코너 패치 속성의 재배열에 의한 콘텐츠 기반 영상검색

Abstract

This paper presents a Content based image retrieval which can retrieve image automatically from image database. Content based image retrieval systems mainly has been used the color, shape and texture features such as primitive. In this study, the shape feature of these features are used in searching image. In this paper, it is implemented the image retrieval method based on interest point of the object by using the histogram of feature vectors to be rearranged. Following steps are implemented. Firstly, after detecting edge of object in image, straight lines are extracted by using hough transform from it. The intersection of extracted straight lines is interest point and it constitutes a corner patch. Secondly, the patch is arranged in circular form based on the interest point and construct table. Then the table is normalized on the basis of an appropriate threshold. In addition, frequency of the normalized feature vector is calculated by using the histogram. And then the similarity between a feature vector based on the calculated frequency and image database is measured. Simulation is performed using MATLAB, and tested against existing algorithm. Performance and recall analysis show that the proposed techniques have superior performance and retrieval accuracy than the conventional ones.