Image Fusion Using Non-Subsampled Contourlet Transform Based on Activity and Local Gradient Energy Methods

Abstract

It is often difficult for a single image to obtain all the details of the same scene. To handle this problem, multiple images can be acquired through a variety of ways, and then the obtained images can be typically combined into one image by image fusion technology. For improving image fusion quality, a new image fusion method based on Non-Subsampled Contourlet Transform (NSCT) is proposed. Source images are initially decomposed via NSCT, the low frequency sub-band image and a series of high frequency sub-band images with different directions and different scales are obtained, low frequency sub-band image fusion is carried out based on the activity with local sharpness changes. In the high frequency region, fusion is carried out based on local gradient energy with edge strength, and finally the fusion image is reconstructed by NSCT inverse transform. Through the fusion experiments of multi-focus image and multimodal images, the proposed method is compared with the other methods such as NSCT_PC, NSCT_EN_PCNN, NSST_PCNN, CWT_SR and JBF. Visually, the fusion image obtained by the proposed method is clearer in detail and has a stronger sense of image hierarchy. Using objective evaluation such as peak signal to noise ratio, structural similarity, edge information retention, and information entropy, they are improved by at least 0.4%, 0.02%, 9.7%, and 1.4% respectively. Generally speaking, the method in this paper retains more important details and shows better fusion performance.