Conditions of Rice statistical model applicability and estimation of the Rician signals parameters by maximum likelihood technique

Abstract

The paper develops a theory of a new so-called two-parametric approach to the random signals' analysis and processing. A mathematical simulation and the task solutions comparison have been implemented for the Gauss and Rice statistical models. The applicability of the Rice statistical model is substantiated for the tasks of data and images processing when the signals envelope is being analyzed. A technique is developed and theoretically substantiated for solving the task of the noise suppression and initial image reconstruction by means of joint calculation of both statistical parameters - an initial signals mean value and noise dispersion - based on the maximum likelihood method within the Rice distribution. The peculiarities of this distributions likelihood function and the following from them possibilities of the signal and noise estimation have been analyzed.

Keywords

• random signal
• Rice distribution
• Gauss distribution
• maximum likelihood technique
• signal-to-noise ratio