Motion Dense Sampling for Video Classification

Abstract

In this paper, we propose the motion dense sampling (MDS) for video classification, which detects very informative interest points from video frames. MDS has two advantages compared to other existing methods. The first advantage is that MDS detects only interest points which belong to foreground regions of all regions of a video frame. Also it can detect the constant number of points even when the size of foreground region in an image drastically changes. The Second one is that MDS enable to describe scale invariable features by computing sampling scale for each frame based on the size of foreground regions. Thus, our method detects much more informative interest points from videos than other methods. Experimental results show a significant improvement over existing methods on YouTube dataset. Our method achieves 86.8% accuracy for video classification by using only one descriptor.

DOI: 10.1109/ICITCS.2013.6717859

8 Figures and Tables

Cite this paper

@article{Aihara2013MotionDS, title={Motion Dense Sampling for Video Classification}, author={Kazuaki Aihara and Terumasa Aoki}, journal={2013 International Conference on IT Convergence and Security (ICITCS)}, year={2013}, pages={1-4} }