Vision Journal Club, Nov 02, 2012, 02:00PM - 03:00PM, Wachman 447
High Resolution Registration, Tracking, and Synthesis of 3D Deforming Objects
Dr. Yang Wang, Siemens Corporate Research
Recent advent of new technologies allows us to capture massive amounts
of high resolution, high frame rate geometry and appearance data in
3D. In order to use such data for the temporal analysis and realistic
synthesis of dynamic motion, an efficient object registration and
tracking algorithm is needed to establish dense correspondences. This
problem remains challenging for non-rigid objects (e.g., faces and
hearts) especially when subtle motion estimation is required. In this
talk, I will present manifold-based approaches for deformable surface
registration and object tracking. An intrinsic shape representation is
introduced in the first part for 3D non-rigid surface matching.
Furthermore, a low-dimensional representation is learned to encode the
non-rigid motion prior based on manifold learning. Due to the strong
implicit and explicit smoothness constraints imposed by the algorithm
and the high-resolution data, we can establish dense inter-frame
correspondences for analyzing and synthesizing subtle non-rigid
motion. Several graphics and medical applications are included to
demonstrate the proposed approaches.
Dr. Yang Wang is a research scientist in Siemens Corporate Research, located at Princeton, NJ. Prior to SCR, he worked as a post-doctoral research fellow in Robotics Institute at CMU from 2006 to 2008. He received his Ph.D. degree from Stony Brook University in 2006. Dr. Wang specializes in non-rigid shape matching and motion tracking, medical image analysis, face recognition, and facial expression analysis. He has more than 35 published papers and 5 patent applications in the above areas. Dr. Wang is a member of ACM and IEEE.