Bilkent University
Department of Computer Engineering


Reconstruction, Matching, and Registration of 3D Shapes


Yusuf Sahillioğlu
University of Pennsylvania (UPenn), Computer Science, Post-doc

This talk covers the key aspects and contributions of my MS and PhD theses as well as my post-doc studies. We first present a generic surface deformation framework for the problem of 3D shape reconstruction. A spatially smooth and topologically plausible surface mesh representation is constructed via a surface evolution based technique, starting from an initial model that represents the bounding surface.

We then focus on matching the featured or all present points of two semantically similar 3D shapes whose surfaces overlap completely or partially up to isometric, i.e., distance-preserving, deformations and scaling. We cast this problem, commonly goes by the name shape correspondence, as an optimization of a distortion function and adapt well-established paradigms such as bipartite graph matching, greedy, EM, combinatorial, and multi-resolution algorithms, and dynamic programming for the required minimization.

Finally we mention our joint-optimization algorithm for a brand new type of 3D shape registration problem that we introduce, namely 3D volumetric shape registration where two shapes to be registered are represented explicitly as volumetric objects. For registration, we deform the high-resolution template tetrahedral mesh towards the underlying geometry of a patient-specific voxel set from a CT scanner.


DATE: 18 June, 2013, Tuesday @ 13:40