Medical Image Fusion Algorithms Based On Multi-scale Analysis

With the development of electronic science and update of microprocessor,hardware and computer equipment,medical imaging technology has made great achievements,and then it assists clinical diagnosis gradually and effectively.Various imaging techniques produce different modal medical images of human,however the single-modal medical image can only display the diagnostic target or peripheral tissue unilaterally and locally,and can't give doctor to provide complete information.In order to solve the problem,multi-modal medical image fusion technology is proposed,which is extracting and integrating information from two or more modal medical images.Moreover it can help clinic diagnose the lesions more intuitive,more comprehensive,more accurate and reliable basis,through combining the complementary message and removing the redundant information of the pictures from the same objective which is in different model.In this paper,according to the characteristics of multi-modal medical image and multi-scale transform,we focus on medical image fusion algorithms based on multi-scale analysis.In this paper,the main research contents are summarized as follows:1)Pointing at the characteristics of multi-scale transform,a novel medical image fusion algorithm based on human visual features and adaptive PCNN is proposed.Firstly,source images after registration are decomposed into low and high frequency sub-bands by nonsubsampled contourlet transform(NSCT).Secondly,majority energy and characteristics of the source image is retained in the low frequency sub-bands,a fusion rule based on region energy(RE)combined with gradient singular value measurement(GSVM)is adopted.Moreover,considering the problem of global image feature,pulse coupled neural network(PCNN)is utilized to fuse the high frequency sub-bands,using the sum of local visual contrast(SLVC)to simulate the human visual feature as the external stimulus input to PCNN,setting the strength connection of PCNN to change with the visual contrast sensitivity adaptively.Finally,the fused image is obtained by performing the inverse NSCT on the combined coefficients.2)According to the characteristics of multi-scale transform,a medical image fusion algorithm based on textural features and information of generalized correlation structure is proposed.Firstly,source images after registration are decomposed into low and high frequency sub-bands by NSCT.Secondly,considering the human visual sensitivity to the textural feature,local differential box counting is compiled statistics of textural information of image;we analyze the strong correlation and structure information of brother' coefficients and father-son coefficients respectively in high frequency sub-bands of NSCT.Moreover,structure similarity in those sub-band and the sum of Laplace energy between of neighborhoods are calculated and used to describe the information of generalized correlation structure.Thirdly,Sigmoid function is proposed on low frequency sub-bands with adaptive fusion,and absolute value rule of information of generalized correlation structure is presented on high frequency coefficients.Finally,the fused image is obtained by performing the inverse NSCT on the combined coefficients.3)Plenty of experiments of fusion of images including gray images and color images are contrasted and analyzed in the dissertation.The experiment results show that the proposed algorithms can keep more outline and texture information and more prominent detailed features of the source image,and can reflect lesions completely with vision clearly.The objective image quality evaluation indexes are compared and analyzed.The algorithms proposed have best datum,and the comprehensive evaluation data is good.It shows that the fusion results contain more information,more scattered gray scale distribution,better edge transitivity,and high space activity.