Research And Application Of Digital Image Template Extraction And Matching Approach

Digital images, as an integral part of the digital media, are becoming more and moreimportant in the field of research and application. The research and application of digitalimage processing technology are combined more tightly, and promoted each other. Especiallyin the understanding and analysis of the digital image technology, the demand of the nationaldefense security, production, living, medical applications, and many other areas is very urgent.Meanwhile, in machine vision, artificial intelligence, pattern recognition, multimediatechnology, and many other research areas, digital image processing technology lie in acritical position.In the technology of image analysis and understanding, the image target representationand recognition technology is an important part of image analysis and understanding, how toselect the image feature and then construct appropriate description determines whether thesuccess of image matching identification. In image matching recognition, template matchingmethod, based on its simple principle, has been researched for years, and achieved fruitfulresults. The methods based on the gray level and based on the features of the templateselection and construction effect are different. In the process of image matching recognition,the methods based on the features are steady when changing the light, displacement, rotation,scale. Therefore, the methods based on the features get in-depth research and extensiveapplication. The critical thing is to find a way to construct the optimal matching template.Based on the existing research, on the basis of feature extraction and image matchingmethods, we start by structuring template of image matching. In order to improve accuracy inimage matching applications, we studied from two aspects: the image template matchingmethods and extraction methods. In this paper, the research work mainly includes:1. In the research of image matching method, first of all, template matching methodsbased on gray level are described and did the matching experiment, and we found in theexperimental process of the NCC algorithm matches the image binarization threshold changeis bigger, then the NCC algorithm is weighted, therefore, the WNCC algorithm is proposed,and experimental results show that the matching results of the WNCC algorithm ofbinarization threshold sensitivity is smaller. At the same time, the contrast experiment basedon the features of the template matching methods, in the light of the angle displacement, rotation and scaling change situation, they got a better result. Thus it verified the importanceof structuring template.2. In the study of the template extraction method, in order to get better recognition effect,combining with the characteristics of template matching methods in the application of imagematching, from the research of the matching template, we found that constructing optimalmatching template can enhance the recognition effect. Therefore, the extraction method basedon the features of the template is proposed in this paper. Edge feature has certainrepresentativeness in the description of the image target, first of all, using the Canny operatorfor edge detection. To solve the problem of how to describe the extracted edge, this papercompared the chain code method with the method of Fourier descriptor, and then we selectthe Fourier descriptor method. To solve the impact of location translation, arbitrary anglerotation, scale and other complicated situations, we proposed an approach which combiningthe elliptical Fourier descriptor and the method of the invariant moments. And theexperiments demonstrate that the proposed template extraction method achieved a bettermatching result in the process of matching applications than the traditional ones.3. We summarized the image Matching methods and the template extraction methodwhich proposed in this paper. And we selected the MATLAB as our simulation environment,then programmed a comparative experimental environment which integrated several typicalmatching methods, and another experiment GUI named TE-Matching which based on theproposed template extraction method of image matching. So it will become more convenientfor our follow-up research and application experiments.In the end, we summarized all the work of this paper, and illustrated the innovationpoints respectively. Moreover, we designed a prospective about the development trend of theresearch content and the emphasis in the future, combining with the present technology ofimage matching recognition.