Practical decision making based on joint geometric-probabilistic analysis

Abstract

This article gives an outline of the novel scheme for increasing performance of object recognition based on approximation of high-resolution wavelet coefficients with morphological deformation operators. The affectivity of statistical pre-processing techniques such as morphological boundary extraction is excavated. The dependency between parent and child wavelet coefficients of natural images was exploited by a generalized Bayesian-based algorithm. The basic idea is to choose upper bands wavelet coefficients between transform outputs of original and morphological presentation of image. It can be applied without any prior knowledge about probability distributions of wavelet coefficients. This is the simplest way to use novel scheme. In other way, the problem can be considered as the estimation of high resolution wavelet coefficients using a priori probability distribution with Bayesian estimation techniques, such as the maximum a posterior (MAP) estimator.