Research On Image Registration Based On Normalized Cross-correlation Matching

Along with the vigorous development of digital image processing,image registration technology as the basic step of image processing is increasingly applied to various emerging fields,such as target detection,three-dimensional reconstruction,virtual reality and so on.Therefore,it is necessary to constantly improve the performance of the registration technology.The purpose of this paper is to use the improved normalized cross-correlation matching algorithm to improve the registration accuracy of image registration techniques and to expand the scope of application of image registration techniques.The main research content includes the following aspects:Firstly,this paper briefly introduces the research background and significance of image registration and the development status at home and abroad,analyzes the mathematical principle,basic elements and technology classification of image registration,and systematically elaborates the framework of image registration technology.At the same time,the principle of Harris corner detection and random sampling consistency algorithm is introduced in detail,which will pave the way for subsequent analysis and research.Secondly,the principle of normalized cross-correlation matching algorithm(NCC)is mainly introduced,and the experiments show that the algorithm has higher sensitivity to rotation and scale transformation.In view of the insufficiency of the algorithm,this paper proposes the combination of the cat swarm algorithm(CSO)and the NCC algorithm to determine the size and direction of the search window used to calculate the NCC.Using the new optimal ratio(s)and rotation angle(?)produced by each optimization of the cat algorithm,the search window of the NCC algorithm is converted and resampling,and the value of the fitness function is maximized.Finally,the optimum size and direction of the search window are found.Through the experimental comparison with the traditional normalized cross-correlation matching algorithm,it can be found that the sensitivity of the improved normalized cross-correlation matching algorithm has been significantly reduced in the rotation transformation and the scale transformation,and the matching accuracy has also been improved.Finally,combined with the improved normalized cross-correlation matching algorithm,an image registration technique based on CSO-NCC is proposed.This technique can effectively register images with obvious rotation and scale transformation.Then compare with the other two kinds of registration algorithms,the results show that the algorithm has high accuracy of registration and stability,and it has certain practical value.Then,it is applied to panoramic image processing to achieve effective position correction of images acquired from the same scene with different angles of view,and an ideal panoramic effect image is finally obtained.