Enhancing SURF Feature Matching Using Colour Histograms

Abstract

A strategy is proposed that enhances the local feature matching capabilities of the SURF descriptor by utilising colour histograms. The results compare variations of the RGB, HSV and Opponent colour spaces on a dataset of image pairs that undergo illumination, viewpoint and translational changes. This study finds the most appropriate colour space that enhances the distinctiveness of a descriptor when applied to the matching of corresponding features in arbitrary image sets.