1 概念

图像直方图均衡化是对图像像素重新映射，使得映射后的像素分布更加均匀，图像显示适合人类视觉。

2 流程

(1) 计算原图像像素的频率
(2) 计算原图像像素的累积频率
(3) 归一化

3 代码实现

（1）方法一

### 3 灰度图像进行直方图均衡化
## histeq(im,nbr_bins=256) ：img --> imhist,bins  -->.cumsum() --> normalize--> interp
def get_pdf(img):
    total = img.size
    return [np.sum(img == i)/total  for i in range(256)]

def histeq_01(img):
    pr = get_pdf(img)
    img0 = img.copy()
    y_points = []
    cum = 0.
    for i in range(256):
        cum = cum + pr[i]
        img0[img==i] = cum * 255.
        y_points.append(cum * 255.)
    return img0, y_points

if __name__ == '__main__':
    import numpy as np
    from pylab import *
    from PIL import Image
	img = np.array(Image.open('luna.png').convert('L'))
    img0, y_points = histeq_01(img)
    img00 = Image.fromarray(np.uint8(img0))

    subplot(121)
    imshow(img, cmap='gray')
    subplot(122)
    imshow(img00,cmap='gray')

    show()

在这里插入图片描述

（1）方法二


def histeq_02(img,bins=256):
    imhist,bins_ = np.histogram(img.flatten(),bins)
    cdf = imhist.cumsum()  #  累积分布函数
    cdf = cdf/cdf[-1]*255  #  归一化
    img0 = np.interp(img.flatten(),bins_[:-1],cdf)
    return img0.reshape(img.shape),cdf

if __name__ == '__main__':
    img = np.array(Image.open('luna.png').convert("L"))
    img0,cdf = histeq_02(img)
    plt.subplot(121)
    plt.imshow(img,cmap='gray')
    plt.subplot(122)
    plt.imshow(img0,cmap='gray')
    plt.show()