四川大学学报（自然科学版）2012，Vol.49Issue(3)：570-576,7.DOI:10.3969/j.issn.0490-6756.2012.03.017
基于广义邻域结构相似度的红外和彩色可见光图像融合
Infrared and color visible images fusion algorithm based on the generalized neighborhood structural similarity
摘要
Abstract
Due to the interchannel non-orthogonality for IHS color space, color distortion is more evident in the IHS image fusion methods. Notice that ια β color space is perceptually decorrelated and the advantages of Nonsubsampled Contourlet Transform (NSCT) whose coefficients have strong dependency among different scales, directions and positions. Therefore, the authors propose a novel fusion algorithm for infrared and color visible images. First, the authors conduct remapping on infrared image with the achromatic channel of color visible image, and then compute the NSCTs of both the remapped and achromatic channel images. Second, the authors exploit the fusion rules, which for approximate coefficients contain the different rules and which for details coefficients are combined the coefficients energy in the neighbourhood with structural similarity among brother subbands in the same scale. After reconstructing the image from the fused NSCT coefficients, the authors remap achromatic channel of the fused image. Finally, the fused color image is transformed from ια β to RGB color space. Experiments indicate that the proposed approach, in target detectability and color distortion aspects, is better than the existing IHS fusion methods, and to a certain extent, can suppress the smoke influence on the targets. Moreover, the resulted image accords with human visual perception while reflecting the effectiveness of the algorithm.关键词
彩色图像融合/非下采样Contourlet变换/邻域特性/结构相似度Key words
color image fusion, nonsubsampled contourlet transform, neighbourhood characteristics, structural similarity分类
信息技术与安全科学引用本文复制引用
薛琴,范勇,李绘卓,王俊波,熊平,唐遵烈..基于广义邻域结构相似度的红外和彩色可见光图像融合[J].四川大学学报（自然科学版）,2012,49(3):570-576,7.基金项目
国家自然科学基金-中国工程物理研究院联合基金资助项目(10676029,10776028) （10676029,10776028）
