Improved Face Model Fitting Using Tensor-Based AAM
Daijin Kim, Hyung-Soo Lee · 2008
In this paper, we propose a tensor-based active appearance model (AAM) which improves the fitting performance of conventional AAM. Tensor-based AAM generates the specific AAM basis vectors by indexing the model tensor in terms of the estimated input image variations. Experimental results show that the proposed tensor-based AAM reduces the average fitting error than the conventional AAM significantly.