Fractional pixel motion estimation of video signals

Claim 1

An improved method of fractional pixel motion estimation for a video compression signal encoder wherein a block of pixels in a current video image of an input video signal is compared with displaced blocks of pixels at different locations in a reference video image to provide a compressed video signal by generating a distortion function for each location and selecting a displacement vector that gives a minimum value of the distortion function as a motion vector for the block of pixels, the motion vector being used by the video compression signal encoder to predict the reference video image, the improvement comprising the steps of: fitting a general surface to the distortion function so that the fitted surface equals the distortion function at integer pixel locations surrounding a center pixel of the block having the minimum value of the distortion function; estimating the distortion function at fractional pixel locations surrounding the center pixel; and selecting, as the motion vector, the displacement vector that gives a minimum value for the fractional pixel distortion function from among the center pixel and the fractional pixel locations surrounding the center pixel, the selected motion vector being transmitted as part of the compressed video for the block of pixels.