shape features for the segments?

hi i have to write a program for shape features for the segments with top 4 largest areas. and i dont know how to do this .

plz help me.

Answer 1

[Image Analyst]

See my Image Segmentation Tutorial: http://www.mathworks.com/matlabcentral/fileexchange/?term=authorid%3A31862

My function that extracts the N biggest or smallest blobs is included in the demo I attached below. Once you have the 4 largest regions you can then call regionprops to measure things in them like the area, perimeter, equivalent circular diameter, etc.

[SAM]

thanks for reply. can u tell me how could i do this with graph cut method. i wants to keep K=5.

basically i have to write for this equation.

G=(1/ area of regioni )sum(k)

where i ( i = 0, 1, 2,4 ) is the index of the largest three regions. (k) is the normalized coordinates (normalized by the width of the image) . plz reply me..

[Image Analyst]

Attach your image and I'll see what I can do, if I have time.

[Image Analyst]

Sorry, I don't know that method. http://matlab.wikia.com/wiki/FAQ#Can_you_program_up_the_algorithm_in_this_article_for_me_and_explain_it_to_me.3F

[SAM]

any way thanks for your help.........

[SAM]

hi can plz tell me a simple way of writing graphcut program with K-means method.

i.e an image divide into 6 parts (cluster).

plz reply.

[Image Analyst]

Steve Eddins has a demo for the k-means method for color segmentation: http://www.mathworks.com/products/demos/image/color_seg_k/ipexhistology.html

[SAM]

i have already try this but the problem with this code is it converting rgb to Lab and i wanted to use rgb image. moreover, its only valid for K=3.