Intra Subpartition Coding With Two-Stage Transform and Quad-Splits
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding technologies face limitations in efficiently encoding and decoding blocks using intra subpartitions due to restrictions on the use of Low Frequency Non-Separable Transform (LFNST) and limited partitioning options, leading to suboptimal coding efficiency and increased bit rates.
Innovation Solution
Implementing a two-stage transform using a primary and secondary transform at the transform unit level for intra subpartitions, allowing for quad-splits and partition-specific intra prediction modes, which enhances coding efficiency by better condensing non-zero coefficients and reducing bit rates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single transform is used for intra subpartitions, then the device complexity is reduced, but the coding efficiency deteriorates due to inability to better condense non-zero coefficients
Solution Approach 1:
The transform processing is segmented into two distinct stages: a primary transform that processes the entire block, and a secondary transform that processes specific partitions. This segmentation allows different transform operations to be applied to different regions, optimizing coefficient condensation while maintaining manageable complexity through structured processing steps.
Solution Approach 2:
The secondary transform is applied selectively to specific partitions (e.g., top-left partition) rather than uniformly across the entire block. This local quality approach applies enhanced transform processing only where needed to condense non-zero coefficients, avoiding unnecessary complexity in regions where the primary transform suffices.
2Device complexity
If intra subpartitions are restricted to limited partitioning options, then the device complexity is reduced, but the coding efficiency deteriorates due to inability to adapt to different block characteristics
Solution Approach 1:
The partitioning structure is made dynamic by introducing conditional logic that selects between different partitioning modes (e.g., horizontal split, vertical split, or no split) based on block characteristics. The secondary transform application is also dynamic, being enabled or disabled based on partitioning decisions and coefficient distribution patterns, allowing adaptation without exhaustive complexity.
3Device complexity
If LFNST is completely disabled for intra subpartitions, then the device complexity is reduced, but the coding efficiency deteriorates due to loss of transform condensation capability
Solution Approach 1:
The system provides self-service by automatically determining whether to apply the secondary transform based on intrinsic properties of the input data (e.g., position of non-zero coefficients, partition characteristics). This self-determination mechanism eliminates the need for complex external control logic while maintaining the ability to condense coefficients when beneficial.
Data Source
AI summary
Decoder for decoding a picture, wherein blocks into which the picture is subdivided are assigned to a set of prediction types, which include intra prediction and inter prediction. For each of predetermined intra-predicted blocks, the decoder is configured to derive, from the data stream, an information on a partitioning of the respective predetermined intra-predicted block into partitions. The decoder is configured to derive, from the data stream, for each partition of the respective predetermined intra-predicted block, a partition-specific prediction residual signal related to a spatial domain prediction residual signal of the respective partition of the respective predetermined intra-predicted block via a predetermined transform. The predetermined transform can be identified out of a set of transforms including a first transform and a second transform, which equals a concatenation of a primary transform and a secondary transform applied onto a subset of coefficients of the primary transform.