Research On Facial Features Location Algorithm

The facial features location is the key issue of research on face. It can provide basic data for post-processing on face, and can be widely used in face recognition, facial expression recognition, face beautification and other fields. In recent years, although the facial features location has been extensively studied, existing methods is still influenced by the illumination, pose, expression and occlusion. Its performance needs to be further improved. Therefore, this thesis undertakes an in-depth study on facial features localization methods.The main work is as follows:(1) This thesis firstly proposes a semi-automatic calibrating features method which greatly improves the efficiency and accuracy of calibrating.(2) In order to determine the initial position of active shape model, it needs to use local features location method to locate the pupils. A pupil location method based on integral projection and radial symmetry transform is proposed. It is found that the method has high location accuracy and robustness, and can provide accurate initial position for active shape model.(3) In view of that the local binary patterns operator can describe the local texture well and has good robustness for light, a modified active shape model is proposed in this thesis by utilizing local binary patterns operator. Finally, simulation experiment verifies the feasibility of the modified algorithm and the result is compared with the original active shape model.According theoretical analysis and experiment results, the modified active shape model improves the location accuracy and robustness for light.