Merge pull request #16586 from themechanicalcoder:video-psnr

* add  python version of video-input-psnr-ssim

* remove ret

* documentation changes

* added link for python file

* command line argument

doc/tutorials/videoio/video-input-psnr-ssim/video_input_psnr_ssim.markdown

@@ -25,7 +25,13 @@ version of it ](https://github.com/opencv/opencv/tree/3.4/samples/data/Megamind_
 You may also find the source code and these video file in the
 `samples/data` folder of the OpenCV source library.
 
+@add_toggle_cpp
 @include cpp/tutorial_code/videoio/video-input-psnr-ssim/video-input-psnr-ssim.cpp
+@end_toggle
+
+@add_toggle_python
+@include samples/python/tutorial_code/videoio/video-input-psnr-ssim.py
+@end_toggle
 
 How to read a video stream (online-camera or offline-file)?
 -----------------------------------------------------------
@@ -139,28 +145,15 @@ an invalid divide by zero operation in the PSNR formula. In this case the PSNR i
 we'll need to handle this case separately. The transition to a logarithmic scale is made because the
 pixel values have a very wide dynamic range. All this translated to OpenCV and a C++ function looks
 like:
-@code{.cpp}
-double getPSNR(const Mat& I1, const Mat& I2)
-{
- Mat s1;
- absdiff(I1, I2, s1);       // |I1 - I2|
- s1.convertTo(s1, CV_32F);  // cannot make a square on 8 bits
- s1 = s1.mul(s1);           // |I1 - I2|^2
 
- Scalar s = sum(s1);        // sum elements per channel
+@add_toggle_cpp
+@include cpp/tutorial_code/videoio/video-input-psnr-ssim/video-input-psnr-ssim.cpp get-psnr
+@end_toggle
 
- double sse = s.val[0] + s.val[1] + s.val[2]; // sum channels
+@add_toggle_python
+@include samples/python/tutorial_code/videoio/video-input-psnr-ssim.py get-psnr
+@end_toggle
 
- if( sse <= 1e-10) // for small values return zero
-     return 0;
- else
- {
-     double  mse =sse /(double)(I1.channels() * I1.total());
-     double psnr = 10.0*log10((255*255)/mse);
-     return psnr;
- }
-}
-@endcode
 Typically result values are anywhere between 30 and 50 for video compression, where higher is
 better. If the images significantly differ you'll get much lower ones like 15 and so. This
 similarity check is easy and fast to calculate, however in practice it may turn out somewhat
@@ -176,60 +169,14 @@ implementation below.
     Simoncelli, "Image quality assessment: From error visibility to structural similarity," IEEE
     Transactions on Image Processing, vol. 13, no. 4, pp. 600-612, Apr. 2004." article.
 
-@code{.cpp}
+@add_toggle_cpp
-Scalar getMSSIM( const Mat& i1, const Mat& i2)
+@include cpp/tutorial_code/videoio/video-input-psnr-ssim/video-input-psnr-ssim.cpp get-mssim
-{
+@end_toggle
- const double C1 = 6.5025, C2 = 58.5225;
- /***************************** INITS **********************************/
- int d     = CV_32F;
 
- Mat I1, I2;
+@add_toggle_python
- i1.convertTo(I1, d);           // cannot calculate on one byte large values
+@include samples/python/tutorial_code/videoio/video-input-psnr-ssim.py get-mssim
- i2.convertTo(I2, d);
+@end_toggle
 
- Mat I2_2   = I2.mul(I2);        // I2^2
- Mat I1_2   = I1.mul(I1);        // I1^2
- Mat I1_I2  = I1.mul(I2);        // I1 * I2
-
- /***********************PRELIMINARY COMPUTING ******************************/
-
- Mat mu1, mu2;   //
- GaussianBlur(I1, mu1, Size(11, 11), 1.5);
- GaussianBlur(I2, mu2, Size(11, 11), 1.5);
-
- Mat mu1_2   =   mu1.mul(mu1);
- Mat mu2_2   =   mu2.mul(mu2);
- Mat mu1_mu2 =   mu1.mul(mu2);
-
- Mat sigma1_2, sigma2_2, sigma12;
-
- GaussianBlur(I1_2, sigma1_2, Size(11, 11), 1.5);
- sigma1_2 -= mu1_2;
-
- GaussianBlur(I2_2, sigma2_2, Size(11, 11), 1.5);
- sigma2_2 -= mu2_2;
-
- GaussianBlur(I1_I2, sigma12, Size(11, 11), 1.5);
- sigma12 -= mu1_mu2;
-
- ///////////////////////////////// FORMULA ////////////////////////////////
- Mat t1, t2, t3;
-
- t1 = 2 * mu1_mu2 + C1;
- t2 = 2 * sigma12 + C2;
- t3 = t1.mul(t2);              // t3 = ((2*mu1_mu2 + C1).*(2*sigma12 + C2))
-
- t1 = mu1_2 + mu2_2 + C1;
- t2 = sigma1_2 + sigma2_2 + C2;
- t1 = t1.mul(t2);               // t1 =((mu1_2 + mu2_2 + C1).*(sigma1_2 + sigma2_2 + C2))
-
- Mat ssim_map;
- divide(t3, t1, ssim_map);      // ssim_map =  t3./t1;
-
- Scalar mssim = mean( ssim_map ); // mssim = average of ssim map
- return mssim;
-}
-@endcode
 This will return a similarity index for each channel of the image. This value is between zero and
 one, where one corresponds to perfect fit. Unfortunately, the many Gaussian blurring is quite
 costly, so while the PSNR may work in a real time like environment (24 frame per second) this will

samples/cpp/tutorial_code/videoio/video-input-psnr-ssim/video-input-psnr-ssim.cpp

@@ -132,6 +132,7 @@ int main(int argc, char *argv[])
     return 0;
 }
 
+// ![get-psnr]
 double getPSNR(const Mat& I1, const Mat& I2)
 {
     Mat s1;
@@ -152,6 +153,9 @@ double getPSNR(const Mat& I1, const Mat& I2)
         return psnr;
     }
 }
+// ![get-psnr]
+
+// ![get-mssim]
 
 Scalar getMSSIM( const Mat& i1, const Mat& i2)
 {
@@ -205,3 +209,4 @@ Scalar getMSSIM( const Mat& i1, const Mat& i2)
     Scalar mssim = mean(ssim_map);   // mssim = average of ssim map
     return mssim;
 }
+// ![get-mssim]

samples/python/tutorial_code/videoio/video-input-psnr-ssim.py

@@ -0,0 +1,148 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+
+# Python 2/3 compatibility
+from __future__ import print_function
+
+import numpy as np
+import cv2 as cv
+import argparse
+import sys
+
+# [get-psnr]
+def getPSNR(I1, I2):
+    s1 = cv.absdiff(I1, I2) #|I1 - I2|
+    s1 = np.float32(s1)     # cannot make a square on 8 bits
+    s1 = s1 * s1            # |I1 - I2|^2
+    sse = s1.sum()          # sum elements per channel
+    if sse <= 1e-10:        # sum channels
+        return 0            # for small values return zero
+    else:
+        shape = I1.shape
+        mse = 1.0 * sse / (shape[0] * shape[1] * shape[2])
+        psnr = 10.0 * np.log10((255 * 255) / mse)
+        return psnr
+# [get-psnr]
+
+# [get-mssim]
+def getMSSISM(i1, i2):
+    C1 = 6.5025
+    C2 = 58.5225
+    # INITS
+
+    I1 = np.float32(i1) # cannot calculate on one byte large values
+    I2 = np.float32(i2)
+
+    I2_2 = I2 * I2 # I2^2
+    I1_2 = I1 * I1 # I1^2
+    I1_I2 = I1 * I2 # I1 * I2
+    # END INITS
+
+    # PRELIMINARY COMPUTING
+    mu1 = cv.GaussianBlur(I1, (11, 11), 1.5)
+    mu2 = cv.GaussianBlur(I2, (11, 11), 1.5)
+
+    mu1_2 = mu1 * mu1
+    mu2_2 = mu2 * mu2
+    mu1_mu2 = mu1 * mu2
+
+    sigma1_2 = cv.GaussianBlur(I1_2, (11, 11), 1.5)
+    sigma1_2 -= mu1_2
+
+    sigma2_2 = cv.GaussianBlur(I2_2, (11, 11), 1.5)
+    sigma2_2 -= mu2_2
+
+    sigma12 = cv.GaussianBlur(I1_I2, (11, 11), 1.5)
+    sigma12 -= mu1_mu2
+
+    t1 = 2 * mu1_mu2 + C1
+    t2 = 2 * sigma12 + C2
+    t3 = t1 * t2                    # t3 = ((2*mu1_mu2 + C1).*(2*sigma12 + C2))
+
+    t1 = mu1_2 + mu2_2 + C1
+    t2 = sigma1_2 + sigma2_2 + C2
+    t1 = t1 * t2                    # t1 =((mu1_2 + mu2_2 + C1).*(sigma1_2 + sigma2_2 + C2))
+
+    ssim_map = cv.divide(t3, t1)    # ssim_map =  t3./t1;
+
+    mssim = cv.mean(ssim_map)       # mssim = average of ssim map
+    return mssim
+# [get-mssim]
+
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument("-d", "--delay", type=int, default=30, help=" Time delay")
+    parser.add_argument("-v", "--psnrtriggervalue", type=int, default=30, help="PSNR Trigger Value")
+    parser.add_argument("-r", "--ref", type=str, default="Megamind.avi", help="Path to reference video")
+    parser.add_argument("-t", "--undertest", type=str, default="Megamind_bugy.avi",
+                        help="Path to the video to be tested")
+    args = parser.parse_args()
+
+    sourceReference = args.ref
+    sourceCompareWith = args.undertest
+    delay = args.delay
+    psnrTriggerValue = args.psnrtriggervalue
+
+    framenum = -1 # Frame counter
+
+    captRefrnc = cv.VideoCapture(sourceReference)
+    captUndTst = cv.VideoCapture(sourceCompareWith)
+
+    if not captRefrnc.isOpened():
+        print("Could not open the reference " + sourceReference)
+        sys.exit(-1)
+    if not captUndTst.isOpened():
+        print("Could not open case test " + sourceCompareWith)
+        sys.exit(-1)
+
+    refS = (int(captRefrnc.get(cv.CAP_PROP_FRAME_WIDTH)), int(captRefrnc.get(cv.CAP_PROP_FRAME_HEIGHT)))
+    uTSi = (int(captUndTst.get(cv.CAP_PROP_FRAME_WIDTH)), int(captUndTst.get(cv.CAP_PROP_FRAME_HEIGHT)))
+
+    if refS != uTSi:
+        print("Inputs have different size!!! Closing.")
+        sys.exit(-1)
+
+    WIN_UT = "Under Test"
+    WIN_RF = "Reference"
+
+    cv.namedWindow(WIN_RF, cv.WINDOW_AUTOSIZE)
+    cv.namedWindow(WIN_UT, cv.WINDOW_AUTOSIZE)
+    cv.moveWindow(WIN_RF, 400, 0) #750,  2 (bernat =0)
+    cv.moveWindow(WIN_UT, refS[0], 0) #1500, 2
+
+    print("Reference frame resolution: Width={} Height={} of nr#: {}".format(refS[0], refS[1],
+                                                                             captRefrnc.get(cv.CAP_PROP_FRAME_COUNT)))
+    print("PSNR trigger value {}".format(psnrTriggerValue))
+
+    while True: # Show the image captured in the window and repeat
+        _, frameReference = captRefrnc.read()
+        _, frameUnderTest = captUndTst.read()
+
+        if frameReference is None or frameUnderTest is None:
+            print(" < < <  Game over!  > > > ")
+            break
+
+        framenum += 1
+        psnrv = getPSNR(frameReference, frameUnderTest)
+        print("Frame: {}# {}dB".format(framenum, round(psnrv, 3)), end=" ")
+
+        if (psnrv < psnrTriggerValue and psnrv):
+            mssimv = getMSSISM(frameReference, frameUnderTest)
+            print("MSSISM: R {}% G {}% B {}%".format(round(mssimv[2] * 100, 2), round(mssimv[1] * 100, 2),
+                                                     round(mssimv[0] * 100, 2)), end=" ")
+
+        print()
+
+        cv.imshow(WIN_RF, frameReference)
+        cv.imshow(WIN_UT, frameUnderTest)
+
+        k = cv.waitKey(delay)
+        if k == 27:
+            break
+
+    sys.exit(0)
+
+
+if __name__ == "__main__":
+    main()