Image match using wavelet — Colour SIFT features

Abstract

This paper presents a new technique for robust Image matching based on the combination of wavelet transform and Colour-SIFT (Scale Invariant Feature Transform). Low frequency and High frequency sub-bands of the image are extracted using wavelet transform and then colour SIFT method is used to extract colour feature descriptors. A wavelet basis having maximum statistical correlation with an image is chosen from a library of wavelets. A novel matching technique is used to validate true matches by establishing relationships between candidate matching features. The proposed method is tested on many image pairs with viewpoint changes, illumination changes and color information. Performance of this method is compared with Hue-SIFT and basic SIFT methods.