Research On Image Matching Algorithm Based On KPCA

Image matching is a process of seeking the corresponding sub image in a strange one based on the image with known mode. It is the basics of image understanding and machine vision. Its applicable field is very extensive. The researches refer to many interrelated ken such as image sampling, transformation, pretreatment, segmentation, feature extraction and so on. It also linked tightly to computer vision, multidimensional signal processing and numerical computation method.This paper firstly analyzes the research actuality of image matching in recent years at home and abroad systematically and summarizes the development of this field of research. Then we discuss its research methods and draw a comparison with each other. Furthermore, we propose some problems for discussion.Though the matching approaches based on gray scale information and characters of image had disadvantages, the combinations of two methods have achieved better matching effect and it has made an appeal to many researchers. This paper emphasizes the matching approaches based on gray scale information and characters of image.Secondly, principal component analysis is the commonest approach of feature extraction and is applied to various fields. This paper specifies its basic principle, mathematical model, geometric significance, deducing and the process of algorithm, then introduces principal component analysis to image matching process and proves its optimality in the experiments. The image matching based on principal component analysis has preferable robustness then traditional normalization cross-correlation matching and can avoid the influence of image size, direction, partial scene and noise jamming.Thirdly, principal component analysis only refers to second-order statistic information of image data and does not utilize the high-order ones. It ignores its nonlinear relativity among a large number of pixels. Research indicates that high-order statistic information sometimes contains image edge or nonlinear relations among multi-pixels. Aiming at this problem, this paper proposes kernel method and studies all its primary components according to the examples of least square regression. Then, we combine kernel method and principal component analysis to present the image matching method based on kernel principal component analysis.Finally, we draw a comparison between principal component analysis and kernel principalcomponent analysis aiming at one of the hot researches on image matching——face matchingand validate its effectiveness.