Video Coding Prediction Buffering Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face challenges in efficiently managing prediction information buffering, leading to increased memory requirements and reduced parsing throughput due to the need to store prediction data from all neighboring blocks, which can hinder parallel processing and increase latency.
Innovation Solution
The proposed solution involves determining prediction information for a block based on neighboring blocks below a top row within a coded unit, excluding prediction information from above-neighboring blocks, thereby reducing the amount of data buffered and enhancing parsing efficiency by eliminating dependencies between processing threads.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If prediction information from all neighboring blocks is buffered, then prediction accuracy is improved, but memory requirements increase
Solution Approach 1:
The patent segments the buffering process by introducing separate buffer structures (line buffer and field buffer) that store prediction information in a structured manner. This segmentation allows efficient access to specific neighboring block data without buffering all possible prediction information simultaneously, thus maintaining prediction accuracy while reducing overall memory requirements.
Solution Approach 2:
The patent extracts only the necessary prediction information from neighboring blocks that is actually needed for the current block's prediction. By selectively taking out and storing only relevant prediction data (而非 all neighboring block data), the system maintains prediction accuracy while minimizing memory usage.
2Measurement precision
If prediction information from all neighboring blocks is buffered, then prediction accuracy is improved, but parsing throughput is reduced
Solution Approach 1:
The patent performs preliminary actions by pre-calculating and storing prediction information from reference blocks in dedicated buffers before the actual decoding process. This preliminary preparation allows the parser to quickly access ready-computed prediction data without performing complex calculations during parsing, thus maintaining prediction accuracy while significantly improving parsing throughput.
3Measurement precision
If prediction information from above-neighboring blocks is used, then prediction accuracy is improved, but processing dependencies increase
Solution Approach 1:
The patent segments the prediction information storage into separate line buffers and field buffers, organizing data by spatial location rather than processing dependency. This segmentation allows the decoder to access prediction information from above-neighboring blocks without creating complex inter-thread dependencies, as each buffer is independently managed and accessed.
4Measurement precision
If prediction information from all neighboring blocks is buffered, then prediction accuracy is improved, but latency is increased
Solution Approach 1:
The patent performs preliminary computation and storage of prediction information in dedicated buffers before the actual decoding occurs. By pre-calculating and storing prediction data from reference blocks, the system eliminates the need for complex real-time calculations during decoding, thus maintaining high prediction accuracy while significantly reducing decoding latency.
Data Source
AI summary
In an example, aspects of this disclosure relate to a method of coding video data that generally includes determining prediction information for a block of video data, where the block is included in a coded unit of video data and positioned below a top row of above-neighboring blocks in the coded unit, and where the prediction information for the block is based on prediction information from one or more other blocks in the coded unit but not based on prediction information from any of the top row of blocks in the coded unit. The method also generally includes coding the block based on the determined prediction information.