A new frontal face image synthesis method was proposed.The performance of most current face recognition algorithm degrades significantly with face pose changing.In this paper,an extended linear object class theorem was proposed.Firstly,the virtual frontal shape and texture of a non-frontal face image were synthesized using the extended linear object-class theorem.Then the synthesized texture was combined with texture of the original test image.Finally,the combined texture was stretched onto the virtual frontal shape to generate the synthesized frontal face image.Experiments show that the proposed method can effectively "rotate" high resolution face images from a single 2D View.After synthesis,the recognition rate increased from 27.3% to 81.9%.