Back to results
US 10,757,431 B2Granted
Method and apparatus for processing video
AssigneeChips & Media, Inc
InventorsHyun Gyu Kim, Tae Young Jung, Dae Yeon Kim
Life of this patent
Estimating the term…
Abstract
The present invention relates to an apparatus for processing a video, comprising: a video central processing unit which communicates with a host, receives a bitstream and parses header information; and a free scan module which generates free scan information by a free scan of a video bitstream inputted from the host according to the control of the video central processing unit, wherein the video central processing unit rearranges the bitstream on the basis of the header information and the free scan information and performs task scheduling so as to perform parallel decoding of the rearranged bitstream by a predetermined unit.
Claims19
A device for processing videos, comprising:
- a video central processing unit which is able to communicate with a host and is provided to receive a bitstream and parse header information;
- and
- a prescan module which is configured to perform a prescan operation for generating prescan information, when the bitstream is determined to be not encoded through a wavefront parallel processing (WPP) method, based on a predetermined parsed syntax element from the bitstream received from the host, by prescanning the bitstream according to a simplified entropy decoding for extracting only initialization information, engine information or context information at a predetermined position for a parallel decoding processing irrespective of any encoding unit structure of the bitstream, and to omit the prescan operation, when the bitstream is determined to be encoded through the wavefront parallel processing (WPP) method,
- wherein the video central processing unit realigns the bitstream based on the prescan information and the header information, and performs a task scheduling to decode in parallel the realigned bitstream in predetermined unit,
- wherein the prescan module generates the prescan information by classifying a video stream including a boundary between a slice and a tile into one or more prescan segments in a unit of a row of a coding tree unit (CTU), the prescan segments being irrespective of the structure of the tile and slice,
- wherein the prescan information in each of the prescan segments includes a CTU context index (context_idx), and
- wherein, when a first block of each of the prescan segments is a start point of the slice or corresponds to a start point of the tile, the context_idx has an invalid value so as to give an implication on an initialization of the context for the entropy decoding.
4 more elements
A method for processing videos, the method comprising:
- parsing header information by receiving a bitstream while communicating with a host;
- performing a prescan operation for generating prescan information, when the bitstream is determined to be not encoded through a wavefront parallel processing (WPP) method, based on a predetermined parsed syntax element from the bitstream received from the host by prescanning the bitstream according to a simplified entropy decoding for extracting only initialization information, engine information or context information at a predetermined position for a parallel decoding processing irrespective of any encoding unit structure of the bitstream, and omitting the prescan operation, when the bitstream is determined to be encoded through the wavefront parallel processing (WPP) method;
- and
- realigning the bitstream based on the prescan information and the header information, and carrying out a task scheduling to decode in parallel the realigned bitstream per unit,
- wherein the prescan operation for generating the prescan information is performed by classifying a video stream including a boundary between a slice and a tile into one or more prescan segments in a unit of a row of a coding tree unit (CTU), the prescan segments being irrespective of the structure of the tile and slice,
- wherein the prescan information in each of the prescan segments includes a CTU context index (context_idx), and
- wherein, when a first block of each of the prescan segments is a start point of the slice or corresponds to a start point of the tile, the context_idx has an invalid value so as to give an implication on an initialization of the context for the entropy decoding.
4 more elements
Dependencies are inferred from the claim text.