Edge detection of a specific object

I have a picture of a cow with a wide background. Now I want to detect the edge of the cow only. Is it possible?

target image : http://s9.postimage.org/iltkypjhr/cow.jpg

Any help is appreciated.

Answer 1

[Image Analyst]

For only that cow in only that image, or for any cow(s) in any possible image? Is this for CBIR (content based image retrieval)? You probably won't be able to get the exact cow outline as if you had hand traced it, particularly if the white part of the cow is adjacent to something else white, like snow or a mountain or another cow's spot.

[Sayak]

Yes it is for CBIR [not context]. And it is more good if I am able to detect the edges for the cow(s) in different images also.But as you told its tough to get the exact edge of the cow then is it possible to crop the cow only? like the following picture:

http://s21.postimage.org/6kd2muuk7/cow_Target.jpg

[Image Analyst]

No. You get the bounding box only after you have all the other edges. Some CBIR methods I've seen are quite sophisticated and may be able to do that, but I'm not familiar with the details of how they do it. The methods are complex, and it's not really my field.

[Sayak]

OK. Then can you provide me any link that actually works on CBIR based algorithms. And what if I subtract the background and try it. Will going to work?

Answer 2

[Jeff E]

The Berkely Segmentation Dataset and Benchmark website: http://www.eecs.berkeley.edu/Research/Projects/CS/vision/bsds/

is a resource you should check out. It contains a wide variety of segmentation algorithms run against a common set of ground truth images, and ranks them on accuracy. Some of them have been implemented in Matlab and can be run without too much trouble...others not so much.

[Sayak]

Thanks. Let me go through it properly first.