IntraBC Coding Reference Area Restriction for Parallel Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The full-frame IntraBC mode in video coding introduces data dependency issues that prevent parallel processing in wavefront parallel processing (WPP) and slice/tile-based parallel processing, limiting the ability to process video blocks concurrently.
Innovation Solution
Restricting the search range for IntraBC mode to a ladder-shaped reference area that includes previously processed blocks and applying block vector clipping to ensure valid reference data, allowing for parallel processing while maintaining improved coding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If full-frame IntraBC mode is used to improve coding efficiency, then compression performance is improved, but data dependency increases preventing parallel processing
Solution Approach 1:
The patent divides the reference area for IntraBC mode into separate non-overlapping regions for different processing parallel regions. Each parallel region can independently select reference blocks from its own designated area without depending on data from other parallel regions, enabling concurrent processing while maintaining coding efficiency through adequate reference options.
2Manufacturing precision
If search range is expanded to improve prediction accuracy, then coding efficiency is improved, but processing complexity and data dependency increase
Solution Approach 1:
The patent defines different reference area configurations for different parallel processing regions (tiles or slices). Each region has its own optimized reference area that provides sufficient prediction accuracy for that specific region while avoiding the complexity of a unified global search space. This localized approach maintains prediction quality without requiring complex coordination across the entire frame.
3Adaptability or versatility
If reference area includes blocks from multiple parallel regions, then coding flexibility is improved, but data dependency between regions increases
Solution Approach 1:
The patent extracts and removes the data dependency between parallel regions by assigning each region its own independent reference area. The reference blocks needed for each parallel region are extracted from within that region's own boundaries, eliminating the need to access or wait for data from other parallel regions. This maintains coding flexibility within each region while ensuring independent processing reliability.
Data Source
Figure 1~2
Figure 3~4
Figure 5A~5C
AI summary
A method of IntraBC coding using restricted reference area is disclosed. A reference block is selected from an available ladder-shaped reference area comprising previously processed blocks before the current working block in the current CTU row and previously processed blocks in one or more previous CTU rows. A location of a last previously processed block of a second previous CTU row that is one CTU row farther away from the current CTU row than a first previous CTU row is always in a same vertical location or after a same vertical position of a last previously processed block of the first previous CTU row. The current picture may be partitioned into multiple CTU rows for applying wavefront parallel processing (WPP) on the multiple CTU rows, where the current working block corresponds to a current working block. Similar restrictions may also be applied to slice/tile-based parallel processing.