Method of Error Compensated Singular Value Decomposition Based Head Pose Estimation for 3D Face Recognition

摘要

The invention proposes a new face pose estimation method for 3D face recognition. First, three-dimensional area (range) when the image is entered automatically apply the facial feature extraction technique based on facial curves. Based on the extracted feature point, we propose a new three-dimensional face pose estimation method used for error compensation singular value decomposition. Using a singular value decomposition to compensate more precisely for the error that exists after acquiring the initial rotation angle. The proposed algorithm is performed using a geometric position of the extracted feature point from the normalized three-dimensional face space. We also propose a method for selecting a candidate faces in the database using a similar three-dimensional candidate (nearest neighbor) classifier for 3D face recognition. Also tested the efficiency and feasibility of the proposed algorithm with respect to various face pose by the experimental results. ; 3-D face recognition, pose estimation, video, extraction, error compensation, normalization, similar candidates