图像的分割（三）

图像的分割之基于区域的分割

基于区域生长法的分割

import numpy as np
import cv2 as cv


# 区域生长法
def region_growth(gray, seeds, thresh=15):
    # 采用8-邻域
    connects = [(-1, -1), (0, -1), (1, -1), (1, 0), (1, 1), (0, 1), (-1, 1), (-1, 0)]
    height, width = gray.shape
    # 标记矩阵，标记所给生长点生长后的区域
    seed_mark = np.zeros(gray.shape)
    # 生长点列表
    seed_list = []
    for seed in seeds:
        seed_list.append(seed)
    # 如果生长点列表中还有生长点
    while len(seed_list) > 0:
        # 获取当前生长点
        current_point = seed_list.pop(0)
        # 将生长点标记为1
        seed_mark[current_point[0], current_point[1]] = 1
        # 对当前生长点的8-邻域进行相似性判断
        for i in range(len(connects)):
            current_x = current_point[0] + connects[i][0]
            current_y = current_point[1] + connects[i][1]
            if current_x < 0 or current_y < 0 or current_x >= height or current_y >= width:
                continue
            # 生长点与邻域的点的像素值之差
            dist = abs(int(gray[current_x, current_y]) - int(gray[current_point[0], current_point[1]]))
            # 如果生长点的8-邻域中有满足相似性准则的并且还没有被标记的点
            if dist < thresh and seed_mark[current_x, current_y] == 0:
                # 在标记矩阵中将该点标记为1
                seed_mark[current_x, current_y] = 1
                # 加入生长点列表
                seed_list.append((current_x, current_y))
    return seed_mark


gray = cv.imread("./watershed.jpg", cv.IMREAD_GRAYSCALE)
seeds = [(2, 2), (2, 50), (2, 100)]
result = region_grow(gray, seeds, thresh=15)
cv.imshow("img", result)
cv.waitKey(0)