python代码实现用pm算法处理图像，实现降噪与图像加强

以下是使用Python实现基于PM算法的自适应中值滤波算法和非局部均值滤波算法的代码：

import cv2
import numpy as np

# 基于PM算法的自适应中值滤波算法
def adaptive_median_filter(img, max_window_size=7, threshold=15):
    img_out = np.zeros_like(img)

    height, width = img.shape[:2]
    for i in range(height):
        for j in range(width):
            window_size = 3
            while window_size <= max_window_size:
                x1 = max(i - window_size, 0)
                x2 = min(i + window_size, height - 1)
                y1 = max(j - window_size, 0)
                y2 = min(j + window_size, width - 1)

                window = img[x1:x2+1, y1:y2+1]
                z_min = np.min(window)
                z_max = np.max(window)
                z_med = np.median(window)
                z_xy = img[i, j]

                if z_med - z_min > threshold and z_max - z_med > threshold:
                    if z_min < z_xy < z_max:
                        img_out[i, j] = z_xy
                    else:
                        img_out[i, j] = z_med
                    break
                else:
                    window_size += 2

            if window_size > max_window_size:
                img_out[i, j] = z_med

    return img_out

# 基于PM算法的非局部均值滤波算法
def non_local_means_filter(img, h=10, search_window_size=21, patch_size=7):
    img_out = np.zeros_like(img)

    height, width = img.shape[:2]
    pad_size = search_window_size // 2
    img_pad = cv2.copyMakeBorder(img, pad_size, pad_size, pad_size, pad_size, cv2.BORDER_REPLICATE)

    for i in range(height):
        for j in range(width):
            x1 = i + pad_size - search_window_size // 2
            x2 = i + pad_size + search_window_size // 2
            y1 = j + pad_size - search_window_size // 2
            y2 = j + pad_size + search_window_size // 2

            search_window = img_pad[x1:x2+1, y1:y2+1]
            search_window_center = search_window[patch_size:patch_size+1, patch_size:patch_size+1]

            weight_sum = 0
            pixel_sum = 0
            for k in range(pad_size, pad_size+height):
                for l in range(pad_size, pad_size+width):
                    patch = img_pad[k-patch_size:k+patch_size+1, l-patch_size:l+patch_size+1]
                    patch_center = patch[patch_size:patch_size+1, patch_size:patch_size+1]

                    diff = np.sum((patch - search_window_center) ** 2)
                    weight = np.exp(-diff / (h ** 2))

                    weight_sum += weight
                    pixel_sum += weight * img_pad[k, l]

            img_out[i, j] = pixel_sum / weight_sum

    return img_out.astype(np.uint8)

# 测试代码
img = cv2.imread('test.jpg', cv2.IMREAD_GRAYSCALE)
img_adaptive_median = adaptive_median_filter(img)
img_non_local_means = non_local_means_filter(img)

cv2.imshow('Original Image', img)
cv2.imshow('Adaptive Median Filter', img_adaptive_median)
cv2.imshow('Non-Local Means Filter', img_non_local_means)
cv2.waitKey(0)
cv2.destroyAllWindows()

需要注意的是，这里使用的是OpenCV库来读取和显示图像，如果没有安装OpenCV库需要先安装。此外，这里只提供了基于PM算法的两种图像处理方法的代码实现，实际应用中需要根据具体情况进行调整和优化。