Abstract

This paper proposes an image matching method that is robust to  illumination variation and affine distortion.  Our idea is to do image matching  through establishing an imaging function that describes the functional  relationship  relating  intensity values between  two  images.   Similar  methodology has been proposed by Viola [11] and Lai & Fang [6].  Viola  proposed to do image matching through establishment of an imaging function  based on a consistency principle.  Lai & Fang proposed a parametric form of  the imaging function.  In cases where the illumination variation is not globally  uniform and the parametric form of imaging function is not obvious, one needs  to have a more robust method.  Our method aims to take care of spatially non- uniform illumination variation and affine distortion.  Central to our method is  the proposal of a localized consistency principle, implemented through a non- parametric way of estimating the imaging function.  The estimation is effected  through optimizing a similarity measure that is robust under spatially non- uniform illumination variation and affine distortion.  Experimental results are  presented from both synthetic and real data.  Encouraging results were obtained.