Bilinear equations and fuzzy image comparison

Abstract

We present an image comparison method based on the greatest solution of a system of bilinear fuzzy relation equations A·x=B·x, where “·” is the max-min composition, A and B are the compared images, normalized in [0,1] and considered as fuzzy relations, and x is an unknown vector. Due to symmetry, A (resp. B) could be the original image and B (resp. A) is an image modified of A (resp. B) (for instance, either noised or watermarked). Our index is more robust than other two comparison indexes already known in literature.