图像分割-传统算法-边缘分割

基于边缘检测的图像分割流程

1、灰度图转变

2、使用边缘检测算子进行边缘提取，如canny、sobel、拉普拉斯算子

3、对边缘进行闭合处理

4、显示闭合后的轮廓

试了下canny、sobel、拉普拉斯，其中拉普拉斯算子可以事先对灰度图平滑处理一下（拉普拉斯对噪声敏感）。

python 复制代码

import cv2
import numpy as np

def edge_detection_canny(img_path, thres=(50,150), ksize=5):
    '''
    :param img_path:
    :param thres: canny算子的限制阈值，[0]为低阈值，[1]为高阈值
    :param ksize: canny的核大小
    :return:
    '''
    img = cv2.imread(img_path)
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

    edges = cv2.Canny(gray, thres[0], thres[1])

    #  形态学闭运算：连接断裂的边缘, cv2.MORPH_RECT结果元素矩形, cv2.MORPH_ELLIPSE椭圆形， CROSS十字形
    kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (ksize, ksize))
    # kernel = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (ksize, ksize))
    # kernel = cv2.getStructuringElement(cv2.MORPH_CROSS, (ksize, ksize))
    #  闭运算：先膨胀后腐蚀，用于填充边缘间的细小间隙，连接相邻的边缘片段
    closed = cv2.morphologyEx(edges, cv2.MORPH_CLOSE, kernel)

    # 查找轮廓
    # 在闭运算后的边缘图像中查找物体的轮廓
    # RETR_EXTERNAL: 只检索最外层轮廓（忽略内部孔洞的轮廓），RETR_LIST检索所有轮廓不建立层级，RETR_CCOMP检索所有轮廓组织为两级层级；检索所有轮廓，重建完整的嵌套层级
    # CHAIN_APPROX_SIMPLE: 压缩水平、垂直和对角线段，只保留端点；CHAIN_APPROX_NONE：存储所有轮廓
    contours, _ = cv2.findContours(closed.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)

    # 在原图上绘制轮廓
    cv2.drawContours(img, contours, -1, (0, 255, 0), 2)

    cv2.imshow('edge', edges)
    cv2.imshow('closed', closed)
    cv2.imshow('img', img)
    cv2.waitKey(0)
    cv2.destroyAllWindows()

def edge_detection_sobel(img_path, ksize=3):
    img = cv2.imread(img_path)
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

    sobel_x = cv2.Sobel(gray, cv2.CV_64F, 1, 0, ksize=ksize)
    sobel_y = cv2.Sobel(gray, cv2.CV_64F, 0, 1, ksize=ksize)
    sobel_magnitude = np.sqrt(sobel_x ** 2 + sobel_y ** 2)

    # 转换为8位无符号整数并归一化到0-255
    sobel_edges = np.uint8(np.clip(sobel_magnitude, 0, 255))

    kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (5, 5))
    closed = cv2.morphologyEx(sobel_edges, cv2.MORPH_CLOSE, kernel)

    contours, _ = cv2.findContours(closed.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)

    cv2.drawContours(img, contours, -1, (0, 255, 0), 2)

    cv2.imshow('edge', sobel_edges)
    cv2.imshow('closed', closed)
    cv2.imshow('img', img)
    cv2.waitKey(0)
    cv2.destroyAllWindows()

def edge_detection_lap(img_path, ksize=3):
    img = cv2.imread(img_path)
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

    blurred = cv2.GaussianBlur(gray, (5, 5), 0)
    edges = cv2.Laplacian(blurred, cv2.CV_64F, ksize=ksize)

    laplacian_abs = np.absolute(edges)
    laplacian_edges = np.uint8(np.clip(laplacian_abs, 0, 255))

    kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (5, 5))
    closed = cv2.morphologyEx(laplacian_edges, cv2.MORPH_CLOSE, kernel)

    contours, _ = cv2.findContours(closed.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)

    cv2.drawContours(img, contours, -1, (0, 255, 0), 2)

    cv2.imshow('edge', laplacian_edges)
    cv2.imshow('closed', closed)
    cv2.imshow('img', img)
    cv2.waitKey(0)
    cv2.destroyAllWindows()


if __name__ == '__main__':
    img_path = './test/test_img.jpg'
    edge_detection_canny(img_path)
    # edge_detection_sobel(img_path)
    # edge_detection_lap(img_path)

参考
理论学习1
理论学习2
代码参考1