A novel approach to image matching using convex layers

Abstract:- This paper presents a novel algorithm for image matching, i.e. comparing them in order to obtain the similarity measure. We extract geometric features of images by binarizing them and calculating a set of convex layers. These serve as unique features for consequent image comparison and similarity measurement. Our approach has linear computational complexity, and copes with rotated, translated, and scaled images, while it is also rather robust up to moderate signal-to-noise ratios