Pan-sharpening of multi-spectral images using over-complete rational-dilation wavelet transform

Abstract

A rational-dilation wavelet transform (RWT) based pan-sharpening method for multi-spectral (MS) images of various oscillatory nature is proposed. The previous multi-scale transforms, such as wavelets, curvelets and contourlets, decompose an image into channels with low constant Q-factors, and aren't suitable for pan-sharpening images with different behavior in frequency domain. The RWT as an over-complete scheme not only increases the sampling in spatial and frequency domain, but also provides a tunable Q-factor approach to be suitable for a given dataset. We studied its multi-scale decomposition scheme and the RWT based pan-sharpening method. The MS image pan-sharpening experiments show that this method using a better suitable parameter set can achieve a promising performance and often outperforms many other widely-used pan-sharpening methods both in visual quality and in term of evaluation indexes.