DocumentCode
3193779
Title
A fast and automatic video object segmentation technique
Author
Lihua, Guo
Author_Institution
Sch. of Electron. & Inf., South China Univ. of Technol., Guangzhou
fYear
2008
fDate
25-27 May 2008
Firstpage
714
Lastpage
717
Abstract
The video object segmentation is a key component of digital video representation, transmission and manipulation, example application including content-based video retrieval, object-based video coding and so on. A fast and automatic video segmentation technique, which aims at foreground and background segmentation via effective combination of color and motion analysis module, is proposed in this paper. Firstly, the watershed segmentation algorithm is employed to provide initial regions according to pixels luminance gradient. Secondly, regions are merged according to their color and motion similarity. Finally, the semantic video object will be obtained after post-processes. The main advantage of this method is its fast and automatic implementation of video object segmentation.
Keywords
content-based retrieval; image colour analysis; image motion analysis; image representation; image resolution; image segmentation; video coding; video retrieval; automatic video object segmentation technique; background segmentation; color analysis module; content-based video retrieval; digital video manipulation; digital video representation; digital video transmission; foreground segmentation; motion analysis module; object-based video coding; pixels luminance gradient; semantic video object; watershed segmentation algorithm; Colored noise; Content based retrieval; Filters; Gaussian noise; Image color analysis; Image segmentation; Information retrieval; Motion analysis; Object segmentation; Video compression;
fLanguage
English
Publisher
ieee
Conference_Titel
Communications, Circuits and Systems, 2008. ICCCAS 2008. International Conference on
Conference_Location
Fujian
Print_ISBN
978-1-4244-2063-6
Electronic_ISBN
978-1-4244-2064-3
Type
conf
DOI
10.1109/ICCCAS.2008.4657872
Filename
4657872
Link To Document