Three-dimensional Intrinsic Shape (TP-P4)

Author(s) : Victor Ha (Samsung Information Systems America, USA) Jose Moura (Carnegie Mellon University, USA)

Abstract : In this paper, we define the \emph{intrinsic shape} of an object in three-dimensional space. The intrinsic shape is invariant to affine-permutation distortions that include translation, rotation, scaling, and skewing. We present an algorithm that blindly recovers from arbitrarily affine-permutation distorted shapes its intrinsic shape. This algorithm referred to as 3D BLAISER (BLind Algorithm for Intrinsic ShapE Recovery) executes as one of its steps the 3D PRA (Point-based Reorientation Algorithm) that solves the reorientation problem of 3D shapes, an important issue in its own right in the image processing and machine vision communities. The 3D PRA extends to 3D the 2D PRA, but is much more complicated due to the multiple axes of rotation and associated shape fold numbers. We describe the 3D BLAISER and the detailed discussion on the 3D PRA.

Menu