Matlab实现SUSAN角点检测

SUSAN（Smallest Univalue Segment Assimilating Nucleus）算法是一种基于像素灰度值的角点检测算法，它可以检测出图像中的角点和边缘。下面是Matlab实现SUSAN角点检测的步骤：

1. 读入图像并转化为灰度图像

img = imread('lena.jpg');
gray_img = rgb2gray(img);

2. 初始化参数

r = 3; % 邻域半径
t = 27; % 阈值

3. 计算像素与邻域内像素的灰度值差异

[m, n] = size(gray_img);
susan_img = zeros(m, n);
for i = r+1:m-r
    for j = r+1:n-r
        % 计算邻域内像素与中心像素的灰度值差异
        diff = gray_img(i-r:i+r,j-r:j+r) - gray_img(i,j);
        % 计算像素灰度值的方差
        var = sum(sum(diff.^2))/(numel(diff)-1); 
        % 计算像素与邻域内像素的灰度值差异的均值
        mean_diff = sum(sum(abs(diff)))/numel(diff); 
        % 计算SUSAN函数值
        susan_img(i,j) = mean_diff/var;
    end
end

4. 根据SUSAN函数值进行角点检测

corners = zeros(m, n);
for i = r+1:m-r
    for j = r+1:n-r
        % 判断SUSAN函数值是否超过阈值
        if susan_img(i,j) > t
            % 标记为角点
            corners(i,j) = 1; 
        end
    end
end

5. 可视化结果

figure;
subplot(1,2,1);
imshow(img);
title('原始图像');
subplot(1,2,2);
imshow(corners);
title('SUSAN角点检测结果');

完整的代码如下：

img = imread('lena.jpg');
gray_img = rgb2gray(img);

r = 3;
t = 27;

[m, n] = size(gray_img);
susan_img = zeros(m, n);
for i = r+1:m-r
    for j = r+1:n-r
        diff = gray_img(i-r:i+r,j-r:j+r) - gray_img(i,j);
        var = sum(sum(diff.^2))/(numel(diff)-1); 
        mean_diff = sum(sum(abs(diff)))/numel(diff); 
        susan_img(i,j) = mean_diff/var;
    end
end

corners = zeros(m, n);
for i = r+1:m-r
    for j = r+1:n-r
        if susan_img(i,j) > t
            corners(i,j) = 1; 
        end
    end
end

figure;
subplot(1,2,1);
imshow(img);
title('原始图像');
subplot(1,2,2);
imshow(corners);
title('SUSAN角点检测结果');