A wavelet and canny based image comparison

Abstract

A robust and reliable image comparison method enables many computer vision applications such as wireless surveillance systems, fingerprint and face recognition. In this paper, we propose an image comparison method that can indicate reliably the similarity and difference between two images. It is constructed by combining an image quality measure and a simple Canny based image comparison metric. Our experiment results show that it is tolerant to random noise, structured noise and invariant to the affine transformation.