Модели и алгоритмы распознавания цифровых изображений в условиях воздействия деформирующих и аддитивных искажений

Abstract

The problem of digital image recognition under influence of warping and additive noise has been considered. Parametric, based on Gaussian approximation, and nonparametric, based on kernel density estimation of likelihood functions, along with neural network, recognition algorithms have been synthesized and analyzed. Modified mixed nonparametric estimation of likelihood functions, based on convolution of kernel estimation for original warped images data set with additive noise probability density function, has been proposed. It has been theoretically shown and experimentally proved that use of mixed nonparametric estimation equiv-alents to data sets augmentation according to known statistical model of additive noise. These algorithms were simulated and compared at different values of signal-to-noise ratios and additive noise correlation coefficients.