i am developing a content based image retrieval system. now i want to recognize the shape from a region segmented image in form of a matrix. i am working on jpeg format. when i used compactness parameter ie perimeter*perimeter/area of the region, the results are at times correct and at times incorrect. could u suggest any other way to identify the regions.
what i could find is the problem is of compression and decompression. is there any method to solve this????