区分正常和不规则形状的斑点

Question

我有一组连接的斑点，我想区分更规则的（a,c和d）与更不规则的区域，如b。

我尝试过使用卷积区域（在blob_area/convolved_blob_area上设置阈值），shape factor and roundness，但没有一个可以很好地区分d和香蕉形状。你会建议使用哪些参数？谢谢！那我脑子里浮现

Answer 1

一个想法是，你可以使用number of corners每size of blob确定regularity/irregularity。测试结果似乎也符合我们的假设。下面的代码 -

im = imread(input_image_path); 
bw= im2bw(im); 

%// Parameter for cutting into four slices into the third dimsensions 
%// corresponding to the four objects 
common_width = 270; 

%// Threshold to decide between regular and irregular ones 
factor1_th = 0.01; 

bw1 = bw(:,1:common_width*floor(size(bw,2)/common_width)); %// Cropped image 
objs =reshape(bw1,size(bw1,1),common_width,[]);%//Objects stored as dim3 slices 
for objc=1:size(objs,3) %// Object counter 
    disp(['-------------- Processing Obj #' num2str(objc)]); 
    obj = objs(:,:,objc); 
    corners = corner(obj); 
    factor1 = size(corners,1)/nnz(obj) 
    if factor1 > factor1_th 
     disp('This is an irregular one.'); %//' 
    else 
     disp('This is a regular one.'); %//' 
    end 
end 

输出 -

-------------- Processing Obj #1 
factor1 = 
    0.0050 
This is a regular one. 
-------------- Processing Obj #2 
factor1 = 
    0.0109 
This is an irregular one. 
-------------- Processing Obj #3 
factor1 = 
    0.0052 
This is a regular one. 
-------------- Processing Obj #4 
factor1 = 
    0.0078 
This is a regular one. 

如果有人有兴趣运行的代码，这里有A，B，C，D去除的符号输入图像 -

链接 - http://i.stack.imgur.com/uPpUU.jpg