ABSTRACT
A PCA-based model with relatively small sample size (100 faces) was used for face recognition and has obtained reasonable results .(Volker & Thomas 2003). The generation of synthetic views from 2D input images was not needed. Instead, the recognition was based on the model coefficients which represent intrinsic shape and texture of faces. In case of 3D face reconstruction, a more diverse set of 3D faces would build a more powerful PCA-based model to generate an accurate 3D representation of the 2D face. On the other hand, PCA-based models have not been quantitatively studied in terms of the effect of 3D reconstruction accuracy on face recognition.
