David E. Tyler

David E. Tyler¹*, Klaus Nordhausen²
¹ Department of Statistics, Rutgers University, New Brunswick, NJ
² Department of Statistics, University of Helsinki, Finland
*Corresponding Author Email: dtyler@stat.rutgers.edu

Abstract: The Hough Transform is a statistical method first introduced in the 1950’s. Although it was originally a line detection method used in computer vision, it has evolved into a more broadly used statistical method for image data and is often cited for its general robustness properties. The Hough Transform does share some interesting similarities to robust statistical methods developed in the 1960’s and later.

The goal of this talk is to review the Hough Transform, and related methods, and to discuss the relevance or irrelevance of their properties from a statistical perspective, and in particular from a statistical robustness perspective.

Keywords: Computer vision, cluster analysis, Hough transform, M-estimation, orthogonal regression