A Robust Pan-Sharpening Scheme for Improving Resolution of Satellite Images in the Domain of the Nonsubsampled Shearlet Transform

Abstract

Pan-sharpening is a substantial post-processing task for captured multispectral remotely sensed satellite imagery. Its main purpose is to fuse the high spectral characteristics of the multispectral (MS) images with the high spatial information of the panchromatic (Pan) image to output a sharper MS image (pan-sharpened) that encompasses higher spectral and spatial resolutions. In this paper, we investigate a conception of a new pan-sharpening scheme using the pulse coupled neural network (PCNN) in the nonsubsampled shearlet transform (NSST) domain. This can be done based on two main steps. In the first step, the input MS and Pan images are individually decomposed into multi-scaled and multi-directional coefficients by NSST. Second, the PCNN is applied to the low-frequency coefficients, which are merged by a weighted firing energy fusion rule utilizing the PCNN firing times. The detail coefficients with higher matching value are chosen to be the fused detail coefficients. Lastly, the pan-sharpened image is generated by the inverse NSST. WorldView-2, GeoEye-1, and QuickBird satellite datasets are employed in the experiments which demonstrate that the investigated scheme gained the ability in preserving both the high spatial details and high spectral characteristics simultaneously without involving abundant computation time. In addition, various image quality metrics such as CC, RMSE, RASE, ERGAS, SAM, Q4, QNR, and SCC are adopted to assess the spectral and spatial qualities of the pan-sharpened image. The experimental results and performance analysis illustrated that our scheme improved performance efficiency and achieved superiority over other conventional techniques.