This　paper　aims　to　solve　the　problem　of　compressing　3D　mesh　sequence　with　different　connectivity.　For　a　non-isomorphic　3D　mesh　sequence,　there　are　both　spatial　correlations　in　each　single　mesh　and　temporal　correlations　between　adjacent　frames,　which　makes　it　insufficient　for　applying　static　3D　mesh　compression　algorithms　to　each　single　mesh.　In　this　paper,　a　novel　method　for　compressing　3D　mesh　sequence　by　using　a　technique　of　3D　human　faces　non-rigid　registration　was　presented.　By　using　the　proposed　registration　algorithm　on　a　3D　mesh　sequence,　the　temporal　redundancy　of　the　mesh　data　was　removed,　thereby　it　reduced　the　storage　cost　drastically.　In　addition,　the　registration　procedure　with　the　inter-coding　idea　in　video　coding　technique　was　combined.　By　adopting　structure　of　bidirectional　multiple　reference　frames,　the　efficiency　of　registration　compared　with　previous　work　was　improved.　The　experimental　results　show　that　better　subjective　and　objective　quality　by　applying　the　proposed　compression　method　to　3D　face　mesh　sequence　can　be　obtained.　©,　2015,　Beijing　University　of　Technology.　All　right　reserved.