내용기반 영상검색 기법에 관한 연구 = Research of content-based image retrieval method

다국어 초록 (Multilingual Abstract)

Most of the current commercial database system supports retrieval of image data. But these system support the retrieval of image data by comment-based method using text information. This paper proposeda method calculates 2-dimensional histogram from a binary image which is obtained from an original gray level image and uses it as shape information of an image. A new similarity measure which uses the 2-dimensional histogram values is proposed in this paper. And the proposed method uses HSI color model. In order to obtain color information, the method subdivides an image into 16 subimages and evaluates similarities based on H and I histograms on each subimages. By evaluating a similarity on each subimage basis, we can obtain some location information which is not available in ordinary color histogram and improve the accuracy of image retrieval. In order to verify the proposed schemes, an image retrieval system is implemented using JAVA and some retrieval simulations are performed. The simulation results show that the proposed scheme retrieves candidate images accurately.