Opencv functions and methods related to binary threshold processing are summarized for comparison and use

Binary threshold processing is the basis of many image processing requirements , And it is the foundation of the foundation , This step has a great impact on the follow-up .

In order to improve the performance of the algorithm , It is necessary to select the appropriate threshold for processing as much as possible .

In this paper OpenCV Functions related to binarization threshold processing in 、 Methods to summarize , It is convenient for comparison and use in actual development .

The basis of this paper is function threshold(), If you want to know more about functions first threshold(), You can refer to the blog https://blog.csdn.net/wenhao_ir/article/details/125592684

01- Manually control the slider to find the best threshold with human eyes

See the blog for this method and sample code https://blog.csdn.net/wenhao_ir/article/details/51539122

02- adopt OTSU( Maximum inter class variance method ) Look for the threshold of binarization

The principle of this method and C++ See the blog for example code https://blog.csdn.net/wenhao_ir/article/details/51179117
The above blog post only gives C++ Example code for , Here's another one Python Code .
Take the following figure as an example ：

Baidu online disk download link of the above picture ：https://pan.baidu.com/s/1IaJ8nrQzGuHt3RA8jbu0GQ?pwd=bjkm
Python The sample code is as follows ：

#  Blogger WeChat /QQ 2487872782
#  If you have any questions, you can contact the blogger 
#  Please contact the blogger if you need image processing 
#  Image processing technology exchange QQ Group  271891601

# !/usr/bin/env python
# -*- coding: utf-8 -*-
# OpenCV The version is 4.4.0

import numpy as np
import cv2 as cv
import sys

image = cv.imread('F:/material/images/2022/2022-06/img_300_320.jpg')
if image is None:
    print('Error: Could not load image')
    sys.exit()

# cv.imshow('Source Image', image)

img_gray = cv.cvtColor(image, cv.COLOR_BGR2GRAY)
cv.imshow('img_gray', img_gray)

OTSU_threshold, img_B = cv.threshold(img_gray, 71, 255, cv.THRESH_OTSU)

cv.imshow('img_B', img_B)

cv.waitKey(0)
cv.destroyAllWindows()

The operation results are as follows ：

From the screenshot above, we can see that OTSU The threshold found by the algorithm is 135

03- adopt TRIANGLE( Triangle method ) Look for the threshold of binarization

Python The sample code is as follows ：

# !/usr/bin/env python
# -*- coding: utf-8 -*-
# OpenCV The version is 4.4.0

import numpy as np
import cv2 as cv
import sys

image = cv.imread('F:/material/images/2022/2022-06/img_300_320.jpg')
if image is None:
    print('Error: Could not load image')
    sys.exit()

# cv.imshow('Source Image', image)

img_gray = cv.cvtColor(image, cv.COLOR_BGR2GRAY)
cv.imshow('img_gray', img_gray)

TRIANGLE_threshold, img_B = cv.threshold(img_gray, 71, 255, cv.THRESH_TRIANGLE)

cv.imshow('img_B', img_B)


cv.waitKey(0)
cv.destroyAllWindows()

The operation results are as follows ：

04- The threshold of image binary segmentation is obtained by the maximum entropy of image histogram

The principle and code of getting the threshold value of image binary segmentation through the maximum entropy of image histogram are shown in the blog https://blog.csdn.net/wenhao_ir/article/details/51671023
For the image

The running results of the code in the above blog post are as follows ：

05- utilize OpenCV Function of adaptiveThreshold() Realize adaptive binary threshold segmentation of image

function adaptiveThreshold() And C++ See the blog for example code https://blog.csdn.net/wenhao_ir/article/details/51565517
Let's add another Python Sample code .
Due to function adaptiveThreshold() In the process of image binarization , The threshold is constantly changing , So we can't act like a function threshold() In that way, the threshold value for image binarization is obtained .
Its Python The prototype is as follows ：

dst	= cv.adaptiveThreshold(src, maxValue, adaptiveMethod, thresholdType, blockSize, C[, dst])

function adaptiveThreshold() Two algorithms for adaptive calculation of binary threshold are provided , Respectively ADAPTIVE_THRESH_MEAN_C and ADAPTIVE_THRESH_GAUSSIAN_C , Let's try both methods .
The sample code is as follows ：

#  Blogger WeChat /QQ 2487872782
#  If you have any questions, you can contact the blogger 
#  Please contact the blogger if you need image processing 
#  Image processing technology exchange QQ Group  271891601

# !/usr/bin/env python
# -*- coding: utf-8 -*-
# OpenCV The version is 4.4.0

import numpy as np
import cv2 as cv
import sys

image = cv.imread('F:/material/images/2022/2022-06/img_300_320.jpg')
if image is None:
    print('Error: Could not load image')
    sys.exit()

# cv.imshow('Source Image', image)

img_gray = cv.cvtColor(image, cv.COLOR_BGR2GRAY)
cv.imshow('img_gray', img_gray)

#  function adaptiveThreshold() Some parameters required 
blockSize = 5
constValue = 0
maxVal = 255

img_B_MEAN = cv.adaptiveThreshold(img_gray, maxVal, cv.ADAPTIVE_THRESH_MEAN_C, cv.THRESH_BINARY, blockSize, constValue)
img_B_GAUSSIAN = cv.adaptiveThreshold(img_gray, maxVal, cv.ADAPTIVE_THRESH_GAUSSIAN_C, cv.THRESH_BINARY, blockSize, constValue)

cv.imshow('img_B_MEAN', img_B_MEAN)
cv.imshow('img_B_GAUSSIAN', img_B_GAUSSIAN)

cv.waitKey(0)
cv.destroyAllWindows()

From the above results, we can see , Due to function adaptiveThreshold() It is to process the original gray-scale image piece by piece , Therefore, the details of the image are preserved more .

06- Integrated approach 1、2、3、5 Code for

Due to the method in this article 01、02、03、05 It's all direct calls OpenCV Library function implementation of , So we can write these four in a sample code , It is convenient for comparison when it is convenient for application , Then choose a method that fits your needs better .
The synthesis code is as follows ：

#  Blogger WeChat /QQ 2487872782
#  If you have any questions, you can contact the blogger 
#  Please contact the blogger if you need image processing 
#  Image processing technology exchange QQ Group  271891601

# !/usr/bin/env python
# -*- coding: utf-8 -*-
# OpenCV The version is 4.4.0

import numpy as np
import cv2 as cv
import sys

image = cv.imread('F:/material/images/2022/2022-06/img_300_320.jpg')
if image is None:
    print('Error: Could not load image')
    sys.exit()

# cv.imshow('Source Image', image)

img_gray = cv.cvtColor(image, cv.COLOR_BGR2GRAY)
cv.imshow('img_gray', img_gray)

#  Method 1: The user specifies the threshold value for binarization 
Manual_threshold, img_bi_manual = cv.threshold(img_gray, 71, 255, cv.THRESH_BINARY)
cv.imshow('img_bi_manual', img_bi_manual)

#  Method 2: use OTSU Method to determine the threshold value of binarization 
OTSU_threshold, img_bi_OTSU = cv.threshold(img_gray, 71, 255, cv.THRESH_OTSU)
cv.imshow('img_bi_OTSU', img_bi_OTSU)

#  Method 3： Determine the threshold value of binarization by triangle method 
TRIANGLE_threshold, img_bi_TRIANGLE = cv.threshold(img_gray, 71, 255, cv.THRESH_TRIANGLE)
cv.imshow('img_bi_TRIANGLE', img_bi_TRIANGLE)

#  Method 5: Using functions adaptiveThreshold() Adaptive threshold processing 
blockSize = 5
constValue = 0
maxVal = 255

#  Method 5-1：MEAN_C Adaptive threshold 
img_B_MEAN = cv.adaptiveThreshold(img_gray, maxVal, cv.ADAPTIVE_THRESH_MEAN_C, cv.THRESH_BINARY, blockSize, constValue)
cv.imshow('img_B_MEAN', img_B_MEAN)

#  Method 5-1：GAUSSIAN_C Adaptive threshold 
img_B_GAUSSIAN = cv.adaptiveThreshold(img_gray, maxVal, cv.ADAPTIVE_THRESH_GAUSSIAN_C, cv.THRESH_BINARY, blockSize, constValue)
cv.imshow('img_B_GAUSSIAN', img_B_GAUSSIAN)

cv.waitKey(0)
cv.destroyAllWindows()

The operation results are as follows ：