Spatial extrapolation of pixel values in intraframe video coding and decoding
AssigneeMicrosoft Corporation
InventorSridhar Srinivasan
Life of this patent
Estimating the term…
Abstract
Techniques and tools for spatial extrapolation of pixel values in intraframe video encoding and/or decoding are described. For example, to predict the pixel values for a block of pixels, a video encoder or decoder uses spatial extrapolation from the pixel values of pixels in a causal neighborhood for the block of pixels.
Claims30
In a computer system, a computer-implemented method of processing blocks of pixels in a video frame of a video sequence, the method comprising:
- for a current block of plural pixels in the video frame of the video sequence,
- identifying plural context pixels in the video frame, wherein the plural context pixels form at least part of a causal neighborhood for the current block;
- predicting a spatial extrapolation mode for the current block based upon plural contextual spatial extrapolation modes, the plural contextual spatial extrapolation modes including a first contextual spatial extrapolation mode for a first causal block above the current block, the plural contextual spatial extrapolation modes further including a second contextual spatial extrapolation mode for a second causal block left of the current block;
- computing an actual spatial extrapolation mode for the current block based at least in part upon the predicted spatial extrapolation mode for the current block;
- and
- computing a spatial extrapolation according to the actual spatial extrapolation mode for pixel values of the plural pixels of the current block based at least in part upon pixel values of the plural context pixels, wherein the spatial extrapolation comprises one or more pixel value extrapolations from one or more of the plural context pixels into the current block to thereby spatially predict the pixel values of the plural pixels of the current block.
(canceled)
(canceled)
(canceled)
(canceled)
In a computer system, a computer-implemented method of processing a current block of pixels of a video frame of a video sequence, the method comprising:
- identifying plural context pixels of the video frame, wherein the plural context pixels form at least part of a causal neighborhood for the current block;
- selecting a spatial extrapolation mode for the current block from among plural spatial extrapolation modes, the plural spatial extrapolation modes including a horizontal mode, a vertical mode, plural diagonal modes and a flat mode, wherein the flat mode uses a DC value based at least in part upon pixel values of the plural context pixels, and wherein the plural diagonal modes include at least one diagonal mode in a right upward direction toward the current block, plural diagonal modes in a right downward direction toward the current block, and plural diagonal modes in a left downward direction toward the current block;
- and
- computing a spatial extrapolation for pixel values of the pixels of the current block according to the selected spatial extrapolation mode, wherein the spatial extrapolation comprises one or more pixel value extrapolations from one or more of the plural context pixels into the current block to thereby spatially predict the pixel values of the pixels of the current block.
(canceled)
-30. (canceled)
In a computer system, a computer-implemented method of processing a current block of pixels of a video frame of a video sequence, the method comprising:
- predicting a spatial extrapolation mode for the current block based upon plural contextual spatial extrapolation modes, the plural contextual spatial extrapolation modes including a first contextual spatial extrapolation mode for a first causal block above the current block in the video frame, the plural contextual spatial extrapolation modes further including a second contextual spatial extrapolation mode for a second causal block left of the current block in the video frame;
- and
- computing an actual spatial extrapolation mode for the current block, wherein the computing is based at least in part upon the predicted spatial extrapolation mode for the current block.
(canceled)
(canceled)
(canceled)
-65. (canceled)
Dependencies are inferred from the claim text.