Design of efficient shape feature for object-based watermarking technology

Abstract

The accelerated development of IT technologies on the Internet for fast provision of commercial multimedia services has stimulated an urgent demand for reliable and secure copyright protection for digital multimedia. Also, information retrieval is required to search valuable information from wide range of image data for various applications such as biometrics, crime prevention, health informatics, and image search. We proposes the shape representation method using angles, orientations, and locations which is called as Oriented Angular Keypoints (OAK) to make help for shape-based watermarking scheme. First, the contour is extracted from input image and is divided into contour blocks. Then, angles and directions from the divided contour blocks are computed to make unique feature. To evaluate the proposed image retrieval algorithm, commonly employed datasets of Gorelick and MPEG-7 are also used in this paper. The performance of the similarity measure that proposed image retrieval algorithm achieves improvement of about 10 % compared with Shape Context in terms of Bull's eye score.