Reducing overhead for multiple-hypothesis temporal prediction
A method is provided for inter-coding video in which transmission bandwidth requirements associated with second motion vectors for bi-directional temporal prediction is reduced. In the method, vector information for one of motion vectors for multi-directional temporal prediction can be transmitted together with information on how to derive or construct the second motion vectors. Thus, rather than sending express information regarding each of the plurality of motion vectors, express information related to only one motion vector along with information related to reconstruction/derivation of the second motion vectors is transmitted, thus reducing bandwidth requirements and increasing coding efficiency.
1. A method of decoding a video included in a bitstream by a decoder, comprising:
(a) receiving said bitstream indicating how a coding tree unit was partitioned into coding units according to a quad tree plus multi tree structure that allows a square parent node to be split with a quaternary tree partitioning that splits said square parent node in half in both horizontal and vertical directions to define leaf nodes that are square in shape each of which are the same size, wherein said quad tree plus multi tree structure allows one of said leaf nodes to be split based upon one selected from a group consisting of,
(i) a symmetric binary tree partitioning that splits one of said leaf nodes of said quaternary tree partitioning in half in either a horizontal direction or a vertical direction resulting in two blocks that are the same size as leaf nodes, and
(ii) an asymmetric tree partitioning that splits one of said leaf nodes of said quaternary tree partitioning in either a horizontal direction or a vertical direction resulting in three blocks two of which are different in size than a third one of said three blocks as leaf nodes;
(b) identifying final coding units to be decoded represented by leaf nodes of the quad tree plus multi tree structure;
(c) wherein one of said final coding units is a rectangular coding unit included in a bi-predictive B-slice of a current frame of said video;
(d) receiving a first motion vector associated with said coding unit of said bi-predictive B-slice of said current frame of said video referencing a temporally previous reference slice of a temporally previous reference frame relative to said current frame of said coding unit;
(e) receiving a second motion vector associated with said coding unit of said bi-predictive B-slice of said current frame of said video referencing a temporally future reference slice of a temporally future reference frame relative to said current frame of said coding unit;
(f) applying an optical flow between said temporally future reference slice and said temporally previous reference slice to perform a sample based motion modification after motion compensation based upon at least one of said second motion vector and said first motion vector; and
(g) decoding said coding unit based upon said sample based motion modification as a result of said applying said optical flow between said temporally previous reference slice and said temporally future reference slice.