The method of a nonlinear three-dimensional filtering of the video images distorted by additive white Gaussian noise is considered. Improvement of noise immunity of reception of noisy images achieved through the effective use of statistical redundancy video-sequence. The method is used filtering theory of conditional Markov processes. The g-bases digital halftone images sequence represented as the set of bit binary images, which is approximated a three-dimensional Markov chain. To improve the quality of the reconstructed image, sliding window was using to evaluate the statistical characteristics of each local area within and between video frames. The results of modeling confirming efficiency of the developed method are given.