Resampling based method for pixel-wise correspondence between 3D faces

This paper presents a novel method for pixel-wise correspondence between 3D face images. The method is based on uniform gridded-resampling of different 3D faces and advances a particular even-partition procedure including uniform face segmentation and iso-curves initialization for correspondence. Pixel-wise correspondence between 3D face images is the basis of a 3D face model, which is capable of analyzing novel face images. Pixel-wise correspondence (which can also be called point-to-point correspondence) between 3D prototypes is currently based mainly on several algorithms derived from optical flow. Using a series of examples, our approach has shown its advantages and been proved to be promising for pixel-wise correspondence between 3D face images.