Improved Multi-pose 2D Face Recognition Using 3D Face Model with Camera Pose Estimation Approach and nD-PCA Recognition Algorithm

In this paper, we will present a novel method to recognize multi-pose face image. In the algorithm we firstly estimate pose of camera by using an estimation algorithm which just used a 2D face image. Then the estimated data can be applied to view position of 3D face model in virtual environment to synthesize 2D verification exemplar DB. Thus face pose in these synthesized 2D verification images can be the same with the input ones. Then an nD-PCA based algorithm will be employed to extract eigenvector of exemplar images and input ones. Finally, the classification machine which is based on cosine distance method will be employed to classify the verifier. We also have carried out a simulation experiment in windows XP system to evaluate efficiency of the proposed algorithm. As we can see from the experiment result that the correction rate can achieve to 92% for front view and 40% for nearly profile-view, robustness of our proposed algorithm is much better than some kinds of conventional approach in multi-pose face recognition field.