2014-11-27 20:39:05 +08:00
|
|
|
Basic Thresholding Operations {#tutorial_threshold}
|
|
|
|
=============================
|
|
|
|
|
|
|
|
Goal
|
|
|
|
----
|
|
|
|
|
|
|
|
In this tutorial you will learn how to:
|
|
|
|
|
|
|
|
- Perform basic thresholding operations using OpenCV function @ref cv::threshold
|
|
|
|
|
|
|
|
Cool Theory
|
|
|
|
-----------
|
|
|
|
|
|
|
|
@note The explanation below belongs to the book **Learning OpenCV** by Bradski and Kaehler. What is
|
2014-11-28 21:21:28 +08:00
|
|
|
|
|
|
|
Thresholding?
|
|
|
|
-------------
|
2014-11-27 20:39:05 +08:00
|
|
|
|
|
|
|
- The simplest segmentation method
|
|
|
|
- Application example: Separate out regions of an image corresponding to objects which we want to
|
|
|
|
analyze. This separation is based on the variation of intensity between the object pixels and
|
|
|
|
the background pixels.
|
|
|
|
- To differentiate the pixels we are interested in from the rest (which will eventually be
|
|
|
|
rejected), we perform a comparison of each pixel intensity value with respect to a *threshold*
|
|
|
|
(determined according to the problem to solve).
|
|
|
|
- Once we have separated properly the important pixels, we can set them with a determined value to
|
|
|
|
identify them (i.e. we can assign them a value of \f$0\f$ (black), \f$255\f$ (white) or any value that
|
|
|
|
suits your needs).
|
|
|
|
|
2014-11-28 21:21:28 +08:00
|
|
|
![](images/Threshold_Tutorial_Theory_Example.jpg)
|
2014-11-27 20:39:05 +08:00
|
|
|
|
|
|
|
### Types of Thresholding
|
|
|
|
|
|
|
|
- OpenCV offers the function @ref cv::threshold to perform thresholding operations.
|
|
|
|
- We can effectuate \f$5\f$ types of Thresholding operations with this function. We will explain them
|
|
|
|
in the following subsections.
|
|
|
|
- To illustrate how these thresholding processes work, let's consider that we have a source image
|
|
|
|
with pixels with intensity values \f$src(x,y)\f$. The plot below depicts this. The horizontal blue
|
|
|
|
line represents the threshold \f$thresh\f$ (fixed).
|
|
|
|
|
2014-11-28 21:21:28 +08:00
|
|
|
![](images/Threshold_Tutorial_Theory_Base_Figure.png)
|
2014-11-27 20:39:05 +08:00
|
|
|
|
|
|
|
#### Threshold Binary
|
|
|
|
|
|
|
|
- This thresholding operation can be expressed as:
|
|
|
|
|
|
|
|
\f[\texttt{dst} (x,y) = \fork{\texttt{maxVal}}{if \(\texttt{src}(x,y) > \texttt{thresh}\)}{0}{otherwise}\f]
|
|
|
|
|
|
|
|
- So, if the intensity of the pixel \f$src(x,y)\f$ is higher than \f$thresh\f$, then the new pixel
|
|
|
|
intensity is set to a \f$MaxVal\f$. Otherwise, the pixels are set to \f$0\f$.
|
|
|
|
|
2014-11-28 21:21:28 +08:00
|
|
|
![](images/Threshold_Tutorial_Theory_Binary.png)
|
2014-11-27 20:39:05 +08:00
|
|
|
|
|
|
|
#### Threshold Binary, Inverted
|
|
|
|
|
|
|
|
- This thresholding operation can be expressed as:
|
|
|
|
|
|
|
|
\f[\texttt{dst} (x,y) = \fork{0}{if \(\texttt{src}(x,y) > \texttt{thresh}\)}{\texttt{maxVal}}{otherwise}\f]
|
|
|
|
|
|
|
|
- If the intensity of the pixel \f$src(x,y)\f$ is higher than \f$thresh\f$, then the new pixel intensity
|
|
|
|
is set to a \f$0\f$. Otherwise, it is set to \f$MaxVal\f$.
|
|
|
|
|
2014-11-28 21:21:28 +08:00
|
|
|
![](images/Threshold_Tutorial_Theory_Binary_Inverted.png)
|
2014-11-27 20:39:05 +08:00
|
|
|
|
|
|
|
#### Truncate
|
|
|
|
|
|
|
|
- This thresholding operation can be expressed as:
|
|
|
|
|
|
|
|
\f[\texttt{dst} (x,y) = \fork{\texttt{threshold}}{if \(\texttt{src}(x,y) > \texttt{thresh}\)}{\texttt{src}(x,y)}{otherwise}\f]
|
|
|
|
|
|
|
|
- The maximum intensity value for the pixels is \f$thresh\f$, if \f$src(x,y)\f$ is greater, then its value
|
|
|
|
is *truncated*. See figure below:
|
|
|
|
|
2014-11-28 21:21:28 +08:00
|
|
|
![](images/Threshold_Tutorial_Theory_Truncate.png)
|
2014-11-27 20:39:05 +08:00
|
|
|
|
|
|
|
#### Threshold to Zero
|
|
|
|
|
|
|
|
- This operation can be expressed as:
|
|
|
|
|
|
|
|
\f[\texttt{dst} (x,y) = \fork{\texttt{src}(x,y)}{if \(\texttt{src}(x,y) > \texttt{thresh}\)}{0}{otherwise}\f]
|
|
|
|
|
|
|
|
- If \f$src(x,y)\f$ is lower than \f$thresh\f$, the new pixel value will be set to \f$0\f$.
|
|
|
|
|
2014-11-28 21:21:28 +08:00
|
|
|
![](images/Threshold_Tutorial_Theory_Zero.png)
|
2014-11-27 20:39:05 +08:00
|
|
|
|
|
|
|
#### Threshold to Zero, Inverted
|
|
|
|
|
|
|
|
- This operation can be expressed as:
|
|
|
|
|
|
|
|
\f[\texttt{dst} (x,y) = \fork{0}{if \(\texttt{src}(x,y) > \texttt{thresh}\)}{\texttt{src}(x,y)}{otherwise}\f]
|
|
|
|
|
|
|
|
- If \f$src(x,y)\f$ is greater than \f$thresh\f$, the new pixel value will be set to \f$0\f$.
|
|
|
|
|
2014-11-28 21:21:28 +08:00
|
|
|
![](images/Threshold_Tutorial_Theory_Zero_Inverted.png)
|
2014-11-27 20:39:05 +08:00
|
|
|
|
|
|
|
Code
|
|
|
|
----
|
|
|
|
|
|
|
|
The tutorial code's is shown lines below. You can also download it from
|
2016-06-14 21:01:36 +08:00
|
|
|
[here](https://github.com/opencv/opencv/tree/master/samples/cpp/tutorial_code/ImgProc/Threshold.cpp)
|
2015-04-29 15:31:53 +08:00
|
|
|
@include samples/cpp/tutorial_code/ImgProc/Threshold.cpp
|
2014-11-28 21:21:28 +08:00
|
|
|
|
2014-11-27 20:39:05 +08:00
|
|
|
Explanation
|
|
|
|
-----------
|
|
|
|
|
2014-11-28 21:21:28 +08:00
|
|
|
-# Let's check the general structure of the program:
|
2015-04-30 17:27:58 +08:00
|
|
|
- Load an image. If it is BGR we convert it to Grayscale. For this, remember that we can use
|
2014-11-27 20:39:05 +08:00
|
|
|
the function @ref cv::cvtColor :
|
2016-07-18 21:32:05 +08:00
|
|
|
@snippet cpp/tutorial_code/ImgProc/Threshold.cpp load
|
2014-11-27 20:39:05 +08:00
|
|
|
|
|
|
|
- Create a window to display the result
|
2016-07-18 21:32:05 +08:00
|
|
|
@snippet cpp/tutorial_code/ImgProc/Threshold.cpp window
|
|
|
|
|
2014-11-27 20:39:05 +08:00
|
|
|
- Create \f$2\f$ trackbars for the user to enter user input:
|
|
|
|
|
|
|
|
- **Type of thresholding**: Binary, To Zero, etc...
|
|
|
|
- **Threshold value**
|
2016-07-18 21:32:05 +08:00
|
|
|
@snippet cpp/tutorial_code/ImgProc/Threshold.cpp trackbar
|
|
|
|
|
2014-11-27 20:39:05 +08:00
|
|
|
- Wait until the user enters the threshold value, the type of thresholding (or until the
|
|
|
|
program exits)
|
|
|
|
- Whenever the user changes the value of any of the Trackbars, the function *Threshold_Demo*
|
|
|
|
is called:
|
2016-07-18 21:32:05 +08:00
|
|
|
@snippet cpp/tutorial_code/ImgProc/Threshold.cpp Threshold_Demo
|
|
|
|
|
2014-11-27 20:39:05 +08:00
|
|
|
As you can see, the function @ref cv::threshold is invoked. We give \f$5\f$ parameters:
|
|
|
|
|
|
|
|
- *src_gray*: Our input image
|
|
|
|
- *dst*: Destination (output) image
|
|
|
|
- *threshold_value*: The \f$thresh\f$ value with respect to which the thresholding operation
|
|
|
|
is made
|
|
|
|
- *max_BINARY_value*: The value used with the Binary thresholding operations (to set the
|
|
|
|
chosen pixels)
|
|
|
|
- *threshold_type*: One of the \f$5\f$ thresholding operations. They are listed in the
|
|
|
|
comment section of the function above.
|
|
|
|
|
|
|
|
Results
|
|
|
|
-------
|
|
|
|
|
2014-11-28 21:21:28 +08:00
|
|
|
-# After compiling this program, run it giving a path to an image as argument. For instance, for an
|
2014-11-27 20:39:05 +08:00
|
|
|
input image as:
|
|
|
|
|
2014-11-28 21:21:28 +08:00
|
|
|
![](images/Threshold_Tutorial_Original_Image.jpg)
|
2014-11-27 20:39:05 +08:00
|
|
|
|
2014-11-28 21:21:28 +08:00
|
|
|
-# First, we try to threshold our image with a *binary threhold inverted*. We expect that the
|
2014-11-27 20:39:05 +08:00
|
|
|
pixels brighter than the \f$thresh\f$ will turn dark, which is what actually happens, as we can see
|
|
|
|
in the snapshot below (notice from the original image, that the doggie's tongue and eyes are
|
|
|
|
particularly bright in comparison with the image, this is reflected in the output image).
|
|
|
|
|
2014-11-28 21:21:28 +08:00
|
|
|
![](images/Threshold_Tutorial_Result_Binary_Inverted.jpg)
|
2014-11-27 20:39:05 +08:00
|
|
|
|
2014-11-28 21:21:28 +08:00
|
|
|
-# Now we try with the *threshold to zero*. With this, we expect that the darkest pixels (below the
|
2014-11-27 20:39:05 +08:00
|
|
|
threshold) will become completely black, whereas the pixels with value greater than the
|
|
|
|
threshold will keep its original value. This is verified by the following snapshot of the output
|
|
|
|
image:
|
|
|
|
|
2014-11-28 21:21:28 +08:00
|
|
|
![](images/Threshold_Tutorial_Result_Zero.jpg)
|