Querying Image Database by Video Content

摘要

A content-based image retrieval (CBIR) system using the features extracted from a video clip is proposed in this paper. The proposed CBIR system is directly applied to access a fish database collected and managed by the Institute of Zoology, Academia Sinica, Taiwan. The major contribution of this work is three-fold: (1) the proposed query-by-video-content scheme can overcome the problems caused by non-rigid motions, serious deformations, and partial occlusions of target objects; (2) the proposed scheme can tolerate different imaging environments; (3) the proposed representation scheme is invariant to translation, scaling, and rotation. Experimental results have confirmed the effectiveness of the proposed CBIR system.