Title :
Temporal gradient pattern for the near-duplicate video clustering
Author :
Lee, Hyundeok ; Byun, Hyeran
Author_Institution :
Dept. of Comput. Sci., Yonsei Univ., Seoul, South Korea
Abstract :
This paper proposes a near-duplicate video clustering method based on temporal gradient pattern (TGP). Most related work that extract feature from video ignored temporal changes of video. To solve this problem we propose feature extraction method that used temporal changes of video that is the most important feature at video. Temporal gradient pattern is extracted from a shot video and patterned by temporal changes of frames within a shot video. We proposed the method that can cluster near-duplicate video effectively and quickly using the temporal gradient pattern. Experiments have performed for speed, feature extraction and video clustering evaluation. Proposed method shows good performance to represent videos using proposed feature extraction method, clustering method can obtain good performance and speed also is good.
Keywords :
feature extraction; gradient methods; image matching; video signal processing; feature extraction; near-duplicate video clustering; temporal gradient pattern; Color; Context; Databases; Feature extraction; Image color analysis; Machine learning; Streaming media; Near-duplicate video; Temporal gradient pattern; Video clustering; Video matching;
Conference_Titel :
Machine Learning and Cybernetics (ICMLC), 2010 International Conference on
Conference_Location :
Qingdao
Print_ISBN :
978-1-4244-6526-2
DOI :
10.1109/ICMLC.2010.5580710