Title :
H.264/SVC scene motion analysis
Author :
Kas, Christian ; Nicolas, Henri
Author_Institution :
LaBRI, Univ. of Bordeaux 1, Talence
Abstract :
We present a simple and lightweight approach to scene analysis in the H.264/SVC domain. The method is entirely based on the motion vectors found in the compressed stream. Motion segmentation and object detection is performed after the estimation of the camera motion. Important object properties are calculated, which are used for object matching and trajectory estimation. The relative distance to the camera is estimated, resulting in a pseudo 3-D representation of the object trajectories.
Keywords :
data compression; image matching; image representation; image segmentation; motion estimation; object detection; video coding; H.264/SVC scene motion analysis; camera motion estimation; compressed stream; motion segmentation; motion vectors; object detection; object matching; pseudo 3D representation; trajectory estimation; Cameras; Image analysis; Layout; Low pass filters; Motion analysis; Motion estimation; Object detection; Static VAr compensators; Streaming media; Video compression; H.264/SVC; compressed domain; scene analysis; trajectory estimation;
Conference_Titel :
Acoustics, Speech and Signal Processing, 2009. ICASSP 2009. IEEE International Conference on
Conference_Location :
Taipei
Print_ISBN :
978-1-4244-2353-8
Electronic_ISBN :
1520-6149
DOI :
10.1109/ICASSP.2009.4959744