A new spatial-temporal video object segmentation algorithm based on region compensation in fixed time interval
Shiping Zhu, Yunyu Lin, Qingrong Zhang · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2008
A post-processing algorithm of video object segmentation that is based on region compensation in fixed time interval is proposed on the base of the traditional spatial-temporal video object segmentation algorithm. Firstly, an image variation detection template is calculated by noise suppression and dilatation linking of frame difference images; Secondly, the morphological gradient is calculated by the simplification of morphological opening and closing reconstruction of the original image. The accurate boundary of the object can be obtained by watershed algorithm through the nonlinear transformation and gradient grade classification of the morphological gradient image. The initial binary template of the video object can be extracted through proportion calculation; Thirdly, the final complete video object template can be obtained through region compensation in fixed time interval. Experimental results indicate the correctness and validity of the proposed video object segmentation algorithm.