Research And Application Of Image Fusion Algorithm Based On Sparse Representation

The sparseness coefficient of the image can describe the characterization information of the image powerfully.Therefore,many scholars focus on the sparse coding theory to construct the fused image efficiently.In order to improve the quality of fused images,this paper focus on image fusion technology based on sparse coding theory.This article describes the background,status and methods of the fusion technology.we also have done some research to improve the defects of existing methods.The main research work is as follows:(1)For the problems that the low-frequency subband coefficients in the NSCT domain do not have high sparsity and sparse representation in the spatial domain is prone to generate a large number of square,this article designs an image fusion algorithm based on NSCT-SR-PCNN by combining three models which were non-subsampled contourlet transform(NSCT),sparse representation(SR)and pulse coupled neural network(PCNN).The low-frequency subband fusion uses Orthogonal Matching Pursuit(OMP)to solve the low-frequency subband coefficients according to the K-SVD's redundant dictionary.Next,three fusion rules and three kinds of fusion mean matrices are designed based on the norm values.According to the characteristics of PCNN model,high-frequency subband fusion is designed with spatial frequency as input,Laplasse energy,visibility and standard deviation are linked intensity values respectively to obtain the ignition map,and a weighted function is designed to obtain a new ignition map.(2)For the problem that "sliding window" technology which would disturb the correlation of image data,an image fusion algorithm based on NSCT-CSR-PCNN is proposed by using the convolutional sparse representation model to model the entire low-frequency subgraph.Last,two kinds of low frequency subgraph fusion rules for multimodal and non-multimodal are proposed in order to prove the applicability of NSCT-CSR-PCNN image fusion algorithm.