Skeletonization of a Grayscale image
in [Matlab]
|
From: Sally on 9 Feb 2010 12:44 Hi, Is there any way of skeletonizing a grayscale image in matlab? I know bwmorph can be used for binary images, but I need to skeletonize a grayscale image. I think this is possible in C++ but is it possible with matlab? Thank you for your help in advance. Sally.
From: Ashish Uthama on 9 Feb 2010 12:49 On Tue, 09 Feb 2010 12:44:04 -0500, Sally <sal_sis(a)hotmail.com> wrote: > Hi, > > Is there any way of skeletonizing a grayscale image in matlab? I know > bwmorph can be used for binary images, but I need to skeletonize a > grayscale image. I think this is possible in C++ but is it possible with > matlab? > > Thank you for your help in advance. > > Sally. Could you extract a binary image from your grayscale image which represents the object of interest?
From: ImageAnalyst on 9 Feb 2010 13:11 No. It's not possible. It doesn't make sense. Not unless you're using some non-standard definition of skeleton, like you mean watershed (http://cmm.ensmp.fr/~beucher/wtshed.html) but you're calling it "skeleton" by mistake. You can't do it in C++ or any other language either. It's only possible to get skeletons of 2D binary images (or medial surfaces of 3D binary images or perhaps 3D skeletons for special cases of 3D binary images). It is possible to find ridgelines and valley-lines in gray scale images, but you wouldn't call those skeletons. Maybe that's what you mean. So, what is your definition of skeleton? Why do you think it might be possible?
From: Sally on 9 Feb 2010 13:47 ImageAnalyst: I am basing this on the following (simplified) algorithm which is found in a journal article here: http://www.wise-t.com/ias/Magazine/2008/3/08_Gavet.pdf Input is a grayscale image 2 I = Filter on Input 3 R = Input-I; 4 C = closing of R 5 L = Supremum of Openings of C by segments; 6 C2 = closing of L; 7 S = skeletonization of C2 8 DM = Distance Map of S; 9 markers = Regional Maxima of DM; 10 segmentation = watershed constrained by markers on inverse of DM; Its at step 7 that I am stuck. I hope this makes it clear what I am trying to implement. I would appreciate your help Sally
From: ImageAnalyst on 9 Feb 2010 14:57
Sally: Nice paper with impressive results. Unfortunately they don't give the resulting image after each stage of the algorithm, nor do they describe exactly what skeletonization of the "closed" image means. I suspect that there is an implicit thresholding in step 7 to get a binary image of the cell walls, then they skeletonize that. (By the way, there is an error in the paper when the Figure 11 images caption says that the skeletonization was the result of line 6 of their algorithm when the algorithm clearly says that the skeletonization occurs at step 7.) |