Moving object detection is a useful tool for the manipulation of video contents for such purposes as intelligent video browsing/analysis and video surveillance systems. There are a variety of methods to detect moving objects not only on pixel domain data but also on coded data domain. In this article, we propose an effective moving object detection method from MPEG-coded data. This method utilizes motion vectors and DCT coefficient data in the MPEG streams. Since motion vectors and DCT coefficients can be obtained without full decoding of original MPEG stream, fast detection can be expected. However, motion vector information does not always represent the real object motion. In addition, there are cases where no motion information is provided from coded data, such as the case of intracoded picture. It is shown in the experiments that the proposed method can detect moving objects successfully in such cases. © 1999 Scripta Technica, Syst Comp Jpn, 30(13): 1–12, 1999
[1]
Akio Yoneyama,et al.
Detection of Moving Objects from MPEG Video Stream
,
1996
.
[2]
Jean-Marc Odobez,et al.
MRF-based motion segmentation exploiting a 2D motion model robust estimation
,
1995,
Proceedings., International Conference on Image Processing.
[3]
Shih-Fu Chang,et al.
CVEPS - a compressed video editing and parsing system
,
1997,
MULTIMEDIA '96.
[4]
Mandyam V. Srinivasan,et al.
Measurement of optical flow by a generalized gradient scheme
,
1991
.
[5]
Anup Basu,et al.
Motion Tracking with an Active Camera
,
1994,
IEEE Trans. Pattern Anal. Mach. Intell..
[6]
Shih-Fu Chang,et al.
Tools for compressed-domain video indexing and editing
,
1996,
Electronic Imaging.