Video Fingerprinting by Using Boosted Features

A fingerprinting system for video copy detection

“…shows the ROC of different methods, with 400 sub-fingerprints and 3 segmentations. From these figures, we can see that, our extraction method has better ROC performances than other extraction methods, which has a result same to that of [5]. The reason why our method can be better others is that our fingerprint method is more effective than these methods [1][2][7] [9], therefore results in good performances for video copyright detection, with same conditions.…”

“…[4], which employ a pairwise boosting on a large set of Viola-Jones features [12], to extract audio fingerprints by learning compact, discriminative, local descriptors on audio spectrograms [4], we has proposed a video fingerprint method, which employ a modified pairwise boosting on visual Viola-Jones features [12], to learn topdiscriminative filters for video fingerprinting [5] [6]. The top four Harr features learned for video fingerprint's extraction are shown in Fig.…”

“…1. We has employed this method for the purpose of video retrieval, which can be seen from [5][6] for more details. In this paper we propose how to use this technique for the task of video copy detection.…”

“…Lee and Yoo proposed a extraction method based on the centroid of gradient orientations [2] [8]. Other video fingerprinting methods can be seen from [5][7] [9][10] [11]. Audio fingerprinting methods were proposed for audio's retrieval more early.…”

“…In our previous works [5][6], we has proposed a video fingerprinting technique for video retrieval and compared with other methods in detail. In this paper, we propose how to adopt the video fingerprinting technique for video copy detection.…”