FIELD: physics, video.
SUBSTANCE: invention relates to video encoding and decoding techniques. Disclosed is a method of encoding video using motion compensation with fractional accuracy. The method includes a step of performing motion search to obtain a motion vector, a step of computing reference probabilities in positions of fractional pixels denoted by the obtained motion vector. The method also includes a step of dividing interpolation positions which are the positions of fractional pixels into multiple groups based on the computed reference probabilities; a step of selecting, for each of the groups of interpolation positions, interpolation filter coefficients to be used to generate an interpolated prediction image from among multiple interpolation filter coefficient candidates.
EFFECT: high encoding and decoding efficiency by adapting interpolation positions depending on the property of the input image.
9 cl, 14 dwg, 2 tbl
