Unsupervised segmentation of moving object by region-based MRF model and occlusion detection
Wei Min Zeng, Wen Gao · 2003
A new algorithm is proposed in this paper for unsupervised segmentation of moving object from the video sequence. The spatial partition of frame is performed by the intensity-based watershed algorithm, which keeps the boundary of ultimate extracted object. The motion-based region classification provides a good initialization for object segmentation. An elaborate occlusion region detection scheme removes the potential miss-classification of the uncovered background regions. The final moving object is successfully segmented by the MRF-based labeling technique plus an efficient region growing process. The performance of the proposed algorithm is evaluated on several real sequences and achieves the plausible results.