边缘检测
-
利用梯度算子进行边缘检测
-
BW =edge( I,type,thresh,direction,‘nothinning’)
-
type: ‘roberts’、‘sobel’、‘prewitt’(算子选择)
-
thresh:敏感度预置参数。灰度低于预置的边缘将不会被检测到。(系统自动设定参数)
-
direction:‘vertical’、‘horizontal’、‘both’
-
‘nothinning’:无细节处理,默认’thinning’
I = imread('C:\Users\Lenovo\Desktop\lena.jpg'); I= rgb2gray(I); BW1 = edge(I,'roberts'); BW2 = edge(I,'sobel'); BW3 = edge(I,'prewitt'); figure subplot(2,2,1),imshow(I),title('orignal') subplot(2,2,2),imshow(BW1),title('roberts') subplot(2,2,3),imshow(BW2),title('sobel') subplot(2,2,4),imshow(BW3),title('prewitt')
-
-
带方向的边缘检测:Robinson算子
-
N=[1,2,1 0,0,0 -1,-2,-1]; edge_n = imfilter(I,N,'symmetric','conv');
-
高斯拉普拉斯算子(LoG):(LoG:低通滤波器)
- 应用LoG算子处理图像
- 检测图像中的零交叉点
- 筛选掉那些不满足条件的点
- BW=edge(I,‘log’,thresh,sigma)
-
高斯差分算子(DoG):
- 步骤和LoG一样,DoG相当于一个带通滤波器
I = double(rgb2gray(imread('C:\Users\Lenovo\Desktop\lena.jpg'))); figure,imshow(uint8(I)) DoG=fspecial('gaussian',5,0.8)-fspecial('gaussian',5,0.6); ImageDoG=imfilter(I,DoG,'symmetric','conv');%构建DoG算子 figure, imshow(ImageDoG) % threshold = 2 proc_Img1 = ImageDoG; proc_Img1(find(proc_Img1 < 2))=0;%选择了不同的阈值 figure, imshow(proc_Img1) % threshold = 3 proc_Img2 = ImageDoG; proc_Img2(find(proc_Img2 < 3))=0; figure, imshow(proc_Img2)
-
Canny边缘检测
- Canny 准则:
- 好的检测效果:对边缘检测的错误尽可能低
- 边缘的定位要准确
- 同一边缘要有低的相应次数
- Canny边缘检测算法:
- Canny实现:BW=edge(I,‘canny’,thresh.sigma)
- Canny 准则: