灰度共生矩阵python代码

下面是使用Python实现灰度共生矩阵的示例代码，代码中使用了Python的numpy和opencv库：

import cv2
import numpy as np

# 读取图像
img = cv2.imread('image.png', 0)

# 灰度共生矩阵参数
d = 1 # 距离
theta = [0, np.pi/4, np.pi/2, 3*np.pi/4] # 方向
levels = 256 # 灰度级数目

# 计算灰度共生矩阵
glcm = np.zeros((levels, levels, len(theta)))
for i, angle in enumerate(theta):
    offset = np.array([d*np.cos(angle), -d*np.sin(angle)])
    for y in range(img.shape[0]):
        for x in range(img.shape[1]):
            yp = int(round(y + offset[1]))
            xp = int(round(x + offset[0]))
            if yp >= 0 and yp < img.shape[0] and xp >= 0 and xp < img.shape[1]:
                glcm[img[y, x], img[yp, xp], i] += 1

# 归一化灰度共生矩阵
glcm /= np.sum(glcm)

# 计算灰度共生矩阵特征
features = np.zeros((len(theta), 4))
for i in range(len(theta)):
    # 能量
    features[i, 0] = np.sum(glcm[:,:,i]**2)
    # 对比度
    features[i, 1] = np.sum((np.arange(levels)[:,np.newaxis,np.newaxis] - np.arange(levels)[np.newaxis,:,np.newaxis])**2 * glcm[:,:,i])
    # 相关性
    mu_x = np.sum(glcm[:,:,i] * np.arange(levels)[:,np.newaxis])
    mu_y = np.sum(glcm[:,:,i] * np.arange(levels)[np.newaxis,:])
    sigma_x = np.sqrt(np.sum(glcm[:,:,i] * (np.arange(levels)[:,np.newaxis] - mu_x)**2))
    sigma_y = np.sqrt(np.sum(glcm[:,:,i] * (np.arange(levels)[np.newaxis,:] - mu_y)**2))
    cov = np.sum(glcm[:,:,i] * (np.arange(levels)[:,np.newaxis] - mu_x) * (np.arange(levels)[np.newaxis,:] - mu_y))
    features[i, 2] = cov / (sigma_x * sigma_y)
    # 熵
    features[i, 3] = -np.sum(glcm[:,:,i] * np.log2(glcm[:,:,i] + 1e-10))

print(features)

该代码读取一张灰度图像，并计算其在四个方向上的灰度共生矩阵，然后归一化灰度共生矩阵，并计算每个方向上的能量、对比度、相关性和熵四个特征。最终输出一个4x4的特征矩阵。