Mixed Secondary Transform Sets for Adaptive Video Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Next-generation video contents with high spatial resolution, high frame rate, and high dimensionality require more efficient memory storage, memory access rate, and processing power, necessitating improved coding tools, particularly in transform efficiency and complexity.
Innovation Solution
A method for configuring a secondary transform set considering various conditions, including intra prediction mode and coding block size, with mixed secondary transform sets comprising 8×8 and 4×4 transform kernels, and a binarized secondary transform index using a truncated unary scheme.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a single transform kernel size (e.g., 4×4) is used for all blocks, then device complexity is reduced, but coding efficiency deteriorates due to inability to adapt to different block characteristics
Solution Approach 1:
The patent applies different transform kernel sizes (4×4 or 8×8) to different regions of the transform block based on local characteristics. Specifically, an 8×8 transform kernel is applied to the top-left 8×8 region when the block size is 16×16 or larger, while 4×4 kernels are applied to other regions. This local differentiation improves coding efficiency by adapting to local frequency characteristics without requiring complex global transform selection.
Solution Approach 2:
The transform block is segmented into multiple regions with different transform kernel applications. The patent divides a 16×16 or larger block into a top-left 8×8 region and other regions, applying different transform kernels to each segment. This segmentation allows the system to achieve adaptability to different block characteristics while maintaining relatively simple device complexity through predefined segmentation rules.
2Productivity
If larger transform kernels (8×8) are used for all blocks, then coding efficiency improves, but computational complexity increases
Solution Approach 1:
The patent applies the larger 8×8 transform kernel only partially - specifically to the top-left 8×8 region of the transform block - rather than to the entire block. This partial application achieves the coding efficiency benefits of larger kernels where they are most beneficial (in the high-energy top-left region) while avoiding the full computational cost of applying 8×8 kernels to the entire block, thus balancing coding efficiency and computational complexity.
3Productivity
If multiple transform sets are configured for different prediction modes, then coding efficiency improves, but device complexity increases due to more transform kernel storage and selection
Solution Approach 1:
The patent creates a universal transform processing mechanism that handles both 4×4 and 8×8 transform kernels through a unified framework. The same inverse transform processing unit processes both kernel sizes by receiving control parameters (transform kernel size indicators) that dictate the specific operation. This multi-functionality allows the system to support multiple transform sets and adapt to different prediction modes while avoiding the need for separate dedicated processing paths for each transform type, thus reducing device complexity.
Data Source
AI summary
Disclosed is a method for decoding a video signal, including: generating a dequantized transform block by performing dequantization for a current block; acquiring an intra prediction mode of the current block; determining a secondary transform set applied to the current block among a plurality of secondary transform sets based on the intra prediction mode; deriving a transform kernel applied to the current block in the determined secondary transform set; and performing secondary transform for a top left specific area of the dequantized transform block by using the derived transform kernel, in which the plurality of secondary transform sets may include at least one mixed secondary transform set.