Title: "Algebraically accurate 2D symmetry detection"

Speakers: Yosi Keller, Yale University, and Yoel Shkolnisky, Tel Aviv University

When/where: Thursday, October 30th, 4:15PM, 200 AKW

We present an algebraically accurate algorithm for detection and estimation of cyclic and dihedral symmetries. Both symmetry types can be detected by the special patterns they induce on the object's Fourier transform. These patterns are effectively detected and analyzed using the pseudo-polar Fourier transform, which rapidly computes the Fourier transform of the object on a near-polar grid. The algorithm detects all symmetry axes of the object using two phases. Firstly, the algorithm detects the number of symmetry axes using robust non-parameteric spectrum estimation. Secondly, it computes the exact locations of the symmetry axes. The presented algorithm is applicable both to centered and non-centered symmetries.