Video coding methods and apparatuses
AssigneeMicrosoft Corporation
InventorsAlexandros Tourapis, Feng Wu, Shipeng Li
Life of this patent
Estimating the term…
Abstract
Video coding methods and apparatuses are provided that make use of various models and/or modes to significantly improve coding efficiency especially for high/complex motion sequences. The methods and apparatuses take advantage of the temporal and/or spatial correlations that may exist within portions of the frames, e.g., at the Macroblocks level, etc. The methods and apparatuses tend to significantly reduce the amount of data required for encoding motion information while retaining or even improving video image quality.
Claims78
A computer-implemented method for use in encoding video data within a sequence of video frames, method comprising:
- encoding at least a portion of at least one reference frame to include motion information associated with said portion of said reference frame;
- defining at least a portion of at least one predictable frame that includes video data predictively correlated to said portion of said reference frame based on said motion information;
- encoding at least said portion of said predictable frame without including corresponding motion information and including mode identifying data that identifies that said portion of said predictable frame can be directly derived using at least said motion information associated with said portion of said reference frame, the mode identifying data defines a type of prediction model required to decode said encoded portion of said predictable frame;
- and
- wherein said type of prediction model includes an enhanced Direct Prediction model that includes at least one submode selected from a group comprising a Motion Projection submode, a Spatial Motion Vector Prediction submode, and a weighted average submode.
A computer-implemented method for use in encoding video data within a sequence of video frames, the method comprising:
- encoding at least a portion of at least one reference frame to include motion information associated with said portion of said reference frame;
- defining at least a portion of at least one predictable frame that includes video data predictively correlated to said portion of said reference frame based on said motion information;
- and
- encoding at least said portion of said predictable frame without including corresponding motion information and including mode identifying data that identifies that said portion of said predictable frame can be directly derived using at least said motion information associated with said portion of said reference frame, said motion information associated with said portion of said reference frame includes one or more of velocity information and acceleration information.
A computer-readable medium having computer-program instructions executable by a processor for performing acts comprising:
- encoding video data for a sequence of video frames into at least one predictable frame selected from a group of predictable frames comprising a P frame and a B frame, by:
- encoding at least a portion of at least one reference frame to include motion information associated with said portion of said reference frame;
- defining at least a portion of at least one predictable frame that includes video data predictively correlated to said portion of said reference frame based on said motion information;
- encoding at least said portion of said predictable frame without including corresponding motion information and including mode identifying data that identifies that said portion of said predictable frame can be directly derived using at least said motion information associated with said portion of said reference frame said mode identifying data defines a type of prediction model required to decode said encoded portion of said predictable frame;
- and
- wherein said type of prediction model includes an enhanced Direct Prediction model that includes at least one submode selected from a group comprising a Motion Projection submode, a Spatial Motion Vector Prediction submode, and a weighted average submode, and wherein said mode identifying data identifies said at least one submode.
A computer-readable medium having computer-implementable instructions for performing acts comprising:
- encoding video data for a sequence of video frames into at least one predictable frame selected from a group of predictable frames comprising a P frame and a B frame, by:
- encoding at least a portion of at least one reference frame to include motion information associated with said portion of said reference frame;
- defining at least a portion of at least one predictable frame that includes video data predictively correlated to said portion of said reference frame based on said motion information;
- and
- encoding at least said portion of said predictable frame without including corresponding motion information and including mode identifying data that identifies that said portion of said predictable frame can be directly derived using at least said motion information associated with said portion of said reference frame;
- and
- wherein said motion information associated with said portion of said reference frame includes information selected from a group comprising velocity information and acceleration information.
An apparatus for use in encoding video data for a sequence of video frames into a plurality of video frames including at least one predictable frame selected from a group of predictable frames comprising a P frame and a B frame, said apparatus comprising:
- memory for storing motion information;
- and
- logic operatively coupled to said memory and configured to encode at least a portion of at least one reference frame to include motion information associated with said portion of said reference frame, determine at least a portion of at least one predictable frame that includes video data predictively correlated to said portion of said reference frame based on said motion information, and encode at least said portion of said predictable frame without including corresponding motion information and including mode identifying data that identifies that said portion of said predictable frame can be directly derived using at least said motion information associated with said portion of said reference frame;
- wherein said mode identifying data defines a type of prediction model required to decode said encoded portion of said predictable frame;
- and
- wherein said type of prediction model includes an enhanced Direct Prediction model that includes at least one submode selected from a group comprising a Motion Projection submode, a Spatial Motion Vector Prediction submode, and a weighted average submode, and wherein said mode identifying data identifies said at least one submode.
A computer-implemented method for use in decoding encoded video data that includes a plurality of video frames comprising at least one predictable frame selected from a group of predictable frames comprising a P frame and a B frame, the method comprising:
- determining motion information associated with at least a portion of at least one reference frame;
- buffering said motion information;
- determining mode identifying data that identifies that at least a portion of a predictable frame can be directly derived using at least said buffered motion information;
- and
- generating said portion of said predictable frame using said buffered motion information;
- wherein said mode identifying data defines a type of prediction model required to decode said encoded portion of said predictable frame;
- and
- wherein said type of prediction model includes an enhanced Direct Prediction model that includes at least one submode selected from a group comprising a Motion Projection submode, a Spatial Motion Vector Prediction submode, and a weighted average submode.
The A computer-implemented method method for use in decoding encoded video data that includes a plurality of video frames comprising at least one predictable frame selected from a group of predictable frames comprising a P frame and a B frame, the method comprising:
- determining motion information associated with at least a portion of at least one reference frame, said motion information associated with said portion of said reference frame includes information selected from a group comprising velocity information and acceleration information;
- buffering said motion information;
- determining mode identifying data that identifies that at least a portion of a predictable frame can be directly derived using at least said buffered motion information;
- and
- generating said portion of said predictable frame using said buffered motion information.
A computer-readable medium having computer-program instructions executable by a processor for performing acts comprising:
- decoding encoded video data that includes a plurality of video frames comprising at least one predictable frame selected from a group of predictable frames comprising a P frame and a B frame, by:
- buffering motion information associated with at least a portion of at least one reference frame;
- determining mode identifying data that identifies that at least a portion of a predictable frame can be directly derived using at least said buffered motion information, said mode identifying data defines a type of prediction model required to decode said encoded portion of said predictable frame;
- generating said portion of said predictable frame using said buffered motion information;
- and
- wherein said type of prediction model includes an enhanced Direct Prediction model that includes at least one submode selected from a group comprising a Motion Projection submode, a Spatial Motion Vector Prediction submode, and a weighted average submode.
An apparatus for use in decoding video data for a sequence of video frames into a plurality of video frames including at least one predictable frame selected from a group of predictable frames comprising a P frame and a B frame, said apparatus comprising:
- memory for storing motion information;
- logic operatively coupled to said memory and configured to buffer in said memory motion information associated with at least a portion of at least one reference frame, ascertain mode identifying data that identifies that at least a portion of a predictable frame can be directly derived using at least said buffered motion information, and generate said portion of said predictable frame using said buffered motion information;
- wherein said made identifying data defines a type of prediction model required to decode said encoded portion of said predictable frame;
- and
- wherein said type of prediction model includes an enhanced Direct Prediction model that includes at least one submode selected from a group comprising a Motion Projection submode, a Spatial Motion Vector Prediction submode, and a weighted average submode.
Dependencies are inferred from the claim text.