使用otsu方法的最优阈值处理

import cv2
import numpy as np
import matplotlib.pyplot as plt
import cv2 as cv
import math

def calc_gray_hist(image):
    rows,cols=image.shape
    gray_hist=np.zeros([256],dtype=np.uint64)
    for i in range(rows):
        for j in range(cols):
            gray_hist[image[i,j]]+=1
    return gray_hist


def otsu_thresh(image):
    rows,cols=image.shape
    #计算灰度直方图
    gray_hist=calc_gray_hist(image)
    #直方图归一化处理
    norm_hist=gray_hist/(rows*cols)
    #计算累计距，一阶累积矩
    zero_cumu_moment=np.zeros([256],dtype=np.float64)
    one_cumu_moment = np.zeros([256], dtype=np.float64)
    variance_G= np.zeros([256], dtype=np.float64)
    for i in range(256):
        if i==0:
            zero_cumu_moment[i]=norm_hist[i]
            one_cumu_moment[i]=0
            variance_G[i] = 0
        else:
            zero_cumu_moment[i]=zero_cumu_moment[i-1]+norm_hist[i]
            one_cumu_moment[i]=one_cumu_moment[i-1]+zero_cumu_moment[i]*i

        #计算类间方差
    mG=one_cumu_moment[255]
    thresh=0
    sigma=0
    for i in range(256):
        if i==0:
            variance_G[i] = 0
        else:
            variance_G[i]=variance_G[i-1]+math.pow(i-mG,2)*norm_hist[i]
    for i in range(256):
        if zero_cumu_moment[i]==0 and zero_cumu_moment[i]==0:
            sigma_temp = 0
        else:
            sigma_temp=math.pow(mG*zero_cumu_moment[i]-one_cumu_moment[i],2)/zero_cumu_moment[i]*(1-zero_cumu_moment[i])
        if sigmathresh]=255
    thresh_img[thresh_img<=thresh] =0
    return thresh,thresh_img,yita


if __name__=='__main__':
    img=cv.imread('../data/1.jpg',0)
    gray_hist=calc_gray_hist(img)
    thresh,thresh_img,yita=otsu_thresh(img)
    print('阈值：',thresh,"可分离性测量度：",yita)
    cv2.imwrite('hhh.jpg',thresh_img)
    plt.plot(gray_hist)
    plt.show()
    cv.waitKey()
    cv.destroyAllWindows()

                                        冈萨雷斯第四版P541