Research On Multi-focus Image Fusion Algorithm Based On Decomposition Theory

Due to the physical characteristics of the sensor itself,the imaging mechanism and the angle of observation,and the sensor itself cannot complete the fusion of two images,a single sensor cannot obtain a comprehensive description of a scene.Image fusion can solve this problem well and has become a hot topic of information fusion.It combines useful information of two or more images into one image,and the merged image has high utilization and reliability.Fusion image can be easily observed by the human eye and further processing by the computer.As an important part of image fusion,multi-focus image fusion technology overcomes the influence of the depth of field of the optical system,extracts the clear parts of different images and finally merges them into a clear image.The traditional method is difficult to extract the focus area in the source image,so the image quality after fusion is not ideal.In order to overcome the shortcomings of traditional fusion methods,this paper based on the existing algorithms in the image space domain research,the research content is as follows:(1)In this paper,it put forward a multi-focus image fusion algorithm that based on total variation and quad-tree decomposition,which can deal with the problem of block effect well.In terms of the definition detection for the source image,the traditional sum-modified-laplacian only consider the horizontal and vertical directions of the image,this paper improved the method of the traditional focus region detection.The algorithm firstly gets the texture image and the structure image by total variation,and then decomposes the texture image with quad-tree to get the optimal partition of the texture image.The modified focus region detection function is used to detect each sub-block to obtain the initial decision diagram.The experimental results show that the mutual information and space frequency of the proposed algorithm are the highest,with the mean values of 6.7225 and 16.5125 respectively,and the t ime complexity is within 4 seconds,which is suitable for real-time processing systems.(2)In this paper,the RPCA theory is applied to the multi-focus image fusion algorithm.Firstly,some source images are decomposed on RPCA to get their sparse matrices,and then executing the operation to the sparse matrices by using focusing detection function and the sliding window technique,which can get the gradient energy matrix,what's more,the gradient energy matrix is further treated by median filtering,then get the smoothing matrix.In addition,the difference matrix is obtained by the subtraction between the gradient energy matrix and the smoothing matrix,and this difference matrix can determine the position of the focus region.The experimental results show that the mean values of the four indicators of the algorithm are 5.9675,6.8325,16.38,and 20.945 seconds,respectively,which is suitable for the fusion system with high requirements for low-integration requirements.