Image Matching Algorithm Based On The Time-varying Images

Image matching has always been a fundamental research in the fields of computer vision and pattern recognition.And the matching of time-varying images is an very important application scene in image matching,such as,target tracking in computer vision,change detection of remote sensing images,clinical diagnosis and treatments in medical field.Presently,progress has been made in matching of time-varying images containing only linear transformations.However,in practical applications,there are often many complex scenes with nonlinear transformation between two time-varying images.In order to facilitate subsequent processing and analyzing,fast and accurate registration of the nonlinear deformations is often necessary in many practical matching problems.Therefore,it is of great importance to make research on the matching algorithm of time-varying images which can achieve nonlinear deformation quickly and accurately.Based on systematical investigation of the research of image matching technology and time-varying image matching algorithm at home and abroad,and focused on the limitations of existing algorithms for matching precision and speed of complex nonlinear time-varying images,the main research content of this article is as follows:(1).Aiming at the complex non-linear change scenes in time-varying images,DAISY feature point extraction operators are used to extract dense features from the image which overcomes the limitation of the current mainstream feature point extraction algorithm SIFT due to the use of fixed-dimensional vector descriptors in nonlinear matching.And also due to the unique circular symmetric template structure of the DAISY itself,the repeated calculation of a large number of gradient images is avoided,which improves the speed of the algorithm.The compare experimental results of DAISY and SIFT operator extraction speeds indicate that the DAISY operator is more than twice as efficient as SIFT in extracting dense features.(2)This paper optimizes the Patch-Match algorithm and proposes an improved algorithm which constrains the initial local window in the random search process.The problem of mis-matching caused by noise and gray consistency and weak texture in the image is solved,and the convergence speed of the algorithm is accelerated and the efficiency is improved.(3)The traditional matching algorithm performs poorly in generality due to its proposing for a certain application in a certain field.This paper applys the time-varying image matching algorithm based on computer vision image to remote sensing time-varying images and medical time-varying images,and has achieved a good research results on cross-domain applications.The results of three groups of time-varying image matching experiments in different fields show that the proposed algorithm has good adaptability and robustness to time-varying images in different fields.