Характеристики двумерной оценки положения фрагмента изображения на основе алгоритма Ньютона – Рафсона

Abstract

This paper considers a two-dimensional estimation of fragment position of image in a frame by using Newton – Rafson type algorithms. These algorithms are implemented as some types of discriminators. Considered a two-dimensional discrete autocorrelation function of test object and the conditions of its factorization to one-dimensional components are defied. The behavior of discriminator characteristics were analyzed for some types of discriminators. It is shown that probability density of process on discriminator output is substantially non-Gaussian with «heavy tails». By using statistical modeling the theoretical calculations are confirmed.