Skin　color　is　an　important　feature　for　face　detection　in　color　images.　By　building　and　applying　a　statistical　skin　color　model,　possible　face　regions　in　color　images　can　be　obtained.　However,　due　to　the　including　of　different　race　of　people＇s　skin　color　points,　this　general　statistical　model　is　not　accurate　enough　to　segment　each　specific　image　as　we　expected.　Besides,　most　2D　skin　color　models,　e.g.　CbCr,　which　is　2D　projection　of　3D　color　distribution,　can＇t　adapt　itself　to　lighting　variation.　So　we　present　a　3D　CrCbCg　model　to　describe　skin　color　distribution　more　precisely.　Meanwhile,　considering　skin　points　in　a　specific　image　have　a　relatively　stable　distribution,　we　present　a　fuzzy　cluster　based　skin　model　to　remove　background　points　which　are　wrongly　retained　by　the　general　model.　Experimental　results　show　that　our　algorithm　can　effectively　improve　segmentation　results.