Connected component labeling
[L, n] = bwlabel(img [, nhood])
img, with the pixels of each connected object having the same number. The numbers vary from 1 to N, where N is the number of connected objects. The background is numbered 0.max(L).Function bwlabel numbers all the objects in a binary image. One common application is to filter out objects that have less than a certain ammount of pixels. See the examples.
You can use the Scilab find function in conjunction with
bwlabel to return vectors of indices for the pixels that
make up a specific object. For example, to return the coordinates
for the pixels in object 3:
[r,c] = find(bwlabel(BW)==3) |  |  |
// // EXAMPLE 1 // Img =[0 0 0 0 0 1 1 0 1 1 0 0 1 1 0 1 1 0 0 1 1 0 0 0 1 0 1 1 0 0 0 1 0 1 1 0 0 0 1 0 1 1 0 0 1 1 0 1 1 0 0 0 0 0 1 1]; L = bwlabel(Img,4) // Objects 2 and 3 are connected if 8-connectivity is used: L = bwlabel(Img) // default: 8-connectivity [r,c] = find(L==2); rc = [r' c'] // coordinates of object 2! // // EXAMPLE 2 // xset('auto clear', 'on'); a = gray_imread(SIPDIR + 'images/disks.bmp'); // Add some noise // a = imnoise(a,'salt & pepper'); a = 1-a; imshow(a,2); // convention: objects are white(1) // Label every connected component with a unique number. // [L, n] = bwlabel(a); // Shows each component with a different color // imshow(L+1, rand(n+1,3)); // Get one specific region (probably a single noise point) reg = (L == 300); imshow(reg*1, 2); // Eliminate regions smaller than 100 pixels (noise) // and those larger than 1000 pixels (cluttered disks) for i=1:n f = find(L==i); // linear coordinates of i-th region reg_size = size(f,'*'); if reg_size < 100 | reg_size > 1000 L(f) = 0; // merge small regions with the background end end imshow(L+1, rand(n+1,3)); // note how the small regions are gone // Just as a side-activity, let's fill the unwanted holes: bw = 1*(L>0); // binarize the image imshow(bw,2) bw = dilate(bw); bw = erode(bw); imshow(bw,2); // every hole is now filled xset('auto clear', 'off'); | | |
We use a simple stack-based flooding implementation written in C, but there exist many faster algorithms. The flood/fill region growing process may be found in most books of imaging science. For instance:
"Shape Analysis and Classification", L. da F. Costa and R. M. Cesar Jr., CRC Press, pp. 335-347.
Example of fast algorithm (not implemented):
Haralick, Robert M., and Linda G. Shapiro, Computer and Robot Vision, Volume I, Addison-Wesley, 1992, pp. 28-48.
http://siptoolbox.sourceforge.net