Filtration of Images Corrupted by Mixture of Additive and Impulsive Noise Based on the Fractional Brownian Motion Model

A novel approach to image filtering when a small image fragment can be represented as fractional brownian motion (fBm) with constant parameters which however depends on fragment position is proposed in this article. The observation model assumes image to be registrated in the presence of mixture of additive and impulsive noise. The case of incomplete data is also considered. For the observation model, the algorithms of model parameters estimation: are derived parameters of fBm, additive noise variance, the probability of impulse noise corruption; image filtering algorithm and impulse noise detection algorithm. The comparative analysis of filter performance is done using quantitative criteria. It is shown that the proposed algorithm outperforms or works as good as chosen for comparison filters. The possibility of quantitative criteria estimation without using a standard image is marked.