3D reconstruction of human faces from range data through HRBF networks

3D reconstruction of human body parts, and faces in particular, is catalysing growing interest in many disciplines ranging from basic image processing to video conferencing, constructive and plastic surgery, rehabilitation and virtual clones. A host of devices (3D scanners), which provide these 3D models, have come to the market in the last few years. They are based on sampling a large number of 3D data points over the surface and of fitting a suitable analytical model to them. There are two main problems which have to be faced: filtering of the noise associated to sampling and interpolation between the samples. These two problems can be reframed in the domain of regularisation. It is shown how a regularised model can be efficiently obtained by using a new neural network called hierarchical radial basis function network (HRBF)