I am using SIFT keypoints atm on matlab to get the keypoints of each image , and then i use a match function that appends 2 images and finds the macthed keypoints between the images , My problem is that the number of keypoints that appears sometimes is very low , When can I say that these images are identical or represent the same object (a money bill paper for example).
Is there a better technique to use to do this??
Firstly, one of the reason your detector returns few keypoints is that your settings for your SIFT detector is not set properly for the type of images you are dealing with. E.g., your images could be lacking in texture/details. Try playing around with the parameter settings first.
Secondly, perhaps a better way for using SIFT for object recognition is the "Bag of Words" (BOW) approach. See this post here for more details.
