改进的基于稀疏表示的多聚焦图像融合  

An Improved Multi-focus Image Fusion Based on Sparse Representation

作　　者：首照宇[1] 胡蓉[2] 欧阳宁[1] 

机构地区：[1]桂林电子科技大学认知无线电与信息处理省部共建教育部重点实验室,广西桂林541004 [2]桂林电子科技大学信息与通信学院,广西桂林541004

出　　处：《电视技术》2014年第7期13-16,共4页Video Engineering

基　　金：国家自然科学基金项目(61363069;11271388);广西自然科学基金项目(2013GXNSFDA019030;2013GXNSFAA019331;2012GXNSFBA053014;2012GXNSFAA053231;2011GXNSFA018158);广西科技开发项目(桂科攻1348020-6;桂科能1298025-7);广西教育厅重点项目(201202ZD040;201202ZD044;2013YB091)

摘　　要：针对基于稀疏表示(Sparse Representation,SR)的融合算法存在冗余性和残余量丢失,提出一种改进的基于稀疏表示(Improved Sparse Representation,ISR)的多聚焦图像融合算法。该算法通过对源图像块去平均后再进行稀疏分解,解决了信息冗余问题;对稀疏表示中的残余量进一步采用小波变换进行融合,保证了信息的完整性;采用加权平均的融合规则,从而获得更多的信息量。实验结果表明,与其他的方法相比,本文提出的算法能实现更好的性能。In view of the fusion algorithm based on sparse representation exists redundancy and residual. In this paper, an improved method based on sparse representation is proposed to overcome this disadvantage. In this method, the source image patches are decomposed sparsely after moving the aver- age value, and the problem of information redundancy is resolved. Then the residual images are continued to be fused through a wavelet fusion method,it ensures the integrity of information. The coefficients are fused using the weighted average rules in which can get more information. The experiments show that the proposed algorithm can achieve better performance compared with other methods.

关 键 词：稀疏表示 小波变换 多聚焦图像融合 融合规则 

分 类 号：TN919.85[电子电信—通信与信息系统]