Generating Textures for 3D Head Models: A Wavelet-Based Approach

A novel method for the automatic generation of a facial texture is proposed in this paper. Multiple pictures from the individual are combined to create the texture using a wavelet-based technique. The different tasks involved in the proposed method include: scaling, histogram equalization, deformation and merging. A set of particular feature points of the MPEG-4 standard for multimedia applications is used in the above mentioned tasks. In the merging stage, wavelet multiresolution decomposition of the images and filtering of the wavelet coefficients are performed to obtain a seamless texture. Finally, the resulting texture is mapped onto a 3D head model which has been adapted to the particular individual based on the same pictures. Since the proposed method relies on the MPEG-4 facial and body animation (FBA) standard, it can be easily integrated with applications using the same standard.