python计算两个图像的互信息

我们常用互信息来衡量两个图像的相似程度。互信息的计算方式如下：

使用python中的numpy包或者sklearn可以很方便的计算互信息，计算代码如下：

import cv2
import numpy as np
import sklearn.metrics as skmdef hxx_forward(x, y):return skm.mutual_info_score(x, y)def hxx(x, y):size = x.shape[-1]px = np.histogram(x, 256, (0, 255))[0] / sizepy = np.histogram(y, 256, (0, 255))[0] / sizehx = - np.sum(px * np.log(px + 1e-8))hy = - np.sum(py * np.log(py + 1e-8))hxy = np.histogram2d(x, y, 256, [[0, 255], [0, 255]])[0]hxy /= (1.0 * size)hxy = - np.sum(hxy * np.log(hxy + 1e-8))r = hx + hy - hxyreturn rimg1 = cv2.imread("F:/BYSY/ACCESS/ImageMatch/1.png", cv2.IMREAD_GRAYSCALE)
img2 = cv2.imread("F:/BYSY/ACCESS/ImageMatch/1.png", cv2.IMREAD_GRAYSCALE)x = np.reshape(img1, -1)
y = np.reshape(img2, -1)print(hxx_forward(x, y))
print(hxx(x, y))

经验证，两种方法计算结果相同