Automatic detection of interest points

Hello,

Please, I try to detect automatically a predefined number of interest points in images ( drosophila wing ) such as those provided in the attached document . I tested several methods, Harris detector, mathematical morphology ... but unfortunately I have not found a good result.

Please if you have any idea let me know (websites, courses, ...), what about face recognition methods ?

Thank you

Answer 1

[Image Analyst]

One part of SIFT and SURF (Computer Vision System Toolbox) is detection of "Salient" points. You might look into that. Fingerprint analysis looks more applicable than face recognition.

[Haykel]

Thank you for your reply! I tested your proposal and I can not find a good resulting output. I preprocessed the image and I detected branch points after skeletonization. I detected about 60 points branches per image and I try to take only the specifique 15. So I look for a method to characterize these points (the distance between them, angles) and am looking for how matching these points with a real model.

- How to calculate distance betewen all branchpoints ?

- Is distance de bhattacharyya will be helpful ? - How can I know if there is a path (connexion) between two branchpoins or not?

- How to cearte a model ?

- In the literature I found Pairewise Geometric Histogram (PGH), is there in matlab?

Please if you have any idea let me know (websites, courses, ...).

Thank you !

[Image Analyst]

Then you might want to try this algorithm on your images. Your images look sort of similar to fundus images so it might work. http://www.mathworks.com/matlabcentral/fileexchange/49172-trainable-cosfire-filters-for-vessel-delineation-with-application-to-retinal-images

[Haykel]

Unfortunately, I have not had good results !