A novel Image Fusion Framework based on Non-Subsampled Shearlet Transform (NSST) Domain

Abstract

As an effective means to integrate the information contained in multiple images in different ways, multi-model image fusion provides more comprehensive and sophisticated information for modern medical diagnosis, remote sensing, video surveillance and other fields. Based on non-subsampled shearlet transform (NSST) domain, this paper proposed an general image fusion framework to well preserve image energy and detail. It can well solve the problem of limited direction in image decomposition. This fusion method applies NSST to decomposition source images into high-pass and low-pass subbands. Sum Modified-laplacian(SML) is employed to fused high-pass bands. A weighted sum of weighted local energy (WLE) and a modified Laplacian (WSEML) based on eight neighborhoods are used as a fusion strategy to fuse low-frequency components. Finally, employing NSST inverse transform to reconstruct the fused high frequency and low frequency components. The experimental results indicate that the framework can save more image details and improve the quality of fused images.