【MATLAB图像融合】[12]canny边缘检测简易版

Canny简介

1、分别计算x,y两个方向的梯度值。
2、由梯度值可以计算出幅值和梯度方向。
3、幅值是梯度的强度，通过梯度方向可以抑制非极大幅值。
4、设置2个阈值。高于高阈值的幅值点，其像素值置为255，低于低阈值的幅值点，该点像素值赋0；而介于两值之间的幅值，检测其是否连通，若连通赋值255，反之赋值0。
5、输出边缘图像。

代码分享

function re=canny(x,min,max)  
    [m,n]=size(x);
    GX=zeros(m,n); GY=zeros(m,n);
    w=1/256*[1;4;6;4;1]*[1,4,6,4,1];
    gaosi_x=conv2(x,w,'same');
    for i=1:m-1
        for j=1:n-1
            GX(i,j)=(gaosi_x(i,j+1)- gaosi_x(i,j)+gaosi_x(i+1,j+1)- gaosi_x(i+1,j))/2;
            GY(i,j)=(gaosi_x(i+1,j)- gaosi_x(i,j)+gaosi_x(i+1,j+1)- gaosi_x(i,j+1))/2;
        end
    end
    
    for t=1:m                                   %梯度计算
        GX(t,n)=GX(t,n-1);GY(t,n)=GY(t,n-1);
    end
    for t=1:n
        GX(m,t)=GX(m-1,t);GY(t,n)=GY(t,n-1);
    end
    
    tfuzhi=zeros(m,n);tpos=zeros(m,n);
    for i=1:m
        for j=1:n
            tem1=(GX(i,j))^2+(GY(i,j))^2;
            tem2=(GX(i,j))/(GY(i,j));
            tfuzhi(i,j)=sqrt(tem1);
            tpos(i,j)=atan(abs(tem2));
        end
    end
    
    pos=zeros(m+2,n+2);fuzhi=zeros(m+2,n+2);   %方向和幅值计算
    for i=2:m+1
        for j=2:n+1
            pos(i,j)=tpos(i-1,j-1);
            fuzhi(i,j)=tfuzhi(i-1,j-1);
        end
    end

    ffuzhi=zeros(m,n);                     %幅值抑制
    for i=2:m+1
        for j=2:n+1
            if (0<=pos(i,j))&&(pos(i,j)<(pi/4))
                tem3=fuzhi(i+1,j-1);tem4=fuzhi(i-1,j+1);
            elseif((pi/4)<=pos(i,j))&&(pos(i,j)<(pi/2))
                tem3=fuzhi(i-1,j);tem4=fuzhi(i+1,j);
            elseif((pi/2)<=pos(i,j))&&(pos(i,j)<3*(pi/4))
                tem3=fuzhi(i-1,j-1);tem4=fuzhi(i+1,j+1);
            elseif(3*(pi/4)<=pos(i,j))&&(pos(i,j)<=pi)
                tem3=fuzhi(i,j-1);tem4=fuzhi(i,j+1);
            end

            if (fuzhi(i-1,j-1)>tem3)&&(fuzhi(i-1,j-1)>tem4)
                ffuzhi(i-1,j-1)=fuzhi(i,j);
            else
                ffuzhi(i-1,j-1)=0;
            end
        end
    end
    B=zeros(m,n);                       %连通检测
    for i=2:m-1
        for j=2:n-1
            if ffuzhi(i,j)>=max
                B(i,j)=255;
            elseif ffuzhi(i,j)<=min
                B(i,j)=0;
              else
                if ffuzhi(i-1,j-1)>max||ffuzhi(i-1,j)>max||ffuzhi(i-1,j+1)>max||ffuzhi(i,j-1)>max...
                    ||ffuzhi(i,j+1)>max||ffuzhi(i+1,j-1)>max||ffuzhi(i+1,j)>max||ffuzhi(i+1,j+1)>max
                    B(i,j)=255;
               end
            end
        end
    end
    re=B;
end

clear;close all;clc;
A=imread('lena.jpg');
A=double(A);
B=canny(A,1,6);
  
subplot 121;imshow(uint8(A));title('源图像');
subplot 122;imshow(uint8(B));title('边缘提取');

运行结果

运行效果感觉有点问题，幅值好像没有被很好抑制，边缘过粗。运行发现错误的话还请斧正，谢谢。