用python实现基于概率分布的超声图像与自然图像性质差异分析，不用greycomatrix和greycoprops，使用opencv，两种图像的文件类型都是jpg,超声图像的文件路径是‘D:\zzz\us2’，自然图像的文件路径是‘D:\zzz\na2’，两种图像都有84张，图片形状都不同，图像名称是随机的，需要将图像归一化，需要对比两种图像的特征必须有颜色，纹理，形状，差异性分析方法也需要多种，包括完整详细最新的代码

首先，我们需要安装必要的库：OpenCV、NumPy、Matplotlib。

pip install opencv-python numpy matplotlib

接下来，我们可以开始编写代码：

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

# 超声图像路径
us_path = 'D:/zzz/us2/'
# 自然图像路径
na_path = 'D:/zzz/na2/'

# 存储图像特征的列表
us_features = []
na_features = []

# 循环读取超声图像
for i in range(1, 85):
    # 拼接文件路径
    filename = us_path + str(i) + '.jpg'
    # 读取图像并灰度化
    img = cv2.imread(filename)
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    # 对图像进行归一化处理
    normalized = cv2.normalize(gray, None, alpha=0, beta=255, norm_type=cv2.NORM_MINMAX)
    # 计算图像的颜色特征
    hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
    hist_h = cv2.calcHist([hsv], [0], None, [180], [0, 180])
    hist_s = cv2.calcHist([hsv], [1], None, [256], [0, 256])
    hist_v = cv2.calcHist([hsv], [2], None, [256], [0, 256])
    # 计算图像的纹理特征
    gray_32 = np.float32(gray)
    dst = cv2.cornerHarris(gray_32, 2, 3, 0.04)
    dst = cv2.dilate(dst, None)
    corners = np.where(dst > 0.01 * dst.max())
    texture_feature = len(corners[0])
    # 将特征存入列表中
    us_features.append([hist_h, hist_s, hist_v, texture_feature])

# 循环读取自然图像，处理方法与超声图像相同
for i in range(1, 85):
    filename = na_path + str(i) + '.jpg'
    img = cv2.imread(filename)
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    normalized = cv2.normalize(gray, None, alpha=0, beta=255, norm_type=cv2.NORM_MINMAX)
    hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
    hist_h = cv2.calcHist([hsv], [0], None, [180], [0, 180])
    hist_s = cv2.calcHist([hsv], [1], None, [256], [0, 256])
    hist_v = cv2.calcHist([hsv], [2], None, [256], [0, 256])
    gray_32 = np.float32(gray)
    dst = cv2.cornerHarris(gray_32, 2, 3, 0.04)
    dst = cv2.dilate(dst, None)
    corners = np.where(dst > 0.01 * dst.max())
    texture_feature = len(corners[0])
    na_features.append([hist_h, hist_s, hist_v, texture_feature])

# 计算图像形状特征
us_shapes = []
na_shapes = []
for i in range(1, 85):
    filename = us_path + str(i) + '.jpg'
    img = cv2.imread(filename)
    shape = img.shape
    us_shapes.append(shape)
for i in range(1, 85):
    filename = na_path + str(i) + '.jpg'
    img = cv2.imread(filename)
    shape = img.shape
    na_shapes.append(shape)

# 计算颜色特征的差异性
color_diffs = []
for i in range(84):
    us_hist_h = us_features[i][0]
    us_hist_s = us_features[i][1]
    us_hist_v = us_features[i][2]
    na_hist_h = na_features[i][0]
    na_hist_s = na_features[i][1]
    na_hist_v = na_features[i][2]
    color_diff = cv2.compareHist(us_hist_h, na_hist_h, cv2.HISTCMP_CORREL) + cv2.compareHist(us_hist_s, na_hist_s, cv2.HISTCMP_CORREL) + cv2.compareHist(us_hist_v, na_hist_v, cv2.HISTCMP_CORREL)
    color_diffs.append(color_diff)

# 计算纹理特征的差异性
texture_diffs = []
for i in range(84):
    us_texture = us_features[i][3]
    na_texture = na_features[i][3]
    texture_diff = abs(us_texture - na_texture)
    texture_diffs.append(texture_diff)

# 计算形状特征的差异性
shape_diffs = []
for i in range(84):
    us_shape = us_shapes[i]
    na_shape = na_shapes[i]
    shape_diff = abs(us_shape[0]-na_shape[0]) + abs(us_shape[1]-na_shape[1]) + abs(us_shape[2]-na_shape[2])
    shape_diffs.append(shape_diff)

# 将三种特征的差异性综合起来
total_diffs = []
for i in range(84):
    total_diff = color_diffs[i] + texture_diffs[i] + shape_diffs[i]
    total_diffs.append(total_diff)

# 绘制差异性直方图
plt.hist(total_diffs, bins=20, alpha=0.5, label='Total Difference')
plt.hist(color_diffs, bins=20, alpha=0.5, label='Color Difference')
plt.hist(texture_diffs, bins=20, alpha=0.5, label='Texture Difference')
plt.hist(shape_diffs, bins=20, alpha=0.5, label='Shape Difference')
plt.legend(loc='upper right')
plt.show()

在这段代码中，我们首先循环读取超声图像和自然图像，对每张图像进行归一化处理，并计算其颜色和纹理特征。然后，我们计算了三种特征的差异性，并将它们综合成一个总的差异性指标。最后，我们使用 Matplotlib 绘制了差异性的直方图。

需要注意的是，这段代码中使用了 OpenCV 的一些函数来计算图像的颜色和纹理特征。如果你对这些函数不熟悉，可以查看 OpenCV 的文档来了解更多信息。