Face is being considered as one of the most commonly used biometric modality. The inaccuracy in two dimensional face recognition systems is mainly due to pose variations, occlusions, illumination etc. Among this, changes in illumination condition do not affect 3D face recognition systems. But pose variation drastically changes the appearance of face images. To solve the problems with depth map and texture images corrupted by head pose variations and the occlusions generated due to these pose variations, a reconstruction method is proposed which consist of three stages. In the first stage, the pose correction is done by Iterative Closest Point (ICP) algorithm and in the second stage the occluded region of the face is reconstructed by a resurfacing method called patch cloning. It is followed by the wrapping of reconstructed depth map by its texture to generate a 3D model. The statistical error between the original face and the reconstructed face is also evaluated. In this work, facial symmetry is used as a prior knowledge. Experiments are done with the FRAV3D database.