Video Classification Using Linear Subspace Methods(International Session 1)
スポンサーリンク
概要
- 論文の詳細を見る
In this paper, we propose a video classifier using linear subspace methods. In the proposed method, the subspace of a training video is formed with color histograms of each frame. When an input video is given, the total sum of projection distances between all frames of the input video and the linear subspace of a training video is computed. This total sum of distances can be regarded as the distance between the input video and the training one. Thus, the input video is classified into the class to which the nearest subspace belongs. By the proposed method, we can classify videos without clustering or shot segmentation. In addition, the learning algorithms of the proposed video classification are presented for reducing memory costs. Furthermore, we propose a shot searching method using subspaces for detecting desired shots or harmful contents. The performance of the proposed method is verified with experiments on short videos downloaded from the Web.
- 社団法人電子情報通信学会の論文
- 2007-10-18
著者
-
Kikuchi Kei
Tokyo University Of Agriculture And Technology
-
Hotta Seiji
Tokyo Univ. Agriculture And Technol. Koganei‐shi Jpn
-
Hotta Seiji
Tokyo University Of Agriculture And Technology
-
Hotta Seiji
Tokyo Univ. Agriculture And Technol.
関連論文
- Color Image Classification Using Block Matching and Learning
- Color Image Classification Using Block Image Replacement and Local Averaging Classifier(Internationa Session 7)
- Video Classification Using Linear Subspace Methods(International Session 1)
- Local Subspace Classifier with Transform-Invariance for Image Classification
- Detection and Retrieval of Nucleated Red Blood Cells Using Linear Subspaces
- Vote-Based Image Classification Using Linear Manifolds
- Generalized Learning Local Averaging Classifier