Adaptive Multiple Transform Block Splitting for Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding technologies, such as HEVC, face limitations in transform selection and block partitioning, particularly for chroma components and intra-inter prediction modes, which affect coding efficiency and hardware implementation costs.
Innovation Solution
The proposed solution introduces an Adaptive Multiple Transform (AMT) scheme using DST-7 and DCT-8 transforms for residual coding, along with a method to split blocks into sub-blocks for efficient transform application, allowing for the use of identity transforms and transform skipping modes to harmonize transform processes across luma and chroma components.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multiple transform selection (MTS) is applied to chroma components and intra-inter prediction modes, then coding efficiency is improved, but hardware implementation complexity increases
Solution Approach 1:
The transform selection process is segmented by prediction mode (intra vs. inter) and chroma component type. Different transform sets are defined for different modes, allowing the system to select appropriate transforms without evaluating all possibilities, thus reducing hardware complexity while maintaining coding efficiency.
Solution Approach 2:
Different transform kernels are applied to different regions (luma vs. chroma components) based on their specific characteristics. The patent defines mode-specific transform sets that apply appropriate transforms locally to each component type, optimizing coding efficiency for each region without uniformly increasing complexity across the entire processing pipeline.
2Productivity
If adaptive multiple transform (AMT) scheme is implemented with transform splitting, then transform processing efficiency is improved, but computational complexity increases
Solution Approach 1:
The transform block is segmented into sub-blocks when the block size exceeds a threshold (e.g., 32x32). Each sub-block undergoes independent transform processing with potentially different transform types, allowing efficient processing of large blocks while managing computational complexity through divide-and-conquer strategy.
Solution Approach 2:
The transform processing dynamically adapts based on block size and prediction mode. The system selectively applies transform splitting only when beneficial (larger blocks), and chooses different transform types (DST-7, DCT-8, identity) based on the specific mode, creating a dynamic processing pipeline that balances efficiency and complexity.
3Device complexity
If transform skipping mode is used to harmonize transform processes, then implementation consistency is improved, but transform flexibility is reduced
Solution Approach 1:
The transform processing architecture is designed to be universal by supporting multiple transform types (DST-7, DCT-8, identity transform) within a single unified framework. The transform skipping mode serves as a special case (identity transform) that can be applied conditionally, allowing the system to maintain implementation consistency while preserving flexibility through a multi-functional transform engine.
Data Source
AI summary
A next-generation video decoding technique performed by at least one computer processor including determining whether at least one parameter of a block is less than or equal to a threshold, signaling, in response to determining the parameter(s) of the block is/are less than or equal to the threshold, a horizontal transform or vertical transform, splitting, in response to determining that the at least one parameter of the block is greater than the threshold, the block into sub-blocks, applying a first signaling scheme on a luma component and a second signaling scheme on a chroma component, performing ones of transforms on the sub-blocks, and decoding a video stream by using the sub-blocks upon which the ones of the transforms are performed, where a maximum block size of the first signaling scheme is different than a maximum block size of the second signaling scheme.