Face parts importance in face and expression recognition

Identification of the crucial face parts for feature extraction plays an important role in face recognition with respect to recognition accuracy and data complexity. The significance of face image blocks was analyzed in LBP (Local Binary Pattern) feature space. To optimize the face parts selection we applied simulated annealing. Proposed approach was tested on standard FERET database. The same optimization approach was applied to the facial expression recognition and tested on JAFFE database. In some cases asymmetrical selection of important image blocks was found to be better than symmetrical selection. The proposed methodology is not limited to LBP but it is applicable to any type of feature space.