Abstract
This thesis develops a method to solve the unpredictable head orientation problem in 2D facial analysis. In this thesis, we extend the expression subspace to the view-based AAM so that it can be applied for multi-view face fitting and pose correction for an input face of any expression. The experimental results will demonstrate that the proposed algorithm can be applied to improve the following facial identification process. We testing our system of the sequence by using Intel C2D 6300 CPU and the frame size is 320*240 pixels. It requires 30~45 ms to fitting a face and 0.35~0.45 ms for warping.