基于视频的人脸识别研究进展_严严.pdf

32 5 Vo l. 32 N o. 5 2009 5 CH INESE JOU RNA L OF CO M PU T ER S M ay 2009 1) , 2) 1) , 2) 严 严 章毓晋 1) 100084) 2) 100084) . , , . 5 ) , , , , . ; ; ; T P 391 DOI : 10. 3724/ SP. J. 10 16 . 2009. 00878 StateoftheArt on VideoBasedFace Recognition 1) , 2) 1) , 2) Y A N Y an ZH A N G Y uJin 1) (N ational L aborator y f or I nf ormation Sc ience and Technology , Tsing hua Univ ersity , Be ij ing 100084) 2) (De ar tment of Electronic E ngineer ing , Tsinghua Univ er sity , B eij ing 100084) Abstract Recent ly, videobased f ace recognit io n has become o ne of t he hot t est t opics in t he do m ain o f face reco gnit ion . H ow t o fully ut ilize bo th spat ial and temporal info rmat ion in video t o overco me t he dif f iculties ex ist ing in t he videobased face recog nit ion, such as lo w reso lut ion of face im ages in video, large variat io ns of f ace scale, radical changes of illum inat ion and pose as w ell as occasionally occlu sion of dif ferent part s of faces , is t he f ocus . T he paper review s most ex ist ing t ypical m et hods f or videobased f ace recog nit ion especially f or the last 5 years) and analy ses t heir respect ive pro s and co ns . T w o com monly u sed video f ace databases and so me ex perimen t al result s are g iven . T he pr ospect s f or fut ur e develo