ONLINEVIDEOFOREGROUND SEGMENTATION USINGGENERAL GAUSSIAN MIXTURE MODELING MohandSaidAllilit , Nizar Bouguila I andDjemel Ziout tDI,Faculty ofScience, University ofSherbrooke,
Sherbrooke Qc · 2007
Inthis paper, wepropose arobust video foreground modeling byusing afinite mixture modelofgeneral Gaussian distributions (GGD). Themodelhasaflexibility tomodelthevideo background inthepresence ofsudden illumination changes andshadows, allowing foranefficient foreground segmentation. Inafirst partofthepresent work, wepropose aderivation oftheonline estimation oftheparameters ofthemixture ofGGDsandwepropose aBayesian approach fortheselection ofthenumberofclasses. Inasecond part, weshow experiments ofvideo foreground segmentation demonstrating theperformance oftheproposed model. IndexTerms-Mixture ofGeneral Gaussians (MoGG), MML,video foreground segmentation.