Image Transform Kernel Selection for Block-Based Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image compression technologies face challenges in efficiently encoding and decoding high-resolution and high-quality images, particularly in handling transforms that do not effectively utilize non-separable primary transforms and lack efficient signaling of transform kernels based on encoding parameters.
Innovation Solution
An image decoding method and apparatus that utilizes non-separable primary transforms, where the choice of transform is based on the size of the current block, employing non-separable primary and secondary transforms or separable primary transforms as needed, and determines the number of transform coefficients based on block size, improving encoding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If non-separable primary transform is applied to all block sizes, then transform performance is improved, but device complexity and processing overhead increase
Solution Approach 1:
The patent segments block sizes into multiple groups (first group and second group) with different transform application rules. The first group includes block sizes where non-separable primary transform is applied, while the second group includes block sizes where only separable primary transform or secondary transform is applied. This segmentation allows selective application of computationally intensive transforms only where beneficial, resolving the contradiction between transform performance and processing overhead.
Solution Approach 2:
The patent applies different transform qualities to different block sizes based on their specific characteristics. For certain block sizes (first group), the higher-quality non-separable primary transform is applied, while for other block sizes (second group), the standard separable transform is sufficient. This local quality adjustment optimizes overall performance without uniformly increasing complexity across all blocks.
2Productivity
If non-separable primary transform is used, then encoding efficiency is improved, but signaling complexity increases
Solution Approach 1:
The patent divides block sizes into groups and applies non-separable primary transform selectively to the first group. This segmentation approach improves encoding efficiency for appropriate block sizes while avoiding the signaling overhead for block sizes where the transform provides minimal benefit, thus resolving the contradiction between encoding efficiency and signaling complexity.
3Measurement precision
If transform kernel is always signaled explicitly, then transform accuracy is improved, but bitstream overhead increases
Solution Approach 1:
The patent segments the signaling approach based on block size groups. For the first group of block sizes, explicit transform kernel signaling is performed to ensure accuracy. For the second group, default or implicit signaling is used, reducing bitstream overhead. This segmented approach resolves the contradiction between transform accuracy and bitstream overhead by applying explicit signaling only where necessary.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
This image decoding method and device according to the present disclosure can: acquire residual information from a bitstream; derive transform coefficients of the current block on the basis of the residual information; derive residual samples of the current block by performing inverse transform on transform coefficients of the current block; and reconstruct the current block on the basis of the residual samples of the current block. Here, the inverse transform is performed on the basis of at least one from among a non-split primary transform, a non-split secondary transform and a split primary transform, and the non-split primary transform can be applied on the basis of the size of the current block.