彩色图像的直方图均衡Python

时间:2017-03-07 15:03:20

标签: python opencv

我是Opencv的新事物和

我试图用直方图均衡手动方式,不知怎的,我的输出就像       this

首先我首先将格式转换为YCR_CB格式,然后将其拆分为每个y,cr和cb。然后我在y层做了直方图均衡化的事情。

这是我的代码:

    from __future__ import division
    import cv2
    import numpy as np

    img1 = cv2.imread("sup.jpg")

    img2 = cv2.cvtColor(img1,cv2.COLOR_BGR2YCR_CB)

    y,cr,cb = cv2.split(img2)

    #y = cv2.equalizeHist(y)

    x =y

    height,width = y.shape

    hist = [0]*256
    pmf = [0]*256
    cdf = [0]*256
    levelBaru = [0]*256

    cv2.imshow("y before",x)
    cv2.waitKey(0)

    for i in range(0,height):  
        for j in range(0,width):
            hist[y.item(i,j)] += 1 
            #hist[y[i,j]] += 1

    for i in range(0,256):  
            #cari pmf
        pmf[i] = round(hist[i]/(height*width),4)

    print "pmf  done"

    cdf[0] = pmf[0]

    for i in range(1,256):              #cari cdf
        cdf[i] = cdf[i-1]+pmf[i]

    for i in range(1,256):              #cari levelBaru
        levelBaru[i] = int(cdf[i]*255)

    for i in range(0,height):           #baru
        for j in range(0,width):
            for k in range (0,256):
                if(y.item(i,j)==k):
                    #print i," ",levelBaru[i]
                    y.itemset((i,j),levelBaru[k])

    img2 = cv2.merge((y,cr,cb))
    img2 = cv2.cvtColor(img2,cv2.COLOR_YCR_CB2BGR) #supaya ga error pas di stack


    #cv2.namedWindow('result', cv2.WINDOW_NORMAL)
    cv2.imshow('result',img2)

    cv2.waitKey(0)
    cv2.destroyAllWindows

有点找不到这个错误,如果有人可以帮助我那将是很棒的

1 个答案:

答案 0 :(得分:2)

根据问题标题,您希望通过在YCrCb域中将其分割来均衡彩色图像的直方图。编写自己的均衡直方图算法是没有意义的,因为在附带的代码片段中,OpenCV已经有一个方法cv2.eqHist()来满足您的目的。

import cv2

img = cv2.imread("path/to/Lenna.png")

img_y_cr_cb = cv2.cvtColor(img, cv2.COLOR_BGR2YCrCb)
y, cr, cb = cv2.split(img_y_cr_cb)

# Applying equalize Hist operation on Y channel.
y_eq = cv2.equalizeHist(y)

img_y_cr_cb_eq = cv2.merge((y_eq, cr, cb))
img_rgb_eq = cv2.cvtColor(img_y_cr_cb_eq, cv2.COLOR_YCR_CB2BGR)

输入:

enter image description here

输出

enter image description here