image segmentation
in [Matlab]
|
From: ImageAnalyst on 16 Jun 2010 16:22 On Jun 16, 3:57 pm, "Li " <x...(a)hotmail.com> wrote: > Thanks! Okay, I will think about your suggestions... > I also tried other image segmentation methods, which have every pixel labeled. In some cases, it is not a foreground vs background problem. It specifies which patch a pixel belongs to. Also, it seems there is nothing I can do to control the number of segments. Thank you. ---------------------------------------------------------------------------------------------------------------------------- Not true. There are many things you can do to affect the number of labeled objects, both before and after calling bwlabel. You just have to know what to do. For example you can use "marker controlled watershed segmentation" rather than the usual "dumb" watershed. There is a MATLAB demo for this. Here's a nice page on watershed with a variety of examples that you might want to check out: http://cmm.ensmp.fr/~beucher/wtshed.html
From: Li on 16 Jun 2010 17:01 It is a nice page. Thank you! But now I have a very strange problem. Suppose I is a 3D matrix representing a color image. I ran the following code: L = watershed(I); It works well with several thousand labels. But if I ran: J = imhmin(I,20); L = watershed(J); I only got two labels. I: 256*256*3 uint8, min: 5; max: 243; J: 256*256*3 uint8, min: 25; max: 243; When I plotted both images using imshow, they also looked almost the same to me. I don't know what is wrong with this?! Thank you.
From: ImageAnalyst on 16 Jun 2010 18:30
Go with the method that gives you what you want. The first method you tried doesn't work because watershed assumes any 3D image you pass to it is a volumetric monochrome image - a completely different beast than a (N x M x 3) color image. |