Presentation 2011-02-17
Video Similarity Measuring using SIFT Trajectories without Pre-learning Process Learning-Free video Retrieval
Hiroaki KUBO, Hideo SAITO, Shin'ichi SATOH,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Video similarity is a basic technique of scene/object recognition. To extract discriminative features from videos, recent popular methods handle videos with visual words i.e., vector quantized descriptors. However, the performance of vector quantization depends on the pre-learning, especially on the clustering. In this paper, we propose a novel video similarity measurement by using marginal distribution on the feature space. Our method uses key-point trajectories extracted from videos and represents videos as histogram representing marginal distribution. Video similarity is defined as combination of similarities between marginal distributions. We tested our method by similar video retrieval tasks. Our method retrieved a kind of hommaged videos which have modified background, are heavyly encoded, rewited by hand and so on.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Video Similarity / Non-learning / Vector Quantization / Keypoint-Trajectory / SIFT
Paper # PRMU2010-215
Date of Issue

Conference Information
Committee PRMU
Conference Date 2011/2/10(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Pattern Recognition and Media Understanding (PRMU)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Video Similarity Measuring using SIFT Trajectories without Pre-learning Process Learning-Free video Retrieval
Sub Title (in English)
Keyword(1) Video Similarity
Keyword(2) Non-learning
Keyword(3) Vector Quantization
Keyword(4) Keypoint-Trajectory
Keyword(5) SIFT
1st Author's Name Hiroaki KUBO
1st Author's Affiliation Science and Technology of Keio University()
2nd Author's Name Hideo SAITO
2nd Author's Affiliation Science and Technology of Keio University
3rd Author's Name Shin'ichi SATOH
3rd Author's Affiliation National Institute of Informatics
Date 2011-02-17
Paper # PRMU2010-215
Volume (vol) vol.110
Number (no) 414
Page pp.pp.-
#Pages 6
Date of Issue