具体参见OpenCV 用户手册

全局阈值

共有5 种二值化方法：

其图像解释为：

import cv2 as cv


def threshold_demo(image):
        src = cv.imread(image)
        gray = cv.cvtColor(src, cv.COLOR_BGR2GRAY) #先转化为灰度图
        #ret, binary = cv.threshold(gray, 0, 255, cv.THRESH_BINARY | cv.THRESH_OTSU)
        ret, binary = cv.threshold(gray, 0, 255, cv.THRESH_BINARY | cv.THRESH_TRIANGLE) #阈值自动赋值，该函数输出两个参数，第一个是阈值，第二个是输出的二值化图像
        print('Threshold value %s ' %ret)
        cv.imshow('binary', binary)


threshold_demo('/home/pi/Desktop/woman.jpg')
cv.waitKey(0)
cv.destroyAllWindows()

THRESH_TRIANGLE 相较于THRESH_OTSU 更加适合于某个像素值的数量远多于其他像素值数量时

若要对阈值进行手动赋值，则:

ret, binary = cv.threshold(gray, 150, 255, cv.THRESH_BINARY)

第二个实参则为阈值，当第二个实参不为零时，对于图像的阈值赋值方式失去作用| cv.THRESH_TRIANGLE

像素取反（大于阈值为黑，反之为白）

ret, binary = cv.threshold(gray, 117, 255, cv.THRESH_BINARY_INV)

ret, binary = cv.threshold(gray, 0, 255, cv.THRESH_BINARY_INV | cv.THRESH_OTSU)

局部阈值

具体参见OpenCV 用户手册

def local_threshold(image):
        src = cv.imread(image)
        gray = cv.cvtColor(src, cv.COLOR_BGR2GRAY)
        binary = cv.adaptiveThreshold(gray, 255, cv.ADAPTIVE_THRESH_GAUSSIAN_C, cv.THRESH_BINARY, 25, $
        binary2 = cv.adaptiveThreshold(gray, 255, cv.ADAPTIVE_THRESH_MEAN_C, cv.THRESH_BINARY, 25, 10)
        cv.imshow('Gaussian', binary)
        cv.imshow('MEAN', binary2)

与全局阈值进行对比：

可见局部阈值对图形边缘绘制更加细腻

Gaussian 局部阈值和MEAN 局部阈值比较：

代码层面知识点

OpenCV 中图像二值化方法

1. OTSU
2. Triangle
3. 自动与手动

---

自适应阈值