSource)
{
Parametr *pParametr = (Parametr *)(&parametrs);
Parametr *pPrevParametr = (Parametr *)(&vd_PrevParametrs);
IAMVideoProcAmp *pProcAmp = NULL;
HRESULT hr = vd_pSource->QueryInterface(IID_PPV_ARGS(&pProcAmp));
if (SUCCEEDED(hr))
{
for(unsigned int i = 0; i < 10; i++)
{
if(pPrevParametr[i].CurrentValue != pParametr[i].CurrentValue || pPrevParametr[i].Flag != pParametr[i].Flag)
hr = pProcAmp->Set(VideoProcAmp_Brightness + i, pParametr[i].CurrentValue, pParametr[i].Flag);
}
pProcAmp->Release();
}
IAMCameraControl *pProcControl = NULL;
hr = vd_pSource->QueryInterface(IID_PPV_ARGS(&pProcControl));
if (SUCCEEDED(hr))
{
for(unsigned int i = 0; i < 7; i++)
{
if(pPrevParametr[10 + i].CurrentValue != pParametr[10 + i].CurrentValue || pPrevParametr[10 + i].Flag != pParametr[10 + i].Flag)
Code review notes fixed.
2013-04-03 18:48:23 +08:00
hr = pProcControl->Set(CameraControl_Pan+i, pParametr[10 + i].CurrentValue, pParametr[10 + i].Flag);
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
}
pProcControl->Release();
}
vd_PrevParametrs = parametrs;
}
}
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
CamParametrs videoDevice::getParametrs()
{
CamParametrs out;
if(vd_IsSetuped)
{
if(vd_pSource)
{
Parametr *pParametr = (Parametr *)(&out);
IAMVideoProcAmp *pProcAmp = NULL;
HRESULT hr = vd_pSource->QueryInterface(IID_PPV_ARGS(&pProcAmp));
if (SUCCEEDED(hr))
{
for(unsigned int i = 0; i < 10; i++)
{
Parametr temp;
hr = pProcAmp->GetRange(VideoProcAmp_Brightness+i, &temp.Min, &temp.Max, &temp.Step, &temp.Default, &temp.Flag);
if (SUCCEEDED(hr))
{
temp.CurrentValue = temp.Default;
pParametr[i] = temp;
}
}
pProcAmp->Release();
}
IAMCameraControl *pProcControl = NULL;
hr = vd_pSource->QueryInterface(IID_PPV_ARGS(&pProcControl));
if (SUCCEEDED(hr))
{
for(unsigned int i = 0; i < 7; i++)
{
Parametr temp;
hr = pProcControl->GetRange(CameraControl_Pan+i, &temp.Min, &temp.Max, &temp.Step, &temp.Default, &temp.Flag);
if (SUCCEEDED(hr))
{
temp.CurrentValue = temp.Default;
pParametr[10 + i] = temp;
}
}
pProcControl->Release();
}
}
}
return out;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#ifdef HAVE_WINRT
long videoDevice::resetDevice(MAKE_WRL_REF(_IDeviceInformation) pDevice)
#else
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
long videoDevice::resetDevice(IMFActivate *pActivate)
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#endif
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
{
HRESULT hr = -1;
vd_CurrentFormats.clear();
if(vd_pFriendlyName)
CoTaskMemFree(vd_pFriendlyName);
vd_pFriendlyName = NULL;
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#ifdef HAVE_WINRT
if (pDevice)
{
ACTIVATE_OBJ(RuntimeClass_Windows_Media_Capture_MediaCapture, MAKE_WRL_OBJ(_MediaCapture), pIMedCap, hr)
if (FAILED(hr)) return hr;
ACTIVATE_OBJ(RuntimeClass_Windows_Media_Capture_MediaCaptureInitializationSettings, MAKE_WRL_OBJ(_MediaCaptureInitializationSettings), pCapInitSet, hr)
if (FAILED(hr)) return hr;
_StringObj str;
WRL_PROP_GET(pDevice, Name, *REF_WRL_OBJ(str), hr)
if (FAILED(hr)) return hr;
unsigned int length = 0;
PCWSTR wstr = WindowsGetStringRawBuffer(reinterpret_cast<HSTRING>(DEREF_WRL_OBJ(str)), &length);
vd_pFriendlyName = (wchar_t*)CoTaskMemAlloc((length + 1) * sizeof(wchar_t));
wcscpy(vd_pFriendlyName, wstr);
WRL_PROP_GET(pDevice, Id, *REF_WRL_OBJ(str), hr)
if (FAILED(hr)) return hr;
WRL_PROP_PUT(pCapInitSet, VideoDeviceId, DEREF_WRL_OBJ(str), hr)
if (FAILED(hr)) return hr;
WRL_PROP_PUT(pCapInitSet, StreamingCaptureMode, _Video, hr)
if (FAILED(hr)) return hr;
MAKE_WRL_REF(_AsyncAction) pAction;
WRL_METHOD(DEREF_WRL_OBJ(pIMedCap), _InitializeWithSettingsAsync, pAction, hr, DEREF_WRL_OBJ(pCapInitSet))
if (FAILED(hr)) return hr;
MAKE_WRL_AGILE_REF(_MediaCapture) pAgileMedCap;
pAgileMedCap = PREPARE_TRANSFER_WRL_OBJ(pIMedCap);
Concurrency::critical_section::scoped_lock _LockHolder(vd_lock);
MAKE_WRL_REF(_AsyncAction) pOldAction = vd_pAction;
SAVE_CURRENT_CONTEXT(context);
vd_pAction = reinterpret_cast<MAKE_WRL_REF(_AsyncAction)>(BEGIN_CREATE_ASYNC(pAction, pOldAction, context, &pAgileMedCap, this)
HRESULT hr;
if (pOldAction) DO_ACTION_SYNCHRONOUSLY(hr, pOldAction, GET_CURRENT_CONTEXT);
DO_ACTION_SYNCHRONOUSLY(hr, pAction, context);
if (SUCCEEDED(hr)) {
//all camera capture calls only in original context
BEGIN_CALL_IN_CONTEXT(hr, context, pAgileMedCap, this)
enumerateCaptureFormats(DEREF_AGILE_WRL_OBJ(pAgileMedCap));
END_CALL_IN_CONTEXT(S_OK)
}
buildLibraryofTypes();
RELEASE_AGILE_WRL(pAgileMedCap)
END_CREATE_ASYNC(hr));
}
#else
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
if(pActivate)
Code review notes fixed.
2013-04-03 18:48:23 +08:00
{
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
IMFMediaSource *pSource = NULL;
hr = pActivate->GetAllocatedString(
MF_DEVSOURCE_ATTRIBUTE_FRIENDLY_NAME,
&vd_pFriendlyName,
NULL
);
hr = pActivate->ActivateObject(
__uuidof(IMFMediaSource),
(void**)&pSource
);
enumerateCaptureFormats(pSource);
buildLibraryofTypes();
SafeRelease(&pSource);
Code review notes fixed.
2013-04-03 18:48:23 +08:00
if(FAILED(hr))
{
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
vd_pFriendlyName = NULL;
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
DPO->printOut(L"VIDEODEVICE %i: IMFMediaSource interface cannot be created \n", vd_CurrentNumber);
}
}
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#endif
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
return hr;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#ifdef HAVE_WINRT
long videoDevice::readInfoOfDevice(MAKE_WRL_REF(_IDeviceInformation) pDevice, unsigned int Num)
{
HRESULT hr = -1;
vd_CurrentNumber = Num;
hr = resetDevice(pDevice);
return hr;
}
#else
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
long videoDevice::readInfoOfDevice(IMFActivate *pActivate, unsigned int Num)
{
HRESULT hr = -1;
vd_CurrentNumber = Num;
hr = resetDevice(pActivate);
return hr;
}
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#endif
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#ifdef HAVE_WINRT
long videoDevice::checkDevice(_DeviceClass devClass, DEFINE_TASK<HRESULT>* pTask, MAKE_WRL_REF(_IDeviceInformation)* ppDevice)
{
HRESULT hr = S_OK;
ACTIVATE_STATIC_OBJ(RuntimeClass_Windows_Devices_Enumeration_DeviceInformation, MAKE_WRL_OBJ(_DeviceInformationStatics), pDevStat, hr)
if (FAILED(hr)) return hr;
MAKE_WRL_REF(_AsyncOperation<MAKE_WRL_REF(_DeviceInformationCollection)>) pAction;
WRL_METHOD(pDevStat, _FindAllAsyncDeviceClass, pAction, hr, devClass)
if (SUCCEEDED(hr)) {
*pTask = CREATE_TASK([pAction, &ppDevice, this]() -> HRESULT {
HRESULT hr;
MAKE_WRL_OBJ(_VectorView<MAKE_WRL_REF(_DeviceInformation)>) pVector;
DO_OPERATION_SYNCHRONOUSLY_VECTOR(hr, pAction, GET_CURRENT_CONTEXT, pVector, _VectorView, _DeviceInformation, _DeviceInformationCollection);
UINT32 count = 0;
if (SUCCEEDED(hr)) WRL_PROP_GET(pVector, Size, count, hr)
if (SUCCEEDED(hr) && count > 0) {
for (UINT32 i = 0; i < count; i++) {
MAKE_WRL_OBJ(_IDeviceInformation) pDevice;
WRL_METHOD(pVector, GetAt, pDevice, hr, i)
if (SUCCEEDED(hr)) {
_StringObj str;
unsigned int length = 0;
WRL_PROP_GET(pDevice, Name, *REF_WRL_OBJ(str), hr)
PCWSTR wstr = WindowsGetStringRawBuffer(reinterpret_cast<HSTRING>(DEREF_WRL_OBJ(str)), &length);
if (wcscmp(wstr, vd_pFriendlyName) == 0) {
*ppDevice = PREPARE_TRANSFER_WRL_OBJ(pDevice);
}
}
}
}
return hr;
});
}
return hr;
}
#else
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
long videoDevice::checkDevice(IMFAttributes *pAttributes, IMFActivate **pDevice)
{
HRESULT hr = S_OK;
IMFActivate **ppDevices = NULL;
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
UINT32 count;
wchar_t *newFriendlyName = NULL;
hr = MFEnumDeviceSources(pAttributes, &ppDevices, &count);
if (SUCCEEDED(hr))
{
if(count > 0)
{
if(count > vd_CurrentNumber)
Code review notes fixed.
2013-04-03 18:48:23 +08:00
{
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
hr = ppDevices[vd_CurrentNumber]->GetAllocatedString(
MF_DEVSOURCE_ATTRIBUTE_FRIENDLY_NAME,
&newFriendlyName,
NULL
);
if (SUCCEEDED(hr))
{
if(wcscmp(newFriendlyName, vd_pFriendlyName) != 0)
{
DPO->printOut(L"VIDEODEVICE %i: Chosen device cannot be found \n", vd_CurrentNumber);
hr = -1;
pDevice = NULL;
}
else
{
*pDevice = ppDevices[vd_CurrentNumber];
(*pDevice)->AddRef();
}
}
else
{
DPO->printOut(L"VIDEODEVICE %i: Name of device cannot be gotten \n", vd_CurrentNumber);
}
}
else
{
DPO->printOut(L"VIDEODEVICE %i: Number of devices more than corrent number of the device \n", vd_CurrentNumber);
hr = -1;
}
for(UINT32 i = 0; i < count; i++)
{
SafeRelease(&ppDevices[i]);
}
SafeRelease(ppDevices);
}
else
hr = -1;
}
else
{
DPO->printOut(L"VIDEODEVICE %i: List of DeviceSources cannot be enumerated \n", vd_CurrentNumber);
}
return hr;
}
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#endif
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
long videoDevice::initDevice()
{
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
HRESULT hr = S_OK;
CoInitialize(NULL);
#ifdef HAVE_WINRT
Concurrency::critical_section::scoped_lock _LockHolder(vd_lock);
MAKE_WRL_REF(_AsyncAction) pOldAction = vd_pAction;
SAVE_CURRENT_CONTEXT(context);
vd_pAction = reinterpret_cast<MAKE_WRL_REF(_AsyncAction)>(BEGIN_CREATE_ASYNC(pOldAction, context, this)
HRESULT hr;
if (pOldAction) DO_ACTION_SYNCHRONOUSLY(hr, pOldAction, GET_CURRENT_CONTEXT);
DEFINE_TASK<HRESULT> pTask;
MAKE_WRL_OBJ(_IDeviceInformation) pDevInfo;
hr = checkDevice(_VideoCapture, &pTask, REF_WRL_OBJ(pDevInfo));
if (SUCCEEDED(hr)) hr = pTask.get();
if (SUCCEEDED(hr)) {
MAKE_WRL_REF(_AsyncAction) pAction;
BEGIN_CALL_IN_CONTEXT(hr, context, pDevInfo, &pAction, context, this)
HRESULT hr;
ACTIVATE_OBJ(RuntimeClass_Windows_Media_Capture_MediaCapture, MAKE_WRL_OBJ(_MediaCapture), pIMedCap, hr)
if (SUCCEEDED(hr)) {
RELEASE_WRL(vd_pMedCap);
vd_pMedCap = PREPARE_TRANSFER_WRL_OBJ(pIMedCap);
ACTIVATE_OBJ(RuntimeClass_Windows_Media_Capture_MediaCaptureInitializationSettings, MAKE_WRL_OBJ(_MediaCaptureInitializationSettings), pCapInitSet, hr)
_StringObj str;
if (SUCCEEDED(hr)) {
WRL_PROP_GET(pDevInfo, Id, *REF_WRL_OBJ(str), hr)
if (SUCCEEDED(hr)) {
WRL_PROP_PUT(pCapInitSet, VideoDeviceId, DEREF_WRL_OBJ(str), hr)
}
}
if (SUCCEEDED(hr))
WRL_PROP_PUT(pCapInitSet, StreamingCaptureMode, _Video, hr)
if (SUCCEEDED(hr)) {
vd_pMedCapFail = create_medcapfailedhandler([this, context](){
HRESULT hr;
BEGIN_CALL_IN_CONTEXT(hr, context, this)
closeDevice();
END_CALL_IN_CONTEXT(S_OK)
});
}
if (SUCCEEDED(hr)) hr = vd_pMedCapFail->AddHandler(reinterpret_cast<ABI::Windows::Media::Capture::IMediaCapture*>(DEREF_AGILE_WRL_OBJ(vd_pMedCap)));
if (SUCCEEDED(hr)) WRL_METHOD(vd_pMedCap, _InitializeWithSettingsAsync, pAction, hr, DEREF_WRL_OBJ(pCapInitSet))
}
END_CALL_IN_CONTEXT(hr)
DO_ACTION_SYNCHRONOUSLY(hr, pAction, context);
}
END_CREATE_ASYNC(hr));
#else
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
ComPtr<IMFAttributes> pAttributes = NULL;
IMFActivate *vd_pActivate = NULL;
hr = MFCreateAttributes(pAttributes.GetAddressOf(), 1);
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
if (SUCCEEDED(hr))
{
hr = pAttributes->SetGUID(
MF_DEVSOURCE_ATTRIBUTE_SOURCE_TYPE,
MF_DEVSOURCE_ATTRIBUTE_SOURCE_TYPE_VIDCAP_GUID
);
Code review notes fixed.
2013-04-03 18:48:23 +08:00
}
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
if (SUCCEEDED(hr))
{
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
hr = checkDevice(pAttributes.Get(), &vd_pActivate);
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
if (SUCCEEDED(hr) && vd_pActivate)
{
SafeRelease(&vd_pSource);
hr = vd_pActivate->ActivateObject(
__uuidof(IMFMediaSource),
(void**)&vd_pSource
);
if (SUCCEEDED(hr))
{
}
SafeRelease(&vd_pActivate);
}
else
{
DPO->printOut(L"VIDEODEVICE %i: Device there is not \n", vd_CurrentNumber);
}
Code review notes fixed.
2013-04-03 18:48:23 +08:00
}
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
else
{
DPO->printOut(L"VIDEODEVICE %i: The attribute of video cameras cannot be getting \n", vd_CurrentNumber);
}
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#endif
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
return hr;
}
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
MediaType videoDevice::getFormat(unsigned int id)
{
if(id < vd_CurrentFormats.size())
{
return vd_CurrentFormats[id];
}
else return MediaType();
}
int videoDevice::getCountFormats()
{
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
return (int)vd_CurrentFormats.size();
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
}
void videoDevice::setEmergencyStopEvent(void *userData, void(*func)(int, void *))
{
vd_func = func;
vd_userData = userData;
}
void videoDevice::closeDevice()
Code review notes fixed.
2013-04-03 18:48:23 +08:00
{
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
if(vd_IsSetuped)
{
vd_IsSetuped = false;
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#ifdef HAVE_WINRT
if (DEREF_AGILE_WRL_OBJ(vd_pMedCap)) {
MAKE_WRL_REF(_AsyncAction) action;
Concurrency::critical_section::scoped_lock _LockHolder(vd_lock);
MAKE_WRL_REF(_AsyncAction) pOldAction = vd_pAction;
vd_pImGr->stopGrabbing(&action);
vd_pMedCapFail->RemoveHandler(reinterpret_cast<ABI::Windows::Media::Capture::IMediaCapture*>(DEREF_AGILE_WRL_OBJ(vd_pMedCap)));
SafeRelease(&vd_pMedCapFail);
vd_pAction = reinterpret_cast<MAKE_WRL_REF(_AsyncAction)>(BEGIN_CREATE_ASYNC(action, pOldAction, this)
HRESULT hr;
if (pOldAction) DO_ACTION_SYNCHRONOUSLY(hr, pOldAction, GET_CURRENT_CONTEXT);
DO_ACTION_SYNCHRONOUSLY(hr, action, GET_CURRENT_CONTEXT);
RELEASE_WRL(vd_pMedCap)
if(vd_LockOut == RawDataLock) {
delete vd_pImGr;
}
vd_pImGr = NULL;
vd_LockOut = OpenLock;
END_CREATE_ASYNC(hr));
return;
}
#endif
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
vd_pSource->Stop();
SafeRelease(&vd_pSource);
if(vd_LockOut == RawDataLock)
{
vd_pImGrTh->stop();
Sleep(500);
delete vd_pImGrTh;
}
vd_pImGrTh = NULL;
Code review notes fixed.
2013-04-03 18:48:23 +08:00
vd_LockOut = OpenLock;
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
DPO->printOut(L"VIDEODEVICE %i: Device is stopped \n", vd_CurrentNumber);
}
}
unsigned int videoDevice::getWidth()
{
if(vd_IsSetuped)
return vd_Width;
else
return 0;
}
unsigned int videoDevice::getHeight()
{
if(vd_IsSetuped)
return vd_Height;
Code review notes fixed.
2013-04-03 18:48:23 +08:00
else
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
return 0;
}
IMFMediaSource *videoDevice::getMediaSource()
{
IMFMediaSource *out = NULL;
if(vd_LockOut == OpenLock)
{
Code review notes fixed.
2013-04-03 18:48:23 +08:00
vd_LockOut = MediaSourceLock;
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
out = vd_pSource;
}
return out;
}
int videoDevice::findType(unsigned int size, unsigned int frameRate)
Code review notes fixed.
2013-04-03 18:48:23 +08:00
{
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
if(vd_CaptureFormats.size() == 0)
return 0;
FrameRateMap FRM = vd_CaptureFormats[size];
if(FRM.size() == 0)
return 0;
UINT64 frameRateMax = 0; SUBTYPEMap STMMax;
if(frameRate == 0)
{
std::map<UINT64, SUBTYPEMap>::iterator f = FRM.begin();
for(; f != FRM.end(); f++)
{
if((*f).first >= frameRateMax)
{
frameRateMax = (*f).first;
STMMax = (*f).second;
}
Code review notes fixed.
2013-04-03 18:48:23 +08:00
}
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
}
else
{
std::map<UINT64, SUBTYPEMap>::iterator f = FRM.begin();
for(; f != FRM.end(); f++)
{
if((*f).first >= frameRateMax)
{
if(frameRate > (*f).first)
{
frameRateMax = (*f).first;
STMMax = (*f).second;
}
}
}
}
if(STMMax.size() == 0)
return 0;
std::map<String, vectorNum>::iterator S = STMMax.begin();
vectorNum VN = (*S).second;
if(VN.size() == 0)
return 0;
return VN[0];
}
GetProperty method for MSMF VideoCapture implemented.
2013-05-23 00:50:54 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
void videoDevice::buildLibraryofTypes()
{
unsigned int size;
unsigned int framerate;
std::vector<MediaType>::iterator i = vd_CurrentFormats.begin();
int count = 0;
for(; i != vd_CurrentFormats.end(); i++)
{
size = (*i).MF_MT_FRAME_SIZE;
GetProperty method for MSMF VideoCapture implemented.
2013-05-23 00:50:54 +08:00
framerate = (*i).MF_MT_FRAME_RATE_NUMERATOR;
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
FrameRateMap FRM = vd_CaptureFormats[size];
SUBTYPEMap STM = FRM[framerate];
String subType((*i).pMF_MT_SUBTYPEName);
vectorNum VN = STM[subType];
VN.push_back(count);
STM[subType] = VN;
FRM[framerate] = STM;
vd_CaptureFormats[size] = FRM;
count++;
}
}
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#ifdef HAVE_WINRT
long videoDevice::setDeviceFormat(MAKE_WRL_REF(_MediaCapture) pSource, unsigned long dwFormatIndex, MAKE_WRL_REF(_AsyncAction)* pAction)
{
HRESULT hr;
MAKE_WRL_OBJ(_VideoDeviceController) pDevCont;
WRL_PROP_GET(pSource, VideoDeviceController, pDevCont, hr)
if (FAILED(hr)) return hr;
GET_WRL_MEDIA_DEVICE_CONTROLLER(pDevCont, pMedDevCont, hr)
if (FAILED(hr)) return hr;
MAKE_WRL_OBJ(_VectorView<MAKE_WRL_REF(_MediaEncodingProperties)>) pVector;
WRL_METHOD(pMedDevCont, GetAvailableMediaStreamProperties, pVector, hr, _VideoPreview)
if (FAILED(hr)) return hr;
MAKE_WRL_OBJ(_MediaEncodingProperties) pMedEncProps;
WRL_METHOD(pVector, GetAt, pMedEncProps, hr, dwFormatIndex)
if (FAILED(hr)) return hr;
WRL_METHOD(pMedDevCont, SetMediaStreamPropertiesAsync, *pAction, hr, _VideoPreview, DEREF_WRL_OBJ(pMedEncProps))
return hr;
}
#endif
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
long videoDevice::setDeviceFormat(IMFMediaSource *pSource, unsigned long dwFormatIndex)
{
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
ComPtr<IMFPresentationDescriptor> pPD = NULL;
ComPtr<IMFStreamDescriptor> pSD = NULL;
ComPtr<IMFMediaTypeHandler> pHandler = NULL;
ComPtr<IMFMediaType> pType = NULL;
HRESULT hr = pSource->CreatePresentationDescriptor(pPD.GetAddressOf());
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
if (FAILED(hr))
{
goto done;
}
BOOL fSelected;
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
hr = pPD->GetStreamDescriptorByIndex(0, &fSelected, pSD.GetAddressOf());
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
if (FAILED(hr))
{
goto done;
}
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
hr = pSD->GetMediaTypeHandler(pHandler.GetAddressOf());
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
if (FAILED(hr))
{
goto done;
}
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
hr = pHandler->GetMediaTypeByIndex((DWORD)dwFormatIndex, pType.GetAddressOf());
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
if (FAILED(hr))
{
goto done;
}
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
hr = pHandler->SetCurrentMediaType(pType.Get());
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
done:
return hr;
}
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
bool videoDevice::isDeviceSetup()
{
return vd_IsSetuped;
}
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
RawImage * videoDevice::getRawImageOut()
{
if(!vd_IsSetuped) return NULL;
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#ifdef HAVE_WINRT
if(vd_pImGr) return vd_pImGr->getRawImage();
#endif
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
if(vd_pImGrTh)
Code review notes fixed.
2013-04-03 18:48:23 +08:00
return vd_pImGrTh->getImageGrabber()->getRawImage();
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
else
{
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
DPO->printOut(L"VIDEODEVICE %i: The instance of ImageGrabberThread class does not exist \n", vd_CurrentNumber);
}
return NULL;
}
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
bool videoDevice::isFrameNew()
{
if(!vd_IsSetuped) return false;
Code review notes fixed.
2013-04-03 18:48:23 +08:00
if(vd_LockOut == RawDataLock || vd_LockOut == OpenLock)
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
{
if(vd_LockOut == OpenLock)
{
vd_LockOut = RawDataLock;
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
//must already be closed
#ifdef HAVE_WINRT
if (DEREF_AGILE_WRL_OBJ(vd_pMedCap)) {
MAKE_WRL_REF(_AsyncAction) action;
if (FAILED(ImageGrabberWinRT::CreateInstance(&vd_pImGr))) return false;
if (FAILED(vd_pImGr->initImageGrabber(DEREF_AGILE_WRL_OBJ(vd_pMedCap), MFVideoFormat_RGB24)) || FAILED(vd_pImGr->startGrabbing(&action))) {
delete vd_pImGr;
return false;
}
Concurrency::critical_section::scoped_lock _LockHolder(vd_lock);
MAKE_WRL_REF(_AsyncAction) pOldAction = vd_pAction;
SAVE_CURRENT_CONTEXT(context);
vd_pAction = reinterpret_cast<MAKE_WRL_REF(_AsyncAction)>(BEGIN_CREATE_ASYNC(action, pOldAction, context, this)
HRESULT hr;
if (pOldAction) DO_ACTION_SYNCHRONOUSLY(hr, pOldAction, GET_CURRENT_CONTEXT);
DO_ACTION_SYNCHRONOUSLY(hr, action, context);
END_CREATE_ASYNC(hr));
return true;
}
#endif
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
HRESULT hr = ImageGrabberThread::CreateInstance(&vd_pImGrTh, vd_pSource, vd_CurrentNumber);
if(FAILED(hr))
{
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
DPO->printOut(L"VIDEODEVICE %i: The instance of ImageGrabberThread class cannot be created.\n", vd_CurrentNumber);
return false;
}
vd_pImGrTh->setEmergencyStopEvent(vd_userData, vd_func);
vd_pImGrTh->start();
return true;
}
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#ifdef HAVE_WINRT
if(vd_pImGr)
return vd_pImGr->getRawImage()->isNew();
#endif
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
if(vd_pImGrTh)
Code review notes fixed.
2013-04-03 18:48:23 +08:00
return vd_pImGrTh->getImageGrabber()->getRawImage()->isNew();
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
}
return false;
}
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
bool videoDevice::isDeviceMediaSource()
{
if(vd_LockOut == MediaSourceLock) return true;
return false;
}
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
bool videoDevice::isDeviceRawDataSource()
{
if(vd_LockOut == RawDataLock) return true;
return false;
}
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
bool videoDevice::setupDevice(unsigned int id)
Code review notes fixed.
2013-04-03 18:48:23 +08:00
{
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
if(!vd_IsSetuped)
{
HRESULT hr = -1;
hr = initDevice();
if(SUCCEEDED(hr))
Code review notes fixed.
2013-04-03 18:48:23 +08:00
{
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#ifdef HAVE_WINRT
Concurrency::critical_section::scoped_lock _LockHolder(vd_lock);
MAKE_WRL_REF(_AsyncAction) pOldAction = vd_pAction;
SAVE_CURRENT_CONTEXT(context);
vd_pAction = reinterpret_cast<MAKE_WRL_REF(_AsyncAction)>(BEGIN_CREATE_ASYNC(pOldAction, context, id, DPO, this)
HRESULT hr;
if (pOldAction) DO_ACTION_SYNCHRONOUSLY(hr, pOldAction, GET_CURRENT_CONTEXT);
#endif
Code review notes fixed.
2013-04-03 18:48:23 +08:00
vd_Width = vd_CurrentFormats[id].width;
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
vd_Height = vd_CurrentFormats[id].height;
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#ifdef HAVE_WINRT
if (DEREF_AGILE_WRL_OBJ(vd_pMedCap)) {
MAKE_WRL_REF(_AsyncAction) pAction;
BEGIN_CALL_IN_CONTEXT(hr, context, id, &pAction, this)
END_CALL_IN_CONTEXT(setDeviceFormat(DEREF_AGILE_WRL_OBJ(vd_pMedCap), (DWORD) id, &pAction))
if (SUCCEEDED(hr)) DO_ACTION_SYNCHRONOUSLY(hr, pAction, context);
} else
#endif
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
hr = setDeviceFormat(vd_pSource, (DWORD) id);
vd_IsSetuped = (SUCCEEDED(hr));
if(vd_IsSetuped)
DPO->printOut(L"\n\nVIDEODEVICE %i: Device is setuped \n", vd_CurrentNumber);
vd_PrevParametrs = getParametrs();
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#ifdef HAVE_WINRT
END_CREATE_ASYNC(hr));
return true;
#else
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
return vd_IsSetuped;
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#endif
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
}
else
{
DPO->printOut(L"VIDEODEVICE %i: Interface IMFMediaSource cannot be got \n", vd_CurrentNumber);
return false;
}
}
else
{
DPO->printOut(L"VIDEODEVICE %i: Device is setuped already \n", vd_CurrentNumber);
return false;
Code review notes fixed.
2013-04-03 18:48:23 +08:00
}
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
}
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
bool videoDevice::setupDevice(unsigned int w, unsigned int h, unsigned int idealFramerate)
Code review notes fixed.
2013-04-03 18:48:23 +08:00
{
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
unsigned int id = findType(w * h, idealFramerate);
return setupDevice(id);
}
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
wchar_t *videoDevice::getName()
{
return vd_pFriendlyName;
}
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
videoDevice::~videoDevice(void)
Code review notes fixed.
2013-04-03 18:48:23 +08:00
{
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
closeDevice();
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#ifdef HAVE_WINRT
RELEASE_WRL(vd_pMedCap)
#endif
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
SafeRelease(&vd_pSource);
if(vd_pFriendlyName)
CoTaskMemFree(vd_pFriendlyName);
}
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#ifdef HAVE_WINRT
HRESULT videoDevice::enumerateCaptureFormats(MAKE_WRL_REF(_MediaCapture) pSource)
{
HRESULT hr;
MAKE_WRL_OBJ(_VideoDeviceController) pDevCont;
WRL_PROP_GET(pSource, VideoDeviceController, pDevCont, hr)
if (FAILED(hr)) return hr;
GET_WRL_MEDIA_DEVICE_CONTROLLER(pDevCont, pMedDevCont, hr)
if (FAILED(hr)) return hr;
MAKE_WRL_OBJ(_VectorView<MAKE_WRL_REF(_MediaEncodingProperties)>) pVector;
WRL_METHOD(pMedDevCont, GetAvailableMediaStreamProperties, pVector, hr, _VideoPreview)
if (FAILED(hr)) return hr;
UINT32 count;
WRL_PROP_GET(pVector, Size, count, hr)
if (FAILED(hr)) return hr;
for (UINT32 i = 0; i < count; i++) {
MAKE_WRL_OBJ(_MediaEncodingProperties) pMedEncProps;
WRL_METHOD(pVector, GetAt, pMedEncProps, hr, i)
if (FAILED(hr)) return hr;
ComPtr<IMFMediaType> pType = NULL;
hr = MediaSink::ConvertPropertiesToMediaType(DEREF_AS_NATIVE_WRL_OBJ(ABI::Windows::Media::MediaProperties::IMediaEncodingProperties, pMedEncProps), &pType);
if (FAILED(hr)) return hr;
MediaType MT = FormatReader::Read(pType.Get());
vd_CurrentFormats.push_back(MT);
}
return hr;
}
#endif
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
HRESULT videoDevice::enumerateCaptureFormats(IMFMediaSource *pSource)
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
{
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
ComPtr<IMFPresentationDescriptor> pPD = NULL;
ComPtr<IMFStreamDescriptor> pSD = NULL;
ComPtr<IMFMediaTypeHandler> pHandler = NULL;
ComPtr<IMFMediaType> pType = NULL;
HRESULT hr = pSource->CreatePresentationDescriptor(pPD.GetAddressOf());
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
if (FAILED(hr))
{
goto done;
}
BOOL fSelected;
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
hr = pPD->GetStreamDescriptorByIndex(0, &fSelected, pSD.GetAddressOf());
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
if (FAILED(hr))
{
goto done;
}
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
hr = pSD->GetMediaTypeHandler(pHandler.GetAddressOf());
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
if (FAILED(hr))
{
goto done;
}
DWORD cTypes = 0;
hr = pHandler->GetMediaTypeCount(&cTypes);
if (FAILED(hr))
{
goto done;
}
for (DWORD i = 0; i < cTypes; i++)
{
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
hr = pHandler->GetMediaTypeByIndex(i, pType.GetAddressOf());
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
if (FAILED(hr))
{
goto done;
}
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
MediaType MT = FormatReader::Read(pType.Get());
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
vd_CurrentFormats.push_back(MT);
}
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
done:
return hr;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
videoDevices::videoDevices(void): count(0)
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
{
#ifdef HAVE_WINRT
vds_enumTask = nullptr;
#endif
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
void videoDevices::clearDevices()
{
std::vector<videoDevice *>::iterator i = vds_Devices.begin();
for(; i != vds_Devices.end(); ++i)
delete (*i);
vds_Devices.clear();
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
videoDevices::~videoDevices(void)
Code review notes fixed.
2013-04-03 18:48:23 +08:00
{
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
clearDevices();
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
videoDevice * videoDevices::getDevice(unsigned int i)
{
if(i >= vds_Devices.size())
{
return NULL;
}
if(i < 0)
{
return NULL;
}
return vds_Devices[i];
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#ifdef HAVE_WINRT
long videoDevices::initDevices(_DeviceClass devClass)
{
HRESULT hr = S_OK;
ACTIVATE_STATIC_OBJ(RuntimeClass_Windows_Devices_Enumeration_DeviceInformation, MAKE_WRL_OBJ(_DeviceInformationStatics), pDevStat, hr)
if (FAILED(hr)) return hr;
MAKE_WRL_REF(_AsyncOperation<MAKE_WRL_REF(_DeviceInformationCollection)>) pAction;
WRL_METHOD(pDevStat, _FindAllAsyncDeviceClass, pAction, hr, devClass)
if (SUCCEEDED(hr)) {
SAVE_CURRENT_CONTEXT(context);
vds_enumTask = reinterpret_cast<MAKE_WRL_REF(_AsyncAction)>(BEGIN_CREATE_ASYNC(pAction, context, this)
HRESULT hr;
MAKE_WRL_OBJ(_VectorView<MAKE_WRL_REF(_DeviceInformation)>) pVector;
DO_OPERATION_SYNCHRONOUSLY_VECTOR(hr, pAction, GET_CURRENT_CONTEXT, pVector, _VectorView, _DeviceInformation, _DeviceInformationCollection);
if (SUCCEEDED(hr)) WRL_PROP_GET(pVector, Size, count, hr)
if (SUCCEEDED(hr) && count > 0) {
for (UINT32 i = 0; i < count; i++) {
videoDevice *vd = new videoDevice;
MAKE_WRL_OBJ(_IDeviceInformation) pDevice;
WRL_METHOD(pVector, GetAt, pDevice, hr, i)
if (SUCCEEDED(hr)) {
BEGIN_CALL_IN_CONTEXT(hr, context, vd, pDevice, i)
vd->readInfoOfDevice(DEREF_WRL_OBJ(pDevice), i);
END_CALL_IN_CONTEXT(S_OK)
vds_Devices.push_back(vd);
}
}
}
END_CREATE_ASYNC(hr));
}
return hr;
}
#else
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
long videoDevices::initDevices(IMFAttributes *pAttributes)
{
HRESULT hr = S_OK;
clearDevices();
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
IMFActivate **ppDevices = NULL;
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
hr = MFEnumDeviceSources(pAttributes, &ppDevices, &count);
if (SUCCEEDED(hr))
{
if(count > 0)
{
for(UINT32 i = 0; i < count; i++)
{
videoDevice *vd = new videoDevice;
vd->readInfoOfDevice(ppDevices[i], i);
vds_Devices.push_back(vd);
SafeRelease(&ppDevices[i]);
}
SafeRelease(ppDevices);
}
else
hr = -1;
}
else
{
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
DPO->printOut(L"VIDEODEVICES: The instances of the videoDevice class cannot be created\n");
}
return hr;
}
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#endif
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Difference in method definition and declaration fixed (Bug #3189).
2013-08-05 02:34:36 +08:00
unsigned int videoDevices::getCount()
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
{
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
return (unsigned int)vds_Devices.size();
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Code review notes fixed.
2013-04-03 18:48:23 +08:00
videoDevices& videoDevices::getInstance()
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
{
static videoDevices instance;
return instance;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
Parametr::Parametr()
{
CurrentValue = 0;
Min = 0;
Max = 0;
Step = 0;
Code review notes fixed.
2013-04-03 18:48:23 +08:00
Default = 0;
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
Flag = 0;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
MediaType::MediaType()
{
pMF_MT_AM_FORMAT_TYPEName = NULL;
pMF_MT_MAJOR_TYPEName = NULL;
pMF_MT_SUBTYPEName = NULL;
Clear();
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
MediaType::~MediaType()
{
Clear();
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
void MediaType::Clear()
{
MF_MT_FRAME_SIZE = 0;
height = 0;
width = 0;
MF_MT_YUV_MATRIX = 0;
MF_MT_VIDEO_LIGHTING = 0;
MF_MT_DEFAULT_STRIDE = 0;
MF_MT_VIDEO_CHROMA_SITING = 0;
MF_MT_FIXED_SIZE_SAMPLES = 0;
MF_MT_VIDEO_NOMINAL_RANGE = 0;
GetProperty method for MSMF VideoCapture implemented.
2013-05-23 00:50:54 +08:00
MF_MT_FRAME_RATE_NUMERATOR = 0;
MF_MT_FRAME_RATE_DENOMINATOR = 0;
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
MF_MT_PIXEL_ASPECT_RATIO = 0;
MF_MT_PIXEL_ASPECT_RATIO_low = 0;
MF_MT_ALL_SAMPLES_INDEPENDENT = 0;
MF_MT_FRAME_RATE_RANGE_MIN = 0;
MF_MT_FRAME_RATE_RANGE_MIN_low = 0;
MF_MT_SAMPLE_SIZE = 0;
MF_MT_VIDEO_PRIMARIES = 0;
MF_MT_INTERLACE_MODE = 0;
MF_MT_FRAME_RATE_RANGE_MAX = 0;
MF_MT_FRAME_RATE_RANGE_MAX_low = 0;
memset(&MF_MT_MAJOR_TYPE, 0, sizeof(GUID));
memset(&MF_MT_AM_FORMAT_TYPE, 0, sizeof(GUID));
memset(&MF_MT_SUBTYPE, 0, sizeof(GUID));
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
videoInput::videoInput(void): accessToDevices(false)
{
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
DPO->printOut(L"\n***** VIDEOINPUT LIBRARY - 2013 (Author: Evgeny Pereguda) *****\n\n");
updateListOfDevices();
if(!accessToDevices)
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
DPO->printOut(L"INITIALIZATION: There is not any suitable video device\n");
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
void videoInput::updateListOfDevices()
{
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
Media_Foundation *MF = &Media_Foundation::getInstance();
accessToDevices = MF->buildListOfDevices();
if(!accessToDevices)
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
DPO->printOut(L"UPDATING: There is not any suitable video device\n");
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
videoInput::~videoInput(void)
{
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
DPO->printOut(L"\n***** CLOSE VIDEOINPUT LIBRARY - 2013 *****\n\n");
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
IMFMediaSource *videoInput::getMediaSource(int deviceID)
{
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
if(accessToDevices)
{
videoDevices *VDS = &videoDevices::getInstance();
videoDevice * VD = VDS->getDevice(deviceID);
if(VD)
{
IMFMediaSource *out = VD->getMediaSource();
if(!out)
DPO->printOut(L"VideoDevice %i: There is not any suitable IMFMediaSource interface\n", deviceID);
return out;
}
}
else
{
DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
}
return NULL;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
bool videoInput::setupDevice(int deviceID, unsigned int id)
{
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
if (deviceID < 0 )
{
DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
return false;
}
if(accessToDevices)
{
videoDevices *VDS = &videoDevices::getInstance();
videoDevice * VD = VDS->getDevice(deviceID);
if(VD)
{
bool out = VD->setupDevice(id);
if(!out)
DPO->printOut(L"VIDEODEVICE %i: This device cannot be started\n", deviceID);
return out;
}
}
else
{
DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
}
return false;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
bool videoInput::setupDevice(int deviceID, unsigned int w, unsigned int h, unsigned int idealFramerate)
{
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
if (deviceID < 0 )
{
DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
return false;
}
if(accessToDevices)
{
videoDevices *VDS = &videoDevices::getInstance();
videoDevice * VD = VDS->getDevice(deviceID);
if(VD)
{
bool out = VD->setupDevice(w, h, idealFramerate);
if(!out)
DPO->printOut(L"VIDEODEVICE %i: this device cannot be started\n", deviceID);
return out;
}
}
else
{
DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n", deviceID);
}
return false;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
MediaType videoInput::getFormat(int deviceID, unsigned int id)
{
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
if (deviceID < 0)
{
DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
return MediaType();
}
if(accessToDevices)
{
videoDevices *VDS = &videoDevices::getInstance();
videoDevice * VD = VDS->getDevice(deviceID);
Code review notes fixed.
2013-04-03 18:48:23 +08:00
if(VD)
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
return VD->getFormat(id);
}
else
{
DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
}
return MediaType();
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
bool videoInput::isDeviceSetup(int deviceID)
{
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
if (deviceID < 0)
{
DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
return false;
}
if(accessToDevices)
{
videoDevices *VDS = &videoDevices::getInstance();
videoDevice * VD = VDS->getDevice(deviceID);
if(VD)
return VD->isDeviceSetup();
}
else
{
DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
}
return false;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
bool videoInput::isDeviceMediaSource(int deviceID)
{
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
if (deviceID < 0)
{
DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
return false;
}
if(accessToDevices)
{
videoDevices *VDS = &videoDevices::getInstance();
videoDevice * VD = VDS->getDevice(deviceID);
if(VD)
return VD->isDeviceMediaSource();
}
else
{
DPO->printOut(L"Device(s): There is not any suitable video device\n");
}
return false;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
bool videoInput::isDeviceRawDataSource(int deviceID)
{
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
if (deviceID < 0)
{
DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
return false;
}
if(accessToDevices)
{
videoDevices *VDS = &videoDevices::getInstance();
videoDevice * VD = VDS->getDevice(deviceID);
if(VD)
{
bool isRaw = VD->isDeviceRawDataSource();
return isRaw;
}
}
else
{
DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
}
return false;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
bool videoInput::isFrameNew(int deviceID)
Code review notes fixed.
2013-04-03 18:48:23 +08:00
{
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
if (deviceID < 0)
{
DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
return false;
}
if(accessToDevices)
{
if(!isDeviceSetup(deviceID))
{
if(isDeviceMediaSource(deviceID))
return false;
}
videoDevices *VDS = &videoDevices::getInstance();
videoDevice * VD = VDS->getDevice(deviceID);
if(VD)
{
return VD->isFrameNew();
}
}
else
{
DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
}
return false;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#ifdef HAVE_WINRT
void videoInput::waitForDevice(int deviceID)
{
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
if (deviceID < 0)
{
DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
return;
}
if(accessToDevices)
{
if(!isDeviceSetup(deviceID))
{
if(isDeviceMediaSource(deviceID))
return;
}
videoDevices *VDS = &videoDevices::getInstance();
videoDevice * VD = VDS->getDevice(deviceID);
if(VD)
{
VD->waitForDevice();
}
}
else
{
DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
}
return;
}
#endif
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
unsigned int videoInput::getCountFormats(int deviceID)
{
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
if (deviceID < 0)
{
DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
return 0;
}
if(accessToDevices)
{
videoDevices *VDS = &videoDevices::getInstance();
videoDevice * VD = VDS->getDevice(deviceID);
if(VD)
return VD->getCountFormats();
}
else
{
DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
}
return 0;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
void videoInput::closeAllDevices()
{
videoDevices *VDS = &videoDevices::getInstance();
for(unsigned int i = 0; i < VDS->getCount(); i++)
closeDevice(i);
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
void videoInput::setParametrs(int deviceID, CamParametrs parametrs)
{
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
if (deviceID < 0)
{
DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
return;
}
if(accessToDevices)
{
videoDevices *VDS = &videoDevices::getInstance();
videoDevice *VD = VDS->getDevice(deviceID);
if(VD)
VD->setParametrs(parametrs);
}
else
{
DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
}
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
CamParametrs videoInput::getParametrs(int deviceID)
{
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
CamParametrs out;
if (deviceID < 0)
{
DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
return out;
}
if(accessToDevices)
{
videoDevices *VDS = &videoDevices::getInstance();
videoDevice *VD = VDS->getDevice(deviceID);
if(VD)
out = VD->getParametrs();
}
else
{
DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
}
return out;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
void videoInput::closeDevice(int deviceID)
{
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
if (deviceID < 0)
{
DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
return;
}
if(accessToDevices)
{
videoDevices *VDS = &videoDevices::getInstance();
videoDevice *VD = VDS->getDevice(deviceID);
if(VD)
VD->closeDevice();
}
else
{
DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
}
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
unsigned int videoInput::getWidth(int deviceID)
{
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
if (deviceID < 0)
{
DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
return 0;
}
if(accessToDevices)
{
videoDevices *VDS = &videoDevices::getInstance();
videoDevice * VD = VDS->getDevice(deviceID);
Code review notes fixed.
2013-04-03 18:48:23 +08:00
if(VD)
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
return VD->getWidth();
}
else
{
DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
}
return 0;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
unsigned int videoInput::getHeight(int deviceID)
{
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
if (deviceID < 0)
{
DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
return 0;
}
if(accessToDevices)
{
videoDevices *VDS = &videoDevices::getInstance();
videoDevice * VD = VDS->getDevice(deviceID);
if(VD)
return VD->getHeight();
}
else
{
DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
}
return 0;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
wchar_t *videoInput::getNameVideoDevice(int deviceID)
{
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
if (deviceID < 0)
{
DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
return NULL;
}
if(accessToDevices)
{
videoDevices *VDS = &videoDevices::getInstance();
videoDevice * VD = VDS->getDevice(deviceID);
if(VD)
return VD->getName();
}
else
{
DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
}
return L"Empty";
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
unsigned int videoInput::listDevices(bool silent)
{
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
int out = 0;
if(accessToDevices)
{
videoDevices *VDS = &videoDevices::getInstance();
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#ifdef HAVE_WINRT
VDS->waitInit();
#endif
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
out = VDS->getCount();
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
if(!silent)DPO->printOut(L"\nVIDEOINPUT SPY MODE!\n\n");
if(!silent)DPO->printOut(L"SETUP: Looking For Capture Devices\n");
for(int i = 0; i < out; i++)
{
if(!silent)DPO->printOut(L"SETUP: %i) %s \n",i, getNameVideoDevice(i));
}
if(!silent)DPO->printOut(L"SETUP: %i Device(s) found\n\n", out);
}
else
{
DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
}
return out;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Code review notes fixed.
2013-04-03 18:48:23 +08:00
videoInput& videoInput::getInstance()
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
{
static videoInput instance;
return instance;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
bool videoInput::isDevicesAcceable()
{
return accessToDevices;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
void videoInput::setVerbose(bool state)
{
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
Code review notes fixed.
2013-04-03 18:48:23 +08:00
DPO->setVerbose(state);
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
void videoInput::setEmergencyStopEvent(int deviceID, void *userData, void(*func)(int, void *))
{
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
if (deviceID < 0)
{
DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
return;
}
if(accessToDevices)
{
if(func)
{
videoDevices *VDS = &videoDevices::getInstance();
videoDevice * VD = VDS->getDevice(deviceID);
Code review notes fixed.
2013-04-03 18:48:23 +08:00
if(VD)
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
VD->setEmergencyStopEvent(userData, func);
}
}
else
{
DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
}
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
bool videoInput::getPixels(int deviceID, unsigned char * dstBuffer, bool flipRedAndBlue, bool flipImage)
{
bool success = false;
unsigned int bytes = 3;
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
if (deviceID < 0)
{
DPO->printOut(L"VIDEODEVICE %i: Invalid device ID\n", deviceID);
return success;
}
if(accessToDevices)
{
bool isRaw = isDeviceRawDataSource(deviceID);
if(isRaw)
{
videoDevices *VDS = &videoDevices::getInstance();
DebugPrintOut *DPO = &DebugPrintOut::getInstance();
RawImage *RIOut = VDS->getDevice(deviceID)->getRawImageOut();
if(RIOut)
{
unsigned int height = VDS->getDevice(deviceID)->getHeight();
Code review notes fixed.
2013-04-03 18:48:23 +08:00
unsigned int width = VDS->getDevice(deviceID)->getWidth();
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
unsigned int size = bytes * width * height;
if(size == RIOut->getSize())
Code review notes fixed.
2013-04-03 18:48:23 +08:00
{
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
processPixels(RIOut->getpPixels(), dstBuffer, width, height, bytes, flipRedAndBlue, flipImage);
success = true;
}
else
{
DPO->printOut(L"ERROR: GetPixels() - bufferSizes do not match!\n");
}
}
else
{
DPO->printOut(L"ERROR: GetPixels() - Unable to grab frame for device %i\n", deviceID);
Code review notes fixed.
2013-04-03 18:48:23 +08:00
}
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
}
else
{
DPO->printOut(L"ERROR: GetPixels() - Not raw data source device %i\n", deviceID);
}
}
else
{
DPO->printOut(L"VIDEODEVICE(s): There is not any suitable video device\n");
}
return success;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Code review notes fixed.
2013-04-03 18:48:23 +08:00
void videoInput::processPixels(unsigned char * src, unsigned char * dst, unsigned int width,
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
unsigned int height, unsigned int bpp, bool bRGB, bool bFlip)
Code review notes fixed.
2013-04-03 18:48:23 +08:00
{
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
unsigned int widthInBytes = width * bpp;
unsigned int numBytes = widthInBytes * height;
int *dstInt, *srcInt;
if(!bRGB)
Code review notes fixed.
2013-04-03 18:48:23 +08:00
{
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
if(bFlip)
{
for(unsigned int y = 0; y < height; y++)
{
dstInt = (int *)(dst + (y * widthInBytes));
srcInt = (int *)(src + ( (height -y -1) * widthInBytes));
memcpy(dstInt, srcInt, widthInBytes);
}
}
else
{
memcpy(dst, src, numBytes);
}
}
else
{
if(bFlip)
{
unsigned int x = 0;
unsigned int y = (height - 1) * widthInBytes;
src += y;
for(unsigned int i = 0; i < numBytes; i+=3)
{
if(x >= width)
{
x = 0;
src -= widthInBytes*2;
}
*dst = *(src+2);
dst++;
*dst = *(src+1);
Code review notes fixed.
2013-04-03 18:48:23 +08:00
dst++;
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
*dst = *src;
Code review notes fixed.
2013-04-03 18:48:23 +08:00
dst++;
src+=3;
x++;
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
}
}
else
Code review notes fixed.
2013-04-03 18:48:23 +08:00
{
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
for(unsigned int i = 0; i < numBytes; i+=3)
{
*dst = *(src+2);
dst++;
*dst = *(src+1);
Code review notes fixed.
2013-04-03 18:48:23 +08:00
dst++;
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
*dst = *src;
Code review notes fixed.
2013-04-03 18:48:23 +08:00
dst++;
src+=3;
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
}
}
}
}
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
/******* Capturing video from camera via Microsoft Media Foundation **********/
class CvCaptureCAM_MSMF : public CvCapture
{
public:
CvCaptureCAM_MSMF();
virtual ~CvCaptureCAM_MSMF();
virtual bool open( int index );
virtual void close();
virtual double getProperty(int);
virtual bool setProperty(int, double);
virtual bool grabFrame();
virtual IplImage* retrieveFrame(int);
virtual int getCaptureDomain() { return CV_CAP_MSMF; } // Return the type of the capture object: CV_CAP_VFW, etc...
protected:
void init();
Test fixes Calib3d badarg tests fixed. Default exception handling enabled; Highgui Video I/O tests fixed on ARM; Code cleanup.
2013-06-27 14:56:50 +08:00
int index, width, height, fourcc;
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
IplImage* frame;
videoInput VI;
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#ifdef HAVE_WINRT
DEFINE_TASK<bool> openTask;
Concurrency::critical_section lock;
#endif
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
};
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
struct SuppressVideoInputMessages
{
SuppressVideoInputMessages() { videoInput::setVerbose(true); }
};
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
static SuppressVideoInputMessages do_it;
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
CvCaptureCAM_MSMF::CvCaptureCAM_MSMF():
index(-1),
width(-1),
height(-1),
fourcc(-1),
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
frame(NULL),
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
VI(videoInput::getInstance())
{
CoInitialize(0);
}
Test fixes Calib3d badarg tests fixed. Default exception handling enabled; Highgui Video I/O tests fixed on ARM; Code cleanup.
2013-06-27 14:56:50 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
CvCaptureCAM_MSMF::~CvCaptureCAM_MSMF()
{
close();
CoUninitialize();
}
Test fixes Calib3d badarg tests fixed. Default exception handling enabled; Highgui Video I/O tests fixed on ARM; Code cleanup.
2013-06-27 14:56:50 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
void CvCaptureCAM_MSMF::close()
{
if( index >= 0 )
{
VI.closeDevice(index);
index = -1;
cvReleaseImage(&frame);
}
Test fixes Calib3d badarg tests fixed. Default exception handling enabled; Highgui Video I/O tests fixed on ARM; Code cleanup.
2013-06-27 14:56:50 +08:00
width = height = -1;
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
// Initialize camera input
bool CvCaptureCAM_MSMF::open( int _index )
{
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#ifdef HAVE_WINRT
SAVE_CURRENT_CONTEXT(context);
auto func = [_index, context, this]() -> bool {
#endif
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
int try_index = _index;
int devices = 0;
close();
devices = VI.listDevices(true);
if (devices == 0)
return false;
try_index = try_index < 0 ? 0 : (try_index > devices-1 ? devices-1 : try_index);
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#ifdef HAVE_WINRT
HRESULT hr;
BEGIN_CALL_IN_CONTEXT(hr, context, this, try_index)
#endif
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
VI.setupDevice(try_index);
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#ifdef HAVE_WINRT
END_CALL_IN_CONTEXT(S_OK)
VI.waitForDevice(try_index);
BEGIN_CALL_IN_CONTEXT(hr, context, this, try_index)
HRESULT hr = S_OK;
#endif
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
if( !VI.isFrameNew(try_index) )
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#ifdef HAVE_WINRT
hr = E_FAIL;
#else
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
return false;
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#endif
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
index = try_index;
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
#ifdef HAVE_WINRT
END_CALL_IN_CONTEXT(hr);
return true;
};
Concurrency::critical_section::scoped_lock _LockHolder(lock);
CREATE_OR_CONTINUE_TASK(openTask, bool, func)
#endif
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
return true;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
bool CvCaptureCAM_MSMF::grabFrame()
{
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
while (VI.isDeviceSetup(index) && !VI.isFrameNew(index))
Sleep(1);
return VI.isDeviceSetup(index);
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
IplImage* CvCaptureCAM_MSMF::retrieveFrame(int)
{
if( !frame || (int)VI.getWidth(index) != frame->width || (int)VI.getHeight(index) != frame->height )
{
if (frame)
cvReleaseImage( &frame );
unsigned int w = VI.getWidth(index), h = VI.getHeight(index);
frame = cvCreateImage( cvSize(w,h), 8, 3 );
}
VI.getPixels( index, (uchar*)frame->imageData, false, true );
return frame;
}
Media Foundation based VideoCapture improved Code formating fixed; GrabFrame method implemented correclty.
2013-04-16 21:25:10 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
double CvCaptureCAM_MSMF::getProperty( int property_id )
{
// image format proprrties
switch( property_id )
{
case CV_CAP_PROP_FRAME_WIDTH:
return VI.getWidth(index);
case CV_CAP_PROP_FRAME_HEIGHT:
return VI.getHeight(index);
}
return -1;
}
bool CvCaptureCAM_MSMF::setProperty( int property_id, double value )
{
// image capture properties
bool handled = false;
switch( property_id )
{
case CV_CAP_PROP_FRAME_WIDTH:
width = cvRound(value);
handled = true;
break;
case CV_CAP_PROP_FRAME_HEIGHT:
height = cvRound(value);
handled = true;
break;
}
Test fixes Calib3d badarg tests fixed. Default exception handling enabled; Highgui Video I/O tests fixed on ARM; Code cleanup.
2013-06-27 14:56:50 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
if ( handled ) {
if( width > 0 && height > 0 )
{
Test fixes Calib3d badarg tests fixed. Default exception handling enabled; Highgui Video I/O tests fixed on ARM; Code cleanup.
2013-06-27 14:56:50 +08:00
if( width != (int)VI.getWidth(index) || height != (int)VI.getHeight(index) && VI.isDeviceSetup(index))//|| fourcc != VI.getFourcc(index) )
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
{
Test fixes Calib3d badarg tests fixed. Default exception handling enabled; Highgui Video I/O tests fixed on ARM; Code cleanup.
2013-06-27 14:56:50 +08:00
VI.closeDevice(index);
VI.setupDevice(index, width, height);
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
}
Test fixes Calib3d badarg tests fixed. Default exception handling enabled; Highgui Video I/O tests fixed on ARM; Code cleanup.
2013-06-27 14:56:50 +08:00
return VI.isDeviceSetup(index);
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
}
return true;
}
Test fixes Calib3d badarg tests fixed. Default exception handling enabled; Highgui Video I/O tests fixed on ARM; Code cleanup.
2013-06-27 14:56:50 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
return false;
}
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
class CvCaptureFile_MSMF : public CvCapture
{
public:
CvCaptureFile_MSMF();
virtual ~CvCaptureFile_MSMF();
virtual bool open( const char* filename );
virtual void close();
virtual double getProperty(int);
virtual bool setProperty(int, double);
virtual bool grabFrame();
virtual IplImage* retrieveFrame(int);
virtual int getCaptureDomain() { return CV_CAP_MSMF; }
protected:
ImageGrabberThread* grabberThread;
IMFMediaSource* videoFileSource;
std::vector<MediaType> captureFormats;
int captureFormatIndex;
IplImage* frame;
bool isOpened;
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
HRESULT enumerateCaptureFormats(IMFMediaSource *pSource);
GetProperty method for MSMF VideoCapture implemented.
2013-05-23 00:50:54 +08:00
HRESULT getSourceDuration(IMFMediaSource *pSource, MFTIME *pDuration);
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
};
CvCaptureFile_MSMF::CvCaptureFile_MSMF():
grabberThread(NULL),
videoFileSource(NULL),
captureFormatIndex(0),
frame(NULL),
isOpened(false)
{
MFStartup(MF_VERSION);
}
CvCaptureFile_MSMF::~CvCaptureFile_MSMF()
{
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
close();
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
MFShutdown();
}
bool CvCaptureFile_MSMF::open(const char* filename)
{
if (!filename)
return false;
wchar_t* unicodeFileName = new wchar_t[strlen(filename)+1];
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
MultiByteToWideChar(CP_ACP, 0, filename, -1, unicodeFileName, (int)strlen(filename)+1);
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
HRESULT hr = S_OK;
MF_OBJECT_TYPE ObjectType = MF_OBJECT_INVALID;
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
ComPtr<IMFSourceResolver> pSourceResolver = NULL;
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
IUnknown* pUnkSource = NULL;
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
hr = MFCreateSourceResolver(pSourceResolver.GetAddressOf());
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
if (SUCCEEDED(hr))
{
hr = pSourceResolver->CreateObjectFromURL(
unicodeFileName,
MF_RESOLUTION_MEDIASOURCE,
NULL, // Optional property store.
&ObjectType,
&pUnkSource
);
}
// Get the IMFMediaSource from the IUnknown pointer.
if (SUCCEEDED(hr))
{
hr = pUnkSource->QueryInterface(IID_PPV_ARGS(&videoFileSource));
}
SafeRelease(&pUnkSource);
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
if (SUCCEEDED(hr))
{
hr = enumerateCaptureFormats(videoFileSource);
}
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
if (SUCCEEDED(hr))
{
GetProperty method for MSMF VideoCapture implemented.
2013-05-23 00:50:54 +08:00
hr = ImageGrabberThread::CreateInstance(&grabberThread, videoFileSource, (unsigned int)-2, true);
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
}
if (SUCCEEDED(hr))
{
grabberThread->start();
}
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
isOpened = SUCCEEDED(hr);
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
return isOpened;
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
}
void CvCaptureFile_MSMF::close()
{
if (grabberThread)
{
isOpened = false;
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
SetEvent(grabberThread->getImageGrabber()->ig_hFinish);
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
grabberThread->stop();
delete grabberThread;
}
if (videoFileSource)
{
Windows RT related code cleanup. Windows RT sample cleanup. Video I/O code cleanup.
2013-06-24 16:10:18 +08:00
videoFileSource->Shutdown();
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
}
}
bool CvCaptureFile_MSMF::setProperty(int property_id, double value)
{
// image capture properties
// FIXME: implement method in VideoInput back end
Test fixes Calib3d badarg tests fixed. Default exception handling enabled; Highgui Video I/O tests fixed on ARM; Code cleanup.
2013-06-27 14:56:50 +08:00
(void) property_id;
(void) value;
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
return false;
}
double CvCaptureFile_MSMF::getProperty(int property_id)
{
// image format proprrties
switch( property_id )
{
case CV_CAP_PROP_FRAME_WIDTH:
return captureFormats[captureFormatIndex].width;
case CV_CAP_PROP_FRAME_HEIGHT:
return captureFormats[captureFormatIndex].height;
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
case CV_CAP_PROP_FRAME_COUNT:
GetProperty method for MSMF VideoCapture implemented.
2013-05-23 00:50:54 +08:00
{
MFTIME duration;
getSourceDuration(this->videoFileSource, &duration);
double fps = ((double)captureFormats[captureFormatIndex].MF_MT_FRAME_RATE_NUMERATOR) /
((double)captureFormats[captureFormatIndex].MF_MT_FRAME_RATE_DENOMINATOR);
return (double)floor(((double)duration/1e7)*fps+0.5);
}
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
case CV_CAP_PROP_FOURCC:
GetProperty method for MSMF VideoCapture implemented.
2013-05-23 00:50:54 +08:00
return captureFormats[captureFormatIndex].MF_MT_SUBTYPE.Data1;
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
case CV_CAP_PROP_FPS:
GetProperty method for MSMF VideoCapture implemented.
2013-05-23 00:50:54 +08:00
return ((double)captureFormats[captureFormatIndex].MF_MT_FRAME_RATE_NUMERATOR) /
((double)captureFormats[captureFormatIndex].MF_MT_FRAME_RATE_DENOMINATOR);
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
}
Deleted all trailing whitespace.
2013-08-21 20:44:09 +08:00
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
return -1;
}
bool CvCaptureFile_MSMF::grabFrame()
{
Test fixes Calib3d badarg tests fixed. Default exception handling enabled; Highgui Video I/O tests fixed on ARM; Code cleanup.
2013-06-27 14:56:50 +08:00
DWORD waitResult = (DWORD)-1;
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
if (isOpened)
{
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
SetEvent(grabberThread->getImageGrabber()->ig_hFrameGrabbed);
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
HANDLE tmp[] = {grabberThread->getImageGrabber()->ig_hFrameReady, grabberThread->getImageGrabber()->ig_hFinish, 0};
waitResult = WaitForMultipleObjects(2, tmp, FALSE, INFINITE);
}
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
return isOpened && grabberThread->getImageGrabber()->getRawImage()->isNew() && (waitResult == WAIT_OBJECT_0);
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
}
IplImage* CvCaptureFile_MSMF::retrieveFrame(int)
{
unsigned int width = captureFormats[captureFormatIndex].width;
unsigned int height = captureFormats[captureFormatIndex].height;
unsigned int bytes = 3;
GetProperty method for MSMF VideoCapture implemented.
2013-05-23 00:50:54 +08:00
if( !frame || (int)width != frame->width || (int)height != frame->height )
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
{
if (frame)
cvReleaseImage( &frame );
frame = cvCreateImage( cvSize(width,height), 8, 3 );
}
RawImage *RIOut = grabberThread->getImageGrabber()->getRawImage();
unsigned int size = bytes * width * height;
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
bool verticalFlip = captureFormats[captureFormatIndex].MF_MT_DEFAULT_STRIDE < 0;
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
if(RIOut && size == RIOut->getSize())
{
Deleted all trailing whitespace.
2013-08-21 20:44:09 +08:00
videoInput::processPixels(RIOut->getpPixels(), (unsigned char*)frame->imageData, width,
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
height, bytes, false, verticalFlip);
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
}
return frame;
}
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
HRESULT CvCaptureFile_MSMF::enumerateCaptureFormats(IMFMediaSource *pSource)
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
{
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
ComPtr<IMFPresentationDescriptor> pPD = NULL;
ComPtr<IMFStreamDescriptor> pSD = NULL;
ComPtr<IMFMediaTypeHandler> pHandler = NULL;
ComPtr<IMFMediaType> pType = NULL;
HRESULT hr = pSource->CreatePresentationDescriptor(pPD.GetAddressOf());
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
if (FAILED(hr))
{
goto done;
}
BOOL fSelected;
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
hr = pPD->GetStreamDescriptorByIndex(0, &fSelected, pSD.GetAddressOf());
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
if (FAILED(hr))
{
goto done;
}
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
hr = pSD->GetMediaTypeHandler(pHandler.GetAddressOf());
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
if (FAILED(hr))
{
goto done;
}
DWORD cTypes = 0;
hr = pHandler->GetMediaTypeCount(&cTypes);
if (FAILED(hr))
{
goto done;
}
for (DWORD i = 0; i < cTypes; i++)
{
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
hr = pHandler->GetMediaTypeByIndex(i, pType.GetAddressOf());
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
if (FAILED(hr))
{
goto done;
}
Media Foundation-based code refactoring. I* + SafeRelease -> ComPtr.
2013-05-15 21:09:03 +08:00
MediaType MT = FormatReader::Read(pType.Get());
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
captureFormats.push_back(MT);
}
done:
return hr;
}
GetProperty method for MSMF VideoCapture implemented.
2013-05-23 00:50:54 +08:00
HRESULT CvCaptureFile_MSMF::getSourceDuration(IMFMediaSource *pSource, MFTIME *pDuration)
{
*pDuration = 0;
IMFPresentationDescriptor *pPD = NULL;
HRESULT hr = pSource->CreatePresentationDescriptor(&pPD);
if (SUCCEEDED(hr))
{
hr = pPD->GetUINT64(MF_PD_DURATION, (UINT64*)pDuration);
pPD->Release();
}
return hr;
}
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
Experimental MS Media Foundation API support added
2013-04-03 09:01:20 +08:00
CvCapture* cvCreateCameraCapture_MSMF( int index )
{
CvCaptureCAM_MSMF* capture = new CvCaptureCAM_MSMF;
try
{
if( capture->open( index ))
return capture;
}
catch(...)
{
delete capture;
throw;
}
delete capture;
return 0;
}
Media Foundation-based VideoWriter added
2013-05-06 18:36:51 +08:00
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
CvCapture* cvCreateFileCapture_MSMF (const char* filename)
{
CvCaptureFile_MSMF* capture = new CvCaptureFile_MSMF;
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
try
{
if( capture->open(filename) )
return capture;
else
{
delete capture;
return NULL;
}
}
catch(...)
{
delete capture;
throw;
}
VideoCapture for video files implemented. Set and Get methods are not implemented; Camera based video capture is broken due to modifications.
2013-05-14 20:17:34 +08:00
}
Media Foundation-based VideoWriter added
2013-05-06 18:36:51 +08:00
//
//
// Media Foundation-based Video Writer
//
//
class CvVideoWriter_MSMF : public CvVideoWriter
{
public:
CvVideoWriter_MSMF();
virtual ~CvVideoWriter_MSMF();
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
virtual bool open(const char* filename, int fourcc,
double fps, CvSize frameSize, bool isColor);
Media Foundation-based VideoWriter added
2013-05-06 18:36:51 +08:00
virtual void close();
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
virtual bool writeFrame(const IplImage* img);
Media Foundation-based VideoWriter added
2013-05-06 18:36:51 +08:00
private:
UINT32 videoWidth;
UINT32 videoHeight;
double fps;
UINT32 bitRate;
UINT32 frameSize;
GUID encodingFormat;
GUID inputFormat;
DWORD streamIndex;
ComPtr<IMFSinkWriter> sinkWriter;
bool initiated;
LONGLONG rtStart;
UINT64 rtDuration;
HRESULT InitializeSinkWriter(const char* filename);
Media Foundation-based VideoWriter improvements. FourCC parameter handlig added; Smart pointers instead SafeRelease call; Windows RT support (vertical mirroring).
2013-05-06 22:17:53 +08:00
static const GUID FourCC2GUID(int fourcc);
Media Foundation-based VideoWriter added
2013-05-06 18:36:51 +08:00
HRESULT WriteFrame(DWORD *videoFrameBuffer, const LONGLONG& rtStart, const LONGLONG& rtDuration);
};
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
CvVideoWriter_MSMF::CvVideoWriter_MSMF():
initiated(false)
Media Foundation-based VideoWriter added
2013-05-06 18:36:51 +08:00
{
}
CvVideoWriter_MSMF::~CvVideoWriter_MSMF()
{
close();
}
Media Foundation-based VideoWriter improvements. FourCC parameter handlig added; Smart pointers instead SafeRelease call; Windows RT support (vertical mirroring).
2013-05-06 22:17:53 +08:00
const GUID CvVideoWriter_MSMF::FourCC2GUID(int fourcc)
{
switch(fourcc)
{
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
case CV_FOURCC_MACRO('d', 'v', '2', '5'):
Media Foundation-based VideoWriter improvements. FourCC parameter handlig added; Smart pointers instead SafeRelease call; Windows RT support (vertical mirroring).
2013-05-06 22:17:53 +08:00
return MFVideoFormat_DV25; break;
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
case CV_FOURCC_MACRO('d', 'v', '5', '0'):
Media Foundation-based VideoWriter improvements. FourCC parameter handlig added; Smart pointers instead SafeRelease call; Windows RT support (vertical mirroring).
2013-05-06 22:17:53 +08:00
return MFVideoFormat_DV50; break;
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
case CV_FOURCC_MACRO('d', 'v', 'c', ' '):
Media Foundation-based VideoWriter improvements. FourCC parameter handlig added; Smart pointers instead SafeRelease call; Windows RT support (vertical mirroring).
2013-05-06 22:17:53 +08:00
return MFVideoFormat_DVC; break;
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
case CV_FOURCC_MACRO('d', 'v', 'h', '1'):
Media Foundation-based VideoWriter improvements. FourCC parameter handlig added; Smart pointers instead SafeRelease call; Windows RT support (vertical mirroring).
2013-05-06 22:17:53 +08:00
return MFVideoFormat_DVH1; break;
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
case CV_FOURCC_MACRO('d', 'v', 'h', 'd'):
Media Foundation-based VideoWriter improvements. FourCC parameter handlig added; Smart pointers instead SafeRelease call; Windows RT support (vertical mirroring).
2013-05-06 22:17:53 +08:00
return MFVideoFormat_DVHD; break;
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
case CV_FOURCC_MACRO('d', 'v', 's', 'd'):
Media Foundation-based VideoWriter improvements. FourCC parameter handlig added; Smart pointers instead SafeRelease call; Windows RT support (vertical mirroring).
2013-05-06 22:17:53 +08:00
return MFVideoFormat_DVSD; break;
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
case CV_FOURCC_MACRO('d', 'v', 's', 'l'):
Media Foundation-based VideoWriter improvements. FourCC parameter handlig added; Smart pointers instead SafeRelease call; Windows RT support (vertical mirroring).
2013-05-06 22:17:53 +08:00
return MFVideoFormat_DVSL; break;
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
case CV_FOURCC_MACRO('H', '2', '6', '3'):
Media Foundation-based VideoWriter improvements. FourCC parameter handlig added; Smart pointers instead SafeRelease call; Windows RT support (vertical mirroring).
2013-05-06 22:17:53 +08:00
return MFVideoFormat_H263; break;
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
case CV_FOURCC_MACRO('H', '2', '6', '4'):
Media Foundation-based VideoWriter improvements. FourCC parameter handlig added; Smart pointers instead SafeRelease call; Windows RT support (vertical mirroring).
2013-05-06 22:17:53 +08:00
return MFVideoFormat_H264; break;
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
case CV_FOURCC_MACRO('M', '4', 'S', '2'):
Media Foundation-based VideoWriter improvements. FourCC parameter handlig added; Smart pointers instead SafeRelease call; Windows RT support (vertical mirroring).
2013-05-06 22:17:53 +08:00
return MFVideoFormat_M4S2; break;
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
case CV_FOURCC_MACRO('M', 'J', 'P', 'G'):
Media Foundation-based VideoWriter improvements. FourCC parameter handlig added; Smart pointers instead SafeRelease call; Windows RT support (vertical mirroring).
2013-05-06 22:17:53 +08:00
return MFVideoFormat_MJPG; break;
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
case CV_FOURCC_MACRO('M', 'P', '4', '3'):
Media Foundation-based VideoWriter improvements. FourCC parameter handlig added; Smart pointers instead SafeRelease call; Windows RT support (vertical mirroring).
2013-05-06 22:17:53 +08:00
return MFVideoFormat_MP43; break;
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
case CV_FOURCC_MACRO('M', 'P', '4', 'S'):
Media Foundation-based VideoWriter improvements. FourCC parameter handlig added; Smart pointers instead SafeRelease call; Windows RT support (vertical mirroring).
2013-05-06 22:17:53 +08:00
return MFVideoFormat_MP4S; break;
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
case CV_FOURCC_MACRO('M', 'P', '4', 'V'):
Media Foundation-based VideoWriter improvements. FourCC parameter handlig added; Smart pointers instead SafeRelease call; Windows RT support (vertical mirroring).
2013-05-06 22:17:53 +08:00
return MFVideoFormat_MP4V; break;
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
case CV_FOURCC_MACRO('M', 'P', 'G', '1'):
Media Foundation-based VideoWriter improvements. FourCC parameter handlig added; Smart pointers instead SafeRelease call; Windows RT support (vertical mirroring).
2013-05-06 22:17:53 +08:00
return MFVideoFormat_MPG1; break;
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
case CV_FOURCC_MACRO('M', 'S', 'S', '1'):
Media Foundation-based VideoWriter improvements. FourCC parameter handlig added; Smart pointers instead SafeRelease call; Windows RT support (vertical mirroring).
2013-05-06 22:17:53 +08:00
return MFVideoFormat_MSS1; break;
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
case CV_FOURCC_MACRO('M', 'S', 'S', '2'):
Media Foundation-based VideoWriter improvements. FourCC parameter handlig added; Smart pointers instead SafeRelease call; Windows RT support (vertical mirroring).
2013-05-06 22:17:53 +08:00
return MFVideoFormat_MSS2; break;
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
case CV_FOURCC_MACRO('W', 'M', 'V', '1'):
Media Foundation-based VideoWriter improvements. FourCC parameter handlig added; Smart pointers instead SafeRelease call; Windows RT support (vertical mirroring).
2013-05-06 22:17:53 +08:00
return MFVideoFormat_WMV1; break;
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
case CV_FOURCC_MACRO('W', 'M', 'V', '2'):
Media Foundation-based VideoWriter improvements. FourCC parameter handlig added; Smart pointers instead SafeRelease call; Windows RT support (vertical mirroring).
2013-05-06 22:17:53 +08:00
return MFVideoFormat_WMV2; break;
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
case CV_FOURCC_MACRO('W', 'M', 'V', '3'):
Media Foundation-based VideoWriter improvements. FourCC parameter handlig added; Smart pointers instead SafeRelease call; Windows RT support (vertical mirroring).
2013-05-06 22:17:53 +08:00
return MFVideoFormat_WMV3; break;
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
case CV_FOURCC_MACRO('W', 'V', 'C', '1'):
Media Foundation-based VideoWriter improvements. FourCC parameter handlig added; Smart pointers instead SafeRelease call; Windows RT support (vertical mirroring).
2013-05-06 22:17:53 +08:00
return MFVideoFormat_WVC1; break;
default:
return MFVideoFormat_H264;
}
}
Media Foundation-based VideoWriter added
2013-05-06 18:36:51 +08:00
bool CvVideoWriter_MSMF::open( const char* filename, int fourcc,
GetProperty method for MSMF VideoCapture implemented.
2013-05-23 00:50:54 +08:00
double _fps, CvSize frameSize, bool /*isColor*/ )
Media Foundation-based VideoWriter added
2013-05-06 18:36:51 +08:00
{
videoWidth = frameSize.width;
videoHeight = frameSize.height;
fps = _fps;
GetProperty method for MSMF VideoCapture implemented.
2013-05-23 00:50:54 +08:00
bitRate = (UINT32)fps*videoWidth*videoHeight; // 1-bit per pixel
Media Foundation-based VideoWriter improvements. FourCC parameter handlig added; Smart pointers instead SafeRelease call; Windows RT support (vertical mirroring).
2013-05-06 22:17:53 +08:00
encodingFormat = FourCC2GUID(fourcc);
Media Foundation-based VideoWriter added
2013-05-06 18:36:51 +08:00
inputFormat = MFVideoFormat_RGB32;
HRESULT hr = CoInitializeEx(NULL, COINIT_APARTMENTTHREADED);
if (SUCCEEDED(hr))
{
hr = MFStartup(MF_VERSION);
if (SUCCEEDED(hr))
{
hr = InitializeSinkWriter(filename);
if (SUCCEEDED(hr))
{
initiated = true;
rtStart = 0;
Media Foundation-based VideoWriter improvements. FourCC parameter handlig added; Smart pointers instead SafeRelease call; Windows RT support (vertical mirroring).
2013-05-06 22:17:53 +08:00
MFFrameRateToAverageTimePerFrame((UINT32)fps, 1, &rtDuration);
Media Foundation-based VideoWriter added
2013-05-06 18:36:51 +08:00
}
}
}
return SUCCEEDED(hr);
}
void CvVideoWriter_MSMF::close()
{
if (!initiated)
{
return;
}
initiated = false;
Test fixes Calib3d badarg tests fixed. Default exception handling enabled; Highgui Video I/O tests fixed on ARM; Code cleanup.
2013-06-27 14:56:50 +08:00
sinkWriter->Finalize();
Media Foundation-based VideoWriter added
2013-05-06 18:36:51 +08:00
MFShutdown();
}
bool CvVideoWriter_MSMF::writeFrame(const IplImage* img)
{
if (!img)
return false;
Media Foundation-based VideoWriter improvements. FourCC parameter handlig added; Smart pointers instead SafeRelease call; Windows RT support (vertical mirroring).
2013-05-06 22:17:53 +08:00
int length = img->width * img->height * 4;
Media Foundation-based VideoWriter added
2013-05-06 18:36:51 +08:00
DWORD* target = new DWORD[length];
for (int rowIdx = 0; rowIdx < img->height; rowIdx++)
{
char* rowStart = img->imageData + rowIdx*img->widthStep;
for (int colIdx = 0; colIdx < img->width; colIdx++)
{
BYTE b = rowStart[colIdx * img->nChannels + 0];
BYTE g = rowStart[colIdx * img->nChannels + 1];
BYTE r = rowStart[colIdx * img->nChannels + 2];
target[rowIdx*img->width+colIdx] = (r << 16) + (g << 8) + b;
}
}
// Send frame to the sink writer.
HRESULT hr = WriteFrame(target, rtStart, rtDuration);
if (FAILED(hr))
{
delete[] target;
return false;
}
rtStart += rtDuration;
delete[] target;
return true;
}
HRESULT CvVideoWriter_MSMF::InitializeSinkWriter(const char* filename)
{
ComPtr<IMFAttributes> spAttr;
ComPtr<IMFMediaType> mediaTypeOut;
ComPtr<IMFMediaType> mediaTypeIn;
ComPtr<IMFByteStream> spByteStream;
MFCreateAttributes(&spAttr, 10);
spAttr->SetUINT32(MF_READWRITE_ENABLE_HARDWARE_TRANSFORMS, true);
wchar_t* unicodeFileName = new wchar_t[strlen(filename)+1];
Update cap_msmf.cpp Add support for WinRT in the MF capture framework by removing the disallowed calls to enumerate devices and create a sample grabber sink and adding framework for the MediaCapture interface and a custom sink which interfaces with the sample grabber callback interface. The change requires discussion for making it completely functional as redundancy is required given that if the source is a video file, the old code pathways must be used. Otherwise all IMFMediaSession, IMFMediaSource, and IMFActivate code must use a MediaCapture code path and all sink code must use the CMediaSink custom sink. Support for the custom sink is extended to non-WinRT not for compatibility as Windows Vista client is a minimum regardless, but because it offers more flexibility, could be faster and is able to be used as an optionally different code path during sink creation based on a future configuration parameter. My discussion and proposal to finish this change: Devices are so easily enumerated through WinRT Windows.Devices namespace that wrapping the calls in a library is quite a chore for little benefit though to get the various modes and formats could still be a worthwhile project. For now conditional compilation to remove videodevices and any offending non-video file related activity in videodevice. In my opinion, this is a different , far less fundamental and important change which can possibly be done as a future project and also much more easily implemented in C++/CX. ImageGrabber has the IMFSampleGrabberSinkCallback replaced with a base class (SharedSampleGrabber) which also be is base class for ImageGrabberRT. This change is necessary as the custom sink does not require a thread to pump events which is done through MediaCapture already. IMFSampleGrabberSinkCallback is the common element between both models and that piece can be shared. Initializing the new ImageGrabberRT is as simple as passing an already initialized MediaCapture object and any video format/encoding parameters. The concurrency event is necessary to wait for completion and is the way the underlying, IAsyncAction wrappers in the task library work as well. Native WIN32 event objects would be an option if HAVE_CONCURRENCY is not defined. I could even imagine doing it with sleep/thread yield and InterlockedCompareExchange yet I am not enthusiastic about that approach either. Since there is a specific compiler HAVE_ for concurrency, I do not like pulling it in though I think for WinRT it is safe to say we will always have it available though should probably conditionally compile with the Interlocked option as WIN32 events would require HAVE_WIN32. It looks like C++/CX cannot be used for the IMediaExtension sink (which should not be a problem) as using COM objects requires WRL and though deriving from IMediaExtension can be done, there is little purpose without COM. Objects from C++/CX can be swapped to interact with objects from native C++ as Inspectable* can reinterpret_cast to the ref object IInspectable^ and vice-versa. A solution to the COM class with C++/CX would be great so we could have dual support. Also without #define for every WRL object in use, the code will get quite muddy given that the */^ would need to be ifdef'd everywhere. Update cap_msmf.cpp Fixed bugs and completed the change. I believe the new classes need to be moved to a header file as the file has become to large and more classes need to be added for handling all the asynchronous problems (one wrapping IAsyncAction in a task and another for making a task out of IAsyncAction). Unfortunately, blocking on the UI thread is not an option in WinRT so a synchronous architecture is considered "illegal" by Microsoft's standards even if implementable (C++/CX ppltasks library throws errors if you try it). Worse, either by design or a bug in the MF MediaCapture class with Custom Sinks causes a crash if stop/start previewing without reinitializing (spPreferredPreviewMediaType is fatally nulled). After decompiling Windows.Media.dll, I worked around this in my own projects by using an activate-able custom sink ID which strangely assigns 1 to this pointer allowing it to be reinitialized in what can only be described as a hack by Microsoft. This would add additional overhead to the project to implement especially for static libraries as it requires IDL/DLL exporting followed by manifest declaration. Better to document that it is not supported. Furthermore, an additional class for IMFAttributes should be implemented to make clean architecture for passing around attributes as opposed to directly calling non-COM interface calls on the objects and making use of SetProperties which would also be a set up for an object that uses the RuntimeClass activation ID. The remaining changes are not difficult and will be complete soon along with debug tracing messages. Update cap_msmf.cpp Create cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update cap_msmf.cpp Update cap_msmf.h Update and rename cap_msmf.h to cap_msmf.hpp Update cap_msmf.cpp Update CMakeLists.txt Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Successful test - samples are grabbed Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Update cap_msmf.cpp Update cap_msmf.hpp Create ppltasks_winrt.h Update cap_msmf.hpp Update cap_msmf.hpp Update cap_msmf.cpp Update ppltasks_winrt.h Library updated and cleaned up with comments, marshaling, exceptions and linker settings Update ppltasks_winrt.h Fixed trailing whitespace
2013-12-14 16:53:30 +08:00
MultiByteToWideChar(CP_ACP, 0, filename, -1, unicodeFileName, (int)strlen(filename)+1);
Media Foundation-based VideoWriter added
2013-05-06 18:36:51 +08:00
HRESULT hr = MFCreateSinkWriterFromURL(unicodeFileName, NULL, spAttr.Get(), &sinkWriter);
delete[] unicodeFileName;
// Set the output media type.
if (SUCCEEDED(hr))
{
hr = MFCreateMediaType(&mediaTypeOut);
}
if (SUCCEEDED(hr))
{
hr = mediaTypeOut->SetGUID(MF_MT_MAJOR_TYPE, MFMediaType_Video);
}
if (SUCCEEDED(hr))
{
hr = mediaTypeOut->SetGUID(MF_MT_SUBTYPE, encodingFormat);
}
if (SUCCEEDED(hr))
{
hr = mediaTypeOut->SetUINT32(MF_MT_AVG_BITRATE, bitRate);
}
if (SUCCEEDED(hr))
{
hr = mediaTypeOut->SetUINT32(MF_MT_INTERLACE_MODE, MFVideoInterlace_Progressive);
}
if (SUCCEEDED(hr))
{
hr = MFSetAttributeSize(mediaTypeOut.Get(), MF_MT_FRAME_SIZE, videoWidth, videoHeight);
}
if (SUCCEEDED(hr))
{
Media Foundation-based VideoWriter improvements. FourCC parameter handlig added; Smart pointers instead SafeRelease call; Windows RT support (vertical mirroring).
2013-05-06 22:17:53 +08:00
hr = MFSetAttributeRatio(mediaTypeOut.Get(), MF_MT_FRAME_RATE, (UINT32)fps, 1);
Media Foundation-based VideoWriter added
2013-05-06 18:36:51 +08:00
}
if (SUCCEEDED(hr))
{
hr = MFSetAttributeRatio(mediaTypeOut.Get(), MF_MT_PIXEL_ASPECT_RATIO, 1, 1);
}
Multiple Media Foundation video i/o fixes. Video i/o tests enabled for media foundation; Negative stride support added to VideoCapture; Error handling improved, dead lock in case of playback error fixed; Some code refacotring done.
2013-05-19 04:04:31 +08:00
Media Foundation-based VideoWriter added
2013-05-06 18:36:51 +08:00
if (SUCCEEDED(hr))
{
hr = sinkWriter->AddStream(mediaTypeOut.Get(), &streamIndex);
}
// Set the input media type.
if (SUCCEEDED(hr))
{
hr = MFCreateMediaType(&mediaTypeIn);
}
if (SUCCEEDED(hr))
{
hr = mediaTypeIn->SetGUID(MF_MT_MAJOR_TYPE, MFMediaType_Video);
}
if (SUCCEEDED(hr))
{
hr = mediaTypeIn->SetGUID(MF_MT_SUBTYPE, inputFormat);
}
if (SUCCEEDED(hr))
{
hr = mediaTypeIn->SetUINT32(MF_MT_INTERLACE_MODE, MFVideoInterlace_Progressive);
}
if (SUCCEEDED(hr))
{
hr = MFSetAttributeSize(mediaTypeIn.Get(), MF_MT_FRAME_SIZE, videoWidth, videoHeight);
}
if (SUCCEEDED(hr))
{
Media Foundation-based VideoWriter improvements. FourCC parameter handlig added; Smart pointers instead SafeRelease call; Windows RT support (vertical mirroring).
2013-05-06 22:17:53 +08:00
hr = MFSetAttributeRatio(mediaTypeIn.Get(), MF_MT_FRAME_RATE, (UINT32)fps, 1);
Media Foundation-based VideoWriter added
2013-05-06 18:36:51 +08:00
}
if (SUCCEEDED(hr))
{
hr = MFSetAttributeRatio(mediaTypeIn.Get(), MF_MT_PIXEL_ASPECT_RATIO, 1, 1);
}
Media foundation video i/o fixes. Bug in Video for Windows capture init fixed; Media Foundation based capture finalization fixed; Highgui tests for video i/o updated.
2013-05-22 19:21:23 +08:00
if (SUCCEEDED(hr))
Media Foundation-based VideoWriter added
2013-05-06 18:36:51 +08:00
{
hr = sinkWriter->SetInputMediaType(streamIndex, mediaTypeIn.Get(), NULL);
}
// Tell the sink writer to start accepting data.
if (SUCCEEDED(hr))
{
hr = sinkWriter->BeginWriting();
}
return hr;
}
HRESULT CvVideoWriter_MSMF::WriteFrame(DWORD *videoFrameBuffer, const LONGLONG& Start, const LONGLONG& Duration)
{
Media Foundation-based VideoWriter improvements. FourCC parameter handlig added; Smart pointers instead SafeRelease call; Windows RT support (vertical mirroring).
2013-05-06 22:17:53 +08:00
ComPtr<IMFSample> sample;
ComPtr<IMFMediaBuffer> buffer;
Media Foundation-based VideoWriter added
2013-05-06 18:36:51 +08:00
const LONG cbWidth = 4 * videoWidth;
const DWORD cbBuffer = cbWidth * videoHeight;
BYTE *pData = NULL;
// Create a new memory buffer.
HRESULT hr = MFCreateMemoryBuffer(cbBuffer, &buffer);
// Lock the buffer and copy the video frame to the buffer.
if (SUCCEEDED(hr))
{
hr = buffer->Lock(&pData, NULL, NULL);
}
if (SUCCEEDED(hr))
{
Test fixes Calib3d badarg tests fixed. Default exception handling enabled; Highgui Video I/O tests fixed on ARM; Code cleanup.
2013-06-27 14:56:50 +08:00
#if defined(_M_ARM)
hr = MFCopyImage(
pData, // Destination buffer.
-cbWidth, // Destination stride.
(BYTE*)videoFrameBuffer, // First row in source image.
cbWidth, // Source stride.
cbWidth, // Image width in bytes.
videoHeight // Image height in pixels.
);
#else
Media Foundation-based VideoWriter added
2013-05-06 18:36:51 +08:00
hr = MFCopyImage(
pData, // Destination buffer.
cbWidth, // Destination stride.
(BYTE*)videoFrameBuffer, // First row in source image.
cbWidth, // Source stride.
cbWidth, // Image width in bytes.
videoHeight // Image height in pixels.
);
Test fixes Calib3d badarg tests fixed. Default exception handling enabled; Highgui Video I/O tests fixed on ARM; Code cleanup.
2013-06-27 14:56:50 +08:00
#endif
Media Foundation-based VideoWriter added
2013-05-06 18:36:51 +08:00
}
if (buffer)
{
buffer->Unlock();
}
// Set the data length of the buffer.
if (SUCCEEDED(hr))
{
hr = buffer->SetCurrentLength(cbBuffer);
}
// Create a media sample and add the buffer to the sample.
if (SUCCEEDED(hr))
{
hr = MFCreateSample(&sample);
}
if (SUCCEEDED(hr))
{
Media Foundation-based VideoWriter improvements. FourCC parameter handlig added; Smart pointers instead SafeRelease call; Windows RT support (vertical mirroring).
2013-05-06 22:17:53 +08:00
hr = sample->AddBuffer(buffer.Get());
Media Foundation-based VideoWriter added
2013-05-06 18:36:51 +08:00
}
// Set the time stamp and the duration.
if (SUCCEEDED(hr))
{
hr = sample->SetSampleTime(Start);
}
if (SUCCEEDED(hr))
{
hr = sample->SetSampleDuration(Duration);
}
// Send the sample to the Sink Writer.
if (SUCCEEDED(hr))
{
Media Foundation-based VideoWriter improvements. FourCC parameter handlig added; Smart pointers instead SafeRelease call; Windows RT support (vertical mirroring).
2013-05-06 22:17:53 +08:00
hr = sinkWriter->WriteSample(streamIndex, sample.Get());
Media Foundation-based VideoWriter added
2013-05-06 18:36:51 +08:00
}
return hr;
}
CvVideoWriter* cvCreateVideoWriter_MSMF( const char* filename, int fourcc,
double fps, CvSize frameSize, int isColor )
{
CvVideoWriter_MSMF* writer = new CvVideoWriter_MSMF;
if( writer->open( filename, fourcc, fps, frameSize, isColor != 0 ))
return writer;
delete writer;
return NULL;
}
Normalized file endings.
2013-08-21 21:26:54 +08:00
#endif
Reference in New Issue Copy Permalink