Research On Image Fusion Based On Sparse Representation

Image fusion is a process that integrates complementary information of the multiple source images captured by multiple sensors at the same time or different time.Compared with a single sensor,multi-sensor image fusion can get more accurate information of image.With the development of sensor technology,image fusion technology has been used widely in various fields,such as remote sensing,military,medical,transportation,industry and so on.The multi focus image fusion technology solves the problem that the camera is difficult to make the different objects in the focusing state by use specific methods to integrates complementary information of the multiple images.Multimodality image fusion includes infrared and visible image fusion,medical image fusion and so on.Because of different modality image have own advantages and disadvantages,So fuse multimodality image makes up for the disadvantage of insufficient information displayed by a single modality image under some conditions.Multi focus image fusion and multi modality image fusion are two important branches in the field of image fusion.The fused image provides a clearer,more comprehensive and reliable image,making it more suitable human perception and machine vision,so image fusion technology attracts more and more researchers' attention.In recent years,there are many algorithm of image fusion with sparse representation have been proposed,because of the superiority of sparse representation,but these algorithm have some problem.In order to further enhances quality of fused image,we propose different image fusion algorithms based on sparse representation,according to characteristics of multimodality image and multi-focus image.The main research work is as follows:1.In the past few years,many methods have been proposed in spatial and transform domain for multi-focus image fusion.Considering limitation of them,we utilize multi-decision fusion algorithm to fuse multi-focus image.Firstly,the source images data are classified into the base and detail data.Then,the base data are merged with weighted average accord to focus measure,while the detail data are merged with decision map which could be obtained by sparse representation.Finally,the fused image is obtained by combining the fused base data and the fused detail data.Numerous experiments demonstrate that the proposed fusion algorithm can retainmore sharpness information of source images to fusion image.2.In this paper,we proposed the algorithm of multimodality image fusion based on visual saliency and sparse representation.Firstly,the source image is filtered by Gauss filter,and then the saliency detection and sparse representation theory are applied to fuse the low frequency and high frequency coefficients according to their characteristics,finally the fusion image is obtained by inverse process.Experimental results show that the proposed fusion scheme has better visual characteristics than the fused image obtained by the compared experimental scheme.At last,the author summarizes the work of this paper,and makes a prospect for future research.