A Novel Mathematical Based Method for Generating Virtual Samples from a Frontal 2D Face Image for Single Training Sample Face Recognition

This paper deals with one sample face recognition which is a new challenging problem inpattern recognition. In the proposed method, the frontal 2D face image of each person isdivided to some sub-regions. After computing the 3D shape of each sub-region, a fusionscheme is applied on them to create the total 3D shape of whole face image. Then, 2Dface image is draped over the corresponding 3D shape to construct 3D face image. Finallyby rotating the 3D face image, virtual samples with different views are generated.Experimental results on ORL dataset using nearest neighbor as classifier reveal animprovement about 5% in recognition rate for one sample per person by enlarging trainingset using generated virtual samples. Compared with other related works, the proposedmethod has the following advantages: 1) only one single frontal face is required for facerecognition and the outputs are virtual images with variant views for each individual 2) itrequires only 3 key points of face (eyes and nose) 3) 3D shape estimation for generatingvirtual samples is fully automatic and faster than other 3D reconstruction approaches 4) itis fully mathematical with no training phase and the estimated 3D model is unique foreach individual.