基于块分类稀疏表示的图像融合算法

针对基于单一字典训练稀疏表示的图像融合算法忽略图像局部特征的问题，提出了基于块分类稀疏表示的图像融合算法。此算法是根据图像局部特征的差异将图像块分为平滑、边缘和纹理三种结构类型，对边缘和纹理结构分别训练出各自的冗余字典。平滑结构利用算术平均法进行融合，边缘和纹理结构由对应字典利用稀疏表示算法进行融合，并对边缘结构稀疏表示中的残余量进行小波变换融合。实验结果证明，该算法相对于单一字典稀疏表示算法，在融合图像的主观评价和客观评价指标上都有显著改进，并且算法速度也有提高。

Image Fusion Algorithms Based on Sparse Representation of Classified Image Patches

To solve the problem of the fusion algorithm based on sparse representation ignores the local characteristics of the image, an algorithm based on sparse representation of classified image patches is proposed in this paper. In this method, image patches are divided into the smooth, the edge and the texture categories according to local features of the image. The edge and texture structure patches are applied into training the corresponding redundant dictionary. During the fusion process, arithmetic average approach is used for smooth structure patches while edge and texture structure patches are fused by sparse representation algorithm using their corresponding dictionary, and the residual images of the sparse edge structure are fused by a wavelet fusion method. Experiment results show that the proposed algorithm significantly improves the subjective performance and objective performance indexes of fused image and has faster speed than other single dictionary methods.