Smooth surface interpolation and texture adaptation for MPEG-4 compliant calibration of 3D head models

Abstract The technique proposed in this paper provides operative indications for implementing 3D head models compliant to face parameters recently standardized in Motion Picture Expert Group, MPEG-4, and capable of being calibrated on any specific human face. Since the set of calibration guide points, defined in MPEG-4 as feature points, is very limited with reference to the complexity of the human head geometry, suitable interpolation must be performed to assure smooth and realistic surface rendering. On the other hand, care must be paid in texture adaptation to avoid annoying artifacts in correspondence to deformable face features, which are even more appreciable when the model is animated. The proposed methodology is based on the use of Radical Basis Functions (RBF) for smooth surface interpolation, applied according to a multi-step procedure for progressive detail definition. The set of interpolating functions has been derived empirically exploiting a priori knowledge on the geometry of human heads. The adaptation of the texture information, achieved using the MPEG-4 feature points as anchor points, is also performed according to a multi-step fashion by organizing the head model polygons into homogeneous subsets (i.e. polygons of the hair region define a specific subset). Experimental results, as reported in the following, prove the effectiveness of the proposed method showing how it is possible to adapt the same generic model to different face parameters. A variety of applications is envisaged in multimedia products, virtual and augmented reality, very low bitrate videophone and efficient three-dimensional (3D) facial data archiving.