Skip to Main Content
Video copy detection has been actively studied in a wide range of multimedia applications. This paper presents a novel content-based video copy detection method using the randomly projected binary features. A very efficient sparse random projection method is employed to encode the image features while retaining their discrimination capability. By taking advantage of the extremely fast similarity computation of binary features using Hamming distance, we present a keyframe-based copy retrieval method that exhaustively searches the copy candidates from the large video database without indexing. Moreover, an effective scoring and localization algorithm is proposed to further refine the retrieved copies and accurately locate the video segments. The experimental evaluation has been performed to show the efficacy of the proposed randomly projected binary features. The promising results in the TRECVID2011  content-based copy detection task demonstrated the effectiveness of our proposed approach.