내용기반 이미지 검색을 위한 효율적인 형태 기술자

초록

An ever-increasing usage of digital images and the wide acceptance of very large volume of image databases naturally give rise to a difficult problem of organizing them for the best and rapid access to their contents. Consequence by, numerous challenges have been attempted to retrieve images using their contents. Content-based image retrievla(CBIR) allows retrieving images that have similar characteristics by extracting image distance using the image features. In this paper, a novel shape descriptor is proposed for the Contents-based Image retrieval applications, and its performance is evaluated quantitatively. The performance of descriptor is also compared with the several descriptors that had been proposed to MPEG-7, such as CSS descriptor, wavelet-based contour descriptor and MLEV descriptor. Experimental results show that ART-C descriptor, proposed in this paper yields the best performance.