Research Of Multi-sensor Image Fusion Based On Matching Pursuit

Multi-sensor image fusion defines as the synthetic process of fusing the visual information obtained from multi-sensors surveying the same scene into a new, single fused image.The fused image can be more suitable for human perception or computer processing. Multi-sensor image fusion makes use of the complementary and redundancy information provided by different sensors to improve the reliability of the system and the image information efficiency. It can also achieve target detection, feature extraction and recognition with better results.Signal sparse decomposition becomes popular in the study of signal processing, and has been applied in the areas of data compression, signal feature extraction, time-frequency analysis,and so on.This paper focuses on the application of signal sparse decomposition in the area of image fusion,and present a new tree based image fusion algorithm based on Matching Pursuit.The paper analyses the characteristic of the sparse decompositon, and introduces the algrithm of sparse decomposition based on Matching pursuit. The algorithm is easy to understand and realise,while it has the huge comutational cost which brings great difficulty to the practical application of the image processing. This paper presents a matching pursuit algorithm based on tree-structure redundancy dictionary. The algorithm proposes a structuring strategy first,which basically groups similar atoms and molecules from bottom to top recursively, leading to a tree structure. It can decrease the search space of the best matched atoms, and reduce the computation of the matching pursuit algorithm greatly.A new pursuit algorithm based on the tree structure redundancy dictionary is also presented to realize the sparse decomposition of the images rapidly and effectively. Taking advantage of the sparse decomposition result, the image fusion is completed. The algorithm is also simulated by Matlab. Through the subjective visual effect and objective experimental data, the presented algorithm is analysed and evaluated.The experimental results verify the efficiency of the proposed algorithm.