Partition-Based Intra Coding for Low-Overhead Accurate Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing block-based video codecs face challenges in achieving higher coding efficiency due to the trade-off between signaling overhead and prediction accuracy, particularly in intra-prediction, where larger blocks result in less accurate spatial prediction and increased signaling overhead.
Innovation Solution
The implementation of a flag-controlled intra-prediction mode that allows for partitioning of blocks into transform partitions, enabling parallel processing and flexible intra-prediction strategies such as global, sequential, or grouped processing based on block size and shape, to optimize prediction residual transformation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If blocks are partitioned into multiple sub-partitions for intra-prediction, then prediction accuracy is improved, but signaling overhead increases
Solution Approach 1:
The block is divided into multiple sub-partitions (e.g., 2x2, 2x1, 1x2 configurations) to improve prediction accuracy for inner block samples. Each sub-partition can have its own intra-prediction mode, allowing more precise local prediction while managing signaling overhead through selective application.
Solution Approach 2:
Different intra-prediction modes are applied to different sub-partitions based on local characteristics. The encoder can select from multiple IPMs for each sub-partition, allowing locally optimized prediction that adapts to regional variations in the block while controlling overall signaling overhead.
2Measurement precision
If the number of sub-partitions per block is increased, then prediction accuracy is improved, but throughput decreases
Solution Approach 1:
The number and configuration of sub-partitions are made dynamic rather than fixed. The encoder can adaptively select the partitioning scheme (e.g., 2x2, 2x1, 1x2, or no partitioning) based on block characteristics, allowing optimization of both accuracy and throughput for different scenarios.
Solution Approach 2:
The partitioning parameters (number of partitions, their dimensions, and associated IPMs) are changed adaptively based on block size, content characteristics, and rate-distortion optimization. This allows the system to adjust between more partitions for accuracy-critical blocks and fewer partitions for throughput-critical blocks.
3Loss of information
If intra-prediction is applied to larger blocks, then signaling overhead is reduced, but prediction accuracy decreases
Solution Approach 1:
Large blocks are segmented into multiple sub-partitions, each with its own intra-prediction mode. This allows the benefits of large-block processing (reduced signaling overhead compared to fine-grained partitioning) while maintaining prediction accuracy through localized mode selection in each sub-partition.
Solution Approach 2:
The intra-prediction system is designed to handle multiple block sizes and partitioning configurations uniformly. The same intra-prediction mechanisms work across different block dimensions, allowing flexible application from small to large blocks while maintaining efficiency.
Data Source
AI summary
A method for decoding a block of a picture includes decoding, from a data stream, an intra-coding mode for the block, deriving a predictor for a prediction partition of the block using at least one already reconstructed samples neighboring the prediction partition and the intra-coding mode, deriving a prediction residual for each of the at least two transform partitions included in the prediction partition, and combining the predictor of the prediction partition and the prediction residual for each of the at least two transform partitions included in the prediction partition to reconstruct the prediction partition. The block is partitioned into multiple transform partitions based on a partition dimension flag, and the prediction partition includes at least two transform partitions of the multiple transform partitions.