Abstract: Video images are communicated between two stations using a method for predicting motion uses up to N frames of the video data, where N is a positive integer and each frame is represented as a plurality of image segments. According to one implementation, combined are at least two spatial displaced image segments from previously decoded frames corresponding to time instances on one side of a temporal axis; and the video data is communicated using the spatial displaced image segments along with corresponding delays as side information. Other aspects of the disclosure include combining using a Lagrangian cost function, and an iterative search method to minimize Lagrangian cost function.

Abstract: Coding and decoding of video images includes the use of multiple reference pictures and determining motion vector and time delay information for the video images. Video images are quantized and transmitted with the motion vector and time delay information from a first station to a second station. A new video image is predicted for display at the second station as function of the motion parameter codes and multiple reference pictures. The invention may be used in connection with a video communication and/or video conferencing system as a long-term memory predictor to improve image-coding efficiencies and motion compensation. The long-term memory prediction extends the spacial displacement utilized in block-based hybrid video coding by a variable time delay. The arrangement and method employ a proposed bit allocation scheme to provide increased coding efficiencies of the video coder.