Range Flow from Stereo-temporal Matching: Application to Skinning

Recently the 3D-MATIC Research Laboratory has developed techniques for the generation of 25 3D models per second of captured data. Our aim is to use these series of 3D models to study deformations of the human body. However since the 3D models have different topologies, they cannot be used directly for analyses of non-rigid motions. Therefore the generation of range flows is a prerequisite to further studies. Since we have acquired a lot of experience in correlating pixels in stereo pairs of images, we have naturally investigated the use of similar techniques to track pixels associated to range data between successive images. We present a method that allows the generation of range flows using stereo and temporal matching. We then demonstrate the efficiency of our techniques by applying the generated range flows to a study of the deformation of human skin around joints in order to perform the skinning of the region of interest.