Image Fusion Algorithm Based On Shearlet Transform And Block Matching

Image fusion is the process of synthesizing two or more images collected by the same or different sensors in the same scene into one image.Through this synthesis,important information from multiple sensors can be filtered and integrated,resulting in a more comprehensive,accurate and rich scene description after the fusion.Image fusion can not only overcome the limitations of a single sensor in image acquisition,so as to preserve richer textures and detailed information in multiple images,but also improve the comprehensibility of the image,and provide more reliable image preparations for the subsequent applications in the field of computer vision.Therefore,in order to further enhance the effect and efficiency of image fusion,this paper proposes an image fusion framework combining spatial domain and transform domain by using shear wave transform and 3D image block matching algorithm,aiming at studying the new image fusion algorithm.Firstly,the image fusion algorithm of image block matching is studied in this thesis.Based on the traditional transform domain fusion framework,with reference to the idea of 3D block matching de-noising algorithm(BM3D),the input image is preprocessed in the spatial domain,where the processing steps of block and similarity grouping are added,thus making the original image block’s inner similarity into the form of three-dimensional array so as to conduct the 3D transform,and then continue into the transform domain to conduct subsequent fusion.In the experimental part,this thesis uses the three-dimensional transformation with uses different multi-scale transformation methods and some common fusion rules to process and compare the source images.Experiments results show that the subjective visual effect and the objective evaluation index of the fusion image obtained by the algorithm framework are improved.Secondly,a multi-focus image fusion algorithm based on non,subsampled shearlet transform and image block matching is studied in this thesis.According to the characteristics of multi-focus images and the features of human’s vision,a fusion rule design based on the Pulse Coupled Neural Network(PCNN)and Sum-modified-Laplacian(SML)is proposed.Experiments results show that the fusion rule algorithm can be combined with the image block matching algorithm very well so as to further improve the multi-focus image’s fusion effect.Finally,the medical image fusion algorithm based on the above fusion frame is studied in this thesis.According to the imaging characteristics of medical images and combining the fusion rules of the second-generation non-dominated sorting genetic algorithm(NSGA-Ⅱ),it proposes a multicolor medical image fusion algorithm which uses IHS color model and non-subsampled shear wave.Experiments results show that the algorithm has some improvement in subjective and objective evaluation indexes.