1

我正在尝试在 OpenCV 的帮助下从头开始实现 Canny 边缘检测算法。我在实施有助于减薄边缘的非最大抑制步骤时遇到问题。

我的逻辑是首先计算强度梯度向量，然后将其分组为 0、45、90、135 度方向，然后尝试找到局部最大值。找到这个局部最大值的方法是确保当前像素大于同一方向上的后续像素和前面的像素。如果不是，我将零值分配给该像素。使用这个逻辑，我仍然无法缩小边缘。我觉得错误是在我计算每个像素的强度梯度向量时。

这是我的代码-

#include <iostream>
#include <bits/stdc++.h>
#include "opencv2/core/core.hpp"
#include "opencv2/highgui/highgui.hpp"
#include "opencv2/opencv.hpp"
#include "opencv2/imgproc/imgproc.hpp"
#include <opencv2/objdetect/objdetect.hpp>
#include <math.h>

using namespace cv; 
using namespace std; 

int main()
{
// Reading image
Mat img = imread("1.jpg");

// Displaying image
//imshow("Original Image",img);
//waitKey(0);

// Converting to grayscale
Mat img_gray,image_gray;
cvtColor(img,image_gray,CV_RGB2GRAY);
GaussianBlur( image_gray, img_gray, Size(15,15), 3, 3);


// Displaying grayscale image
imshow("Original Image",img_gray);
waitKey(0);


int cols = img_gray.cols;
int rows = img_gray.rows;

// Creating sobel operator in x direction
int sobel_x[3][3] = {-1,0,1,-2,0,2,-1,0,1};
// Creating sobel operator in y direction
int sobel_y[3][3] = {1,2,1,0,0,0,-1,-2,-1};


int radius = 1;

// Handle border issues
Mat _src;
copyMakeBorder(img_gray, _src, radius, radius, radius, radius, BORDER_REFLECT101);

// Create output matrix
Mat gradient_x = img_gray.clone();
Mat gradient_y = img_gray.clone();
Mat gradient_f = img_gray.clone();
Mat gradient_mag = img_gray.clone();

// Conrrelation loop in x direction 

// Iterate on image 
for (int r = radius; r < _src.rows - radius; ++r)
{
    for (int c = radius; c < _src.cols - radius; ++c)
    {
        int s = 0;

        // Iterate on kernel
        for (int i = -radius; i <= radius; ++i)
        {
            for (int j = -radius; j <= radius; ++j)
            {
                s += _src.at<uchar>(r + i, c + j) * sobel_x[i + radius][j + radius];
            }
        }
        gradient_x.at<uchar>(r - radius, c - radius) = s/8;

        /*if(s>200)
            gradient.at<uchar>(r - radius, c - radius) = 255;
        else
            gradient.at<uchar>(r - radius, c - radius) = 0;
        */    
    }
}

// Conrrelation loop in y direction 

// Iterate on image 
for (int r = radius; r < _src.rows - radius; ++r)
{
    for (int c = radius; c < _src.cols - radius; ++c)
    {
        int s = 0;

        // Iterate on kernel
        for (int i = -radius; i <= radius; ++i)
        {
            for (int j = -radius; j <= radius; ++j)
            {
                s += _src.at<uchar>(r + i, c + j) * sobel_y[i + radius][j + radius];
            }
        }
        gradient_y.at<uchar>(r - radius, c - radius) = s/8;

        /*if(s>200)
            gradient.at<uchar>(r - radius, c - radius) = 255;
        else
            gradient.at<uchar>(r - radius, c - radius) = 0;
        */    
    }
}

///cout<<endl<<"max:"<<max;
//cout<<img_gray.rows;
//cout<<endl<<_src.rows;
cout<<endl<<gradient_x.rows;
cout<<endl<<gradient_y.rows;
cout<<endl<<gradient_f.rows<<gradient_f.cols;   


//Calculating gradient magnitude
for(int i=0; i<gradient_mag.rows; i++)
{
    for(int j=0; j<gradient_mag.cols; j++)
    {
        gradient_mag.at<uchar>(i,j) = sqrt( pow(gradient_x.at<uchar>(i,j),2) + pow(gradient_y.at<uchar>(i,j),2) );  

         if(gradient_mag.at<uchar>(i,j) >250)
            gradient_f.at<uchar>(i,j) = 255;
        else
            gradient_f.at<uchar>(i,j) = 0;
    }
}


/*
imshow("grad x",gradient_x);
waitKey(0);

imshow("grad y",gradient_y);
waitKey(0);
*/

imshow("grad magnitude",gradient_f);
waitKey(0); 

int max=0;


// Performing Non-Maximum Surpression
float theta; // Calculate intensity gradient vector theta=atan2(Gy,Gx);
Mat nonMaxSupp= Mat(gradient_mag.rows-2, gradient_mag.cols-2, CV_8UC1); //CV_8UC1 is 8-bit single channel image i.e grayscale

for(int i=1; i<gradient_x.rows-1; i++)
{
    for(int j=1; j<gradient_x.cols-1; j++)
    {
        //if(gradient_x.at<uchar>(i,j) ==0) //Arctan Fix
         //   theta = 90;
        //else
            theta = atan2(gradient_y.at<uchar>(i,j),gradient_x.at<uchar>(i,j))*(180/3.14);
            //theta = atan(gradient_y.at<uchar>(i,j)/gradient_x.at<uchar>(i,j))*(180/3.14);
        //cout<<theta<<endl;
        //if(theta>max)
        //    max=theta;
        nonMaxSupp.at<uchar>(i-1, j-1) = gradient_mag.at<uchar>(i,j);

        // For horizontal edge
        if(((-22.5 < theta) && (theta <= 22.5)) || ((157.5 < theta) && (theta <= -157.5)))
        {
            if ((gradient_mag.at<uchar>(i,j) < gradient_mag.at<uchar>(i,j+1)) || (gradient_mag.at<uchar>(i,j) < gradient_mag.at<uchar>(i,j-1)))
                nonMaxSupp.at<uchar>(i-1, j-1) = 0;
        }

        //For vertical edge
        if (((-112.5 < theta) && (theta <= -67.5)) || ((67.5 < theta) && (theta <= 112.5)))
        {
            if ((gradient_mag.at<uchar>(i,j) < gradient_mag.at<uchar>(i+1,j)) || (gradient_mag.at<uchar>(i,j) < gradient_mag.at<uchar>(i-1,j)))
                nonMaxSupp.at<uchar>(i-1, j-1) = 0;
        }

        // For 135 degree or -45 degree edge
        if (((-67.5 < theta) && (theta <= -22.5)) || ((112.5 < theta) && (theta <= 157.5)))
        {
            if ((gradient_mag.at<uchar>(i,j) < gradient_mag.at<uchar>(i-1,j+1)) || (gradient_mag.at<uchar>(i,j) < gradient_mag.at<uchar>(i+1,j-1)))
                nonMaxSupp.at<uchar>(i-1, j-1) = 0;
        }

        // For 45 Degree Edge
        if (((-157.5 < theta) && (theta <= -112.5)) || ((22.5 < theta) && (theta <= 67.5)))
        {
            if ((gradient_mag.at<uchar>(i,j) < gradient_mag.at<uchar>(i+1,j+1)) || (gradient_mag.at<uchar>(i,j) < gradient_mag.at<uchar>(i-1,j-1)))
                nonMaxSupp.at<uchar>(i-1, j-1) = 0;
        }

    }

}
//cout<<endl<<"max"<<max;
imshow("Non-Maximum Surpression",nonMaxSupp);
waitKey(0);

return 0;

}

4

0 回答 0