Identifying subjects with variations caused by poses is one of the most challenging problems in face recognition, essentially, a misalignment problem. In this paper, we propose a simple, practical but effective continuous pose normalization method to handle pose variations. First, 2D-3D correspondence is constructed based on five facial landmarks of query image. A single reference 3D mesh is projected onto query image and appearance of query face is assigned to the reference mesh. Frontal view of query face is obtained by rendering the appearance-assigned 3D mesh at frontal pose. Large scale recognition experiments conducted on MultiPIE and FERET databases show that our method achieves competitive, high recognition accuracy, with advantage of database independent and running fast, which is very suitable for practical applications.