A fast and automatic video object segmentation technique

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.