基于Welsh算法的灰度图像彩色化处理及优化python实现源码+项目说明+详细注释.zip

import cv2 # 导入opencv import bisect # 导入bisect,用于有序序列的插入和查找 import numpy as np # 导入numpy(可用来存储和处理大型矩阵)并起别名 from Colorize import WeightPixel # 函数导入 from Colorize import Pixel from Colorize import ConsultImage if __name__ == '__main__': # 创建参考图像的分析类； print("---Initializes the reference image---\n") consult_image = ConsultImage("Reference_image2.jpg") print("---Initialization complete---\n") # 读取灰度图像；opencv默认读取的是3通道的，不需要扩展通道； gray_image = cv2.imread("Gray_image2.jpg") cv2.imshow("Grayscale_image", gray_image) # 展示灰度图像 gray_image = cv2.cvtColor(gray_image, cv2.COLOR_BGR2LAB) # 实现了RGB和Lab之间的转换 height, width, channel = gray_image.shape # 读取矩阵 # 获取灰度图像的亮度信息； gray_image_l_origin, gray_image_a, gray_image_b = cv2.split(gray_image) gray_image_l = np.zeros(gray_image_l_origin.shape, np.uint8) max_pixel, min_pixel = np.max( gray_image_l_origin), np.min(gray_image_l_origin) consult_max_pixel, consult_min_pixel = consult_image.max_min_pixel() pixel_ratio = (consult_max_pixel - consult_min_pixel) / \ (max_pixel - min_pixel) # 把灰度图像的亮度值映射到参考图像范围内； print("---Mapping---\n") for i in range(height): for j in range(width): pixel_light = consult_min_pixel + \ (gray_image_l_origin[i, j] - min_pixel) * pixel_ratio gray_image_l[i, j] = pixel_light # 获取领域的窗口大小； window_size = consult_image.get_window_size() ratio = consult_image.get_ratio() print("---Start colorization---\n") for row in range(height): for col in range(width): pixel = Pixel(row, col) pixel_light = gray_image_l[pixel.x, pixel.y] # 求窗口内像素方差； window_left = max(pixel.x - window_size, 0) window_right = min(pixel.x + window_size + 1, height) window_top = max(pixel.y - window_size, 0) window_bottom = min(pixel.y + window_size + 1, width) window_slice = gray_image_l[window_left: window_right, window_top: window_bottom] pixel_std = np.std(window_slice) weight_pixel = WeightPixel( ratio * pixel_light + (1 - ratio) * pixel_std, 0, 0) search_pixel = bisect.bisect( consult_image.get_weight_list(), weight_pixel) search_pixel = 1 if search_pixel == 0 else search_pixel search_pixel = len(consult_image.get_weight_list()) - 1 \ if search_pixel == len(consult_image.get_weight_list()) else search_pixel left_pixel = consult_image.get_weight_list()[search_pixel - 1] right_pixel = consult_image.get_weight_list()[search_pixel] nearest_pixel = left_pixel if left_pixel.weight + \ right_pixel.weight > 2 * weight_pixel.weight else right_pixel gray_image_a[row, col] = nearest_pixel.a gray_image_b[row, col] = nearest_pixel.b merge_image = cv2.merge([gray_image_l, gray_image_a, gray_image_b]) rgb_image = cv2.cvtColor(merge_image, cv2.COLOR_LAB2BGR) cv2.imshow("Shaded_image", rgb_image) # 展示着色后的图像 cv2.waitKey(0) cv2.destroyAllWindows() cv2.imwrite("Shaded_image2.jpg", rgb_image) # 写入