Motion Object Segmentation Using Regions Classification and Energy Model
An automatic video object segmentation is proposed. The video scene is partitioned into some homogeneous regions by an automatically cluster to form regions method. Then these regions are initially classified into three categories: moving object, candidate and background using the difference information between two successive frames. The spatio-temporal energy model is constructed to determine the candidate regions to moving object or background. Some post-processing methods are utilized to achieve the more accurate segmentation object. Experimental results show that the spatial accuracy of our proposed algorithm improves about 30–50% and temporal coherency improves about 0.05–0.70 than COST211 AM.
KeywordsVideo processing Motion segmentation Regions selection
- 5.Patras I, Hendricks EA, Lagendijk RL (2001) Video segmentation by MAP Labeling of watershed segments. IEEE Trans Pattern Anal Mach Intell 23(3):326–331Google Scholar
- 9.Wollborn M, Mech R (1998) Refined procedure for objective evaluation of video object segmentation algorithms [R]. Doc. ISO/IEC JTC1/SC29/WG11 M3448, March 1998Google Scholar
- 10.COST211 AM. Working site for sequences and algorithms exchange. http://www.tele.ucl.ac.be/exchange