Intra-coding Rectangular Chroma Blocks Using Square Sub-blocks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding standards, such as HEVC, do not support efficient transform unit partitioning for chroma components in 4:2:2 and 4:4:4 color formats, leading to increased hardware complexity and inefficiencies in using rectangular transforms, which are not fully defined and available for all sizes and shapes of chroma block transforms.
Innovation Solution
The technique involves splitting rectangular chroma blocks into square sub-blocks to allow for the use of square transforms, enabling efficient encoding and decoding by processing coded block flags (CBFs) for each sub-block and intra-predicting one sub-block based on the reconstructed samples of the other, thereby optimizing transform unit partitioning and reducing hardware complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If rectangular transforms are used for chroma blocks in 4:2:2 and 4:4:4 formats, then the transform can directly match the block shape, but hardware complexity increases and the transforms are not fully defined for all sizes and shapes
Solution Approach 1:
The rectangular chroma block is divided into multiple square sub-blocks (e.g., two 8x8 sub-blocks from a 16x8 block). Each square sub-block is then transformed independently using standard square transforms (DST or DCT), avoiding the need for complex rectangular transforms while maintaining coding efficiency.
2Productivity
If rectangular chroma blocks are transformed directly, then the transform operation can be performed on the entire block, but the transforms are not fully defined and available for all sizes and shapes of chroma block transforms
Solution Approach 1:
The chroma block is segmented into square sub-blocks that match the supported transform sizes. This segmentation enables the use of well-defined square transforms (4x4, 8x8, 16x16) that are fully specified in the standard, ensuring compatibility and availability across all block sizes.
Solution Approach 2:
Different transform types (DST or DCT) can be selectively applied to different square sub-blocks based on local characteristics. The transform type is determined by the presence of coded block flags (CBFs) in adjacent blocks, allowing adaptive transform selection for each sub-block.
3Device complexity
If square transforms are used for rectangular chroma blocks by splitting into sub-blocks, then hardware complexity is reduced and transforms are well-defined, but additional processing steps are required for partitioning and reconstruction
Solution Approach 1:
The chroma block is partitioned into square sub-blocks before transformation. This segmentation simplifies the transform operation to use standard square transforms, reducing hardware complexity while the partitioning logic is automatically managed by the coding standard.
Solution Approach 2:
The chroma block is pre-partitioned into square sub-blocks before the transform operation. This preliminary action ensures that the subsequent transform can use standardized square transforms, and the partitioning information is prepared in advance for efficient processing.
4Productivity
If intra-prediction is performed for one sub-block using reconstructed samples from another sub-block, then coding efficiency is improved, but additional dependency and processing steps are introduced
Solution Approach 1:
The first square sub-block is reconstructed first using standard intra-prediction methods. This preliminary reconstruction provides reference samples that are then used for predicting the second sub-block, enabling efficient intra-prediction while following a deterministic processing order.
Solution Approach 2:
Intra-prediction for the second sub-block uses reference samples specifically from the reconstructed first sub-block when available. This local quality approach allows the prediction to adapt to the actual reconstructed content, improving coding efficiency by utilizing locally available information.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Systems, methods, and devices for video coding that may receive a rectangular chroma block are disclosed. The rectangular chroma block may be divided into a first square sub-block and a second square sub-block. The systems, methods, and devices may reconstruct the first square sub-block and intra-predict the second square sub-block using reference samples from the reconstructed first square sub-block.