Lighting Analysis and Texture Modification of 3D Human Face Scans

Abstract

The 3D scanning of human faces is an important tool to acquire accurate shape and texture information of heads and faces for both animation and recognition purposes. One of the most popular scanning tools is Cyberware head and face 3D colour scanner. Currently, 3D face databases scanned using the Cyberware head and face 3D colour scanner databases are public available and have been used in many face recognition system as the training data, one of which is USF human ID 3-D database. The mechanism of facial texture capturing, however, remains unclear due to the special settings of the light sources during scanning. The direct application of the acquired texture intensity as the albedo of the face surfaces leads to suboptimal modelling results. This paper analyses the lighting configuration of the scanner and proposes an appropriate reflectance model to describe the reflecting mechanism when scanning. The texture modifications can be carried on in the existing scanning results without any further scans. Experiment results show that the proposed reflectance model can better approximate the reflecting situation of scanning so that the acquired textures can be directly used as diffuse reflectance coefficients and achieve better rendering results.