一种视频对象分割技术的研究与实现.pdf

19 4 Vol19 No4 JOURNAL OF ELECTRONIC 2005 8 MEAS UREMENT AND INSTR UMENT 43 王东升 李在铭 ( , 610054) : , Neyman- Pearson , , , : ; ; Neyman- Pearson ; ; Research and Implementation of A Video Object Segmentation Technology Wang Dongsheng Li Zaiming ( S hool of Communi ation and Information Engineering, UESTC, Chengdu 610054, China) Abstract: In this paper a te hnology of video obje t segmentation is presented. Firstly, the urrent frame is ompen sated after globe moving estimation, and the differen e operation is made between the urrent frame that has been om pensated and the prior frame; then the moving region is obtained from the differen e image based on Neyman- Pearson riteria, and in the differen e image a new region growing algorithm whi h over ome the default of the traditional algo rithm of region growing in the time- domain is used to obtain the binary segmentation model of video obje t; finally, based on the moving dete tion algorithm the labeled blo ks that are resulted from noise are eliminated. The experiment result shows that the te hnology an segment the video obje t effi iently. Keywords: video obje ts;globe moving estimation; Neyman- Pearson riteria; region growing;MRF model [ 1- 3] 1 MPEG- 4 , , Neyman- Pearson , ( VOP) , ,