In this paper, we propose a novel face recognition approach based on 2.5D/3D shape matching. While most of existing methods use facial intensity image, we aim to develop a method using three-dimensional information of the human face. This is the main innovation of our technology. In our approach, the 3D dimensional information is introduced in order to overcome classical face recognition problems which are pose, illumination and facial expression variations. The paradigm is to build a 3D face gallery using a laser-based scanner: the off-line stage. In the on-line stage, the recognition, we capture one 2.5D face model at any view point and with any facial expressions. Our processing allows the identification of the presented person by performing the captured model with all faces from the database. Here, the Iterative Closest Point-based matching algorithm provides the pose of the probe whereas the region-based metric provides a spatial deviation between the probe and each face from the gallery. In this metric, we calculate the global recognition score as a weighted sum of region-based distances already labelled as mimic or static regions. For automatic 3D face segmentation, we use an immersion version of watershed segmentation algorithm. This paper also presents some experiments in order to shown illumination, pose and facial expression compensations.