Probabilistic finite modeling of stochastic estimation of image inter-frame geometric deformations

Abstract

The paper proposes an approach to probabilistic finite modeling of the process of stochastic estimation of the parameters of interframe geometric deformations of images. As the quantities characterizing the state of the stochastic procedure at each iteration of the estimation, the demolition probabilities of the formed parameter estimates (improvements, deterioration, or no change in the vector of estimates in the parameter space) are used. At the same time, the models of images and noises are given by probability distribution densities and autocorrelation functions. This made it possible to simplify the description of a complex set of factors influencing the formation of estimates. The use of adaptive limitation of the parameter space used for modeling reduces the computational cost several times while maintaining the adequacy of the model. In practice, probabilistic modeling can be used to find the accuracy and probability characteristics of stochastic algorithms for estimating interframe deformations of images for a given number of iterations.