Pseudo-invariant image transformations on a hexagonal lattice

Abstract

A novel data structure for a machine vision system is presented in this paper. The approach is inspired from anatomical considerations of the primate's vision system. The geometrical arrangements of cones (photo receptors) on a primate's retina can be described in terms of a hexagonal lattice. The importance of this lattice is that it possesses special computational features pertinent to the vision process.

Pseudo- (quasi-) invariant transformations are relevant to both biological and machine vision systems. They provide a means by which an image can be represented independent of a coordinate system. This facilitates the object recognition process. This paper describes a method of producing pseudo-invariant image transformations on a hexagonal lattice. A method of implementing the invariant transformations on a computer is provided. Application areas where the transformations can be employed are discussed.