Eigenspace interpolation for appearance-based object recognition

All eigel/space imerjHJfl1liull IIu!/!/Ull s/Iluulhly illler? polares between 111'0 differem eigellwaces IIsillg high dimCllsional rolatio/!. Howcl'('r, up 10 IIOW its effcctiveness ill objeci recognitioll alld the mlidit)' of Ihe imerpolarioll algorithm hal'l! IlOt been discllssed slIfficiell1/)'. \Ve therefore propo.{e all appearance.based 0/)' jeer recogllition method combillillg the eigenspace interpolation method alld a subspace method. We COlidlleted face recogllition experimems usillg il/lages caplIIredfromlllultiple camera positiolls with mrious iIluminatioll cOllditiollS. ?fperimemal results demollstrate tlte effectiveness ofthe proposed metlrod and the mfidit)' ofthe interpolatioll algorithm.

• Type: Conference
• Year: 2008
• Where: ICPR
• Authors: Tomokazu Takahashi, Lina, Ichiro Ide, Yoshito Mekada, Hiroshi Murase
• Keywords: Computer Vision, Eigenspace Interpolation Method, ICPR 2008, Jeer Recogllition Method, Subspace Method