E.P. Petrov, E.V. Medvedeva, A.P. Metelyov

The adaptive algorithm of combined nonlinear filtration of digital grayscale images videosequence allows to restore digital grayscale images with the signal-to-noise ratio by power of much less than 1 dB. In synthesis of the filtration algorithm the digital grayscale image is represented by g-digit binary numbers and it is transmitted digit-by-digit (in parallel or in sequence) as binary signals. Using approximation of digit binary image sequence by three-dimensional Markov chain with two states, an equation of nonlinear filtration for the digit binary image sequence of the l-th digit in the digital grayscale image is received. A set of these equations forms a system of g equations for filtration of the digital grayscale images videosequence. Adaptation is carried out by calculating estimations of elements in transition probability matrixes and their substitution in the algorithm. Time of receiving estimations close to the true ones does not exceed three  five slots at the signal-to-noise ratio by power on the input of the receiving device from ρ2=0 dB to ρ2=12 dB. After processing by the adaptive nonlinear filter on the filtered digital grayscale images there are distortions in the form of noise spike with casual values of impulses. The probability of occurrence of such distortions by that more than is less signal-to-noise ratio. For elimination of noise spike the simple median filter in realization is used. The combined method of nonlinear filtration of correlated video images allows to improve visual quality of video images and possesses high efficiency at the signal-to-noise ratio from 12 dB to 0 dB. The developed method is simple in realization and demands small computing resources.