dog 高斯差分边缘检测

Applies two Gaussian blurs to the drawable, and subtracts the results.  This is robust and widely used method for detecting edges.

 

cv::Mat input;// source image
cv::Mat edge;// output image
cv::Mat blur1;
    cv::Mat blur2;
    int r1 = 1; // 奇数
    int r2 = 49;// 奇数
    float s1 =sqrt (-(r1 * r1) / (2 * log (1.0 / 255.0))) ;
    float s2 =sqrt (-(r2 * r2) / (2 * log (1.0 / 255.0))) ;
    // 两个不同半径的高斯滤波
    cv::GaussianBlur(input, blur1, cv::Size(r1,r1),s1);
    cv::GaussianBlur(input, blur2, cv::Size(r2,r2),s2);
    //作差 
    cv::subtract(blur1,blur2,edge);

    //反色
    cv::bitwise_not(edge,edge);

   

Gimp 插件高斯差分边缘检测 源码：

https://github.com/piksels-and-lines-orchestra/gimp/blob/master/plug-ins/common/edge-dog.c