Face recognition using computer-generated database

Abstract

Many face database and recognition systems have been constructed in specially designed studios with various illuminations, poses, and expressions. However, none of these databases yet satisfies a large variation of poses and illuminations that permit the study of systematic 3D human face information, which results in an unsatisfactory success rate. It is caused primarily by the difficulty of data collection of facial images to satisfy the large variation of poses and illuminations to fully represent the 3D characteristics of human faces. We present how computer-generated database can be used for face recognition experiment focused on multiview face recognition/descriptor. We show multiview face data collection using rendering of 3D models. We also illustrate our approach how to build a face descriptor containing 3D information of human face using multiview concepts. This multiview face recognition descriptor is a 3D descriptor using the concept how much powerful a view influences over nearby views, so called as quasiview size