The invention teaches correlation analysis of the stochastic relationship between video images for use in motion estimation. On the basis of the method of analysis of correlation values as temporal series, a method was created for the analysis of the dynamics of the objects in the scenes, the isolation of quasi-static fragments of the images and the use of the obtained information for the construction of optimal procedures for making decisions in the control of motion estimation and the isolation of different types of motion of the objects. The dynamics of motion of objects or blocks in a video scene can be predicted, and the results of the predicting permit the isolation in motion of each element of the scene--the following components: camera motion, object motion, object distortion thereby allowing the possibility of isolating groups of blocks for which a given coding strategy has maximum effectiveness.