Imaging applications of stochastic minimal graphs

Abstract

This paper presents an overview of some of the theory and application of stochastic minimal graphs in the context of entropy estimation for imaging applications. Stochastic graphs which span a set of extracted image features can be constructed to yield consistent estimators of Jensen's entropy difference for between pairs of images. Unlike traditional plug-in entropy estimates based on density estimation, stochastic graph methods provide direct estimates of these quantities. We review the stochastic graph approach to entropy estimation, compare convergence rates to that of plug-in estimators, and discuss a geo-registration application.