Reduced-Eigenvector Transform Coding for High-Resolution Video
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face challenges in efficiently compressing digital video data while maintaining video quality, particularly with high-resolution video, due to limitations in transform matrix selection and computational complexity in processes like LFNST and KLT.
Innovation Solution
Implementing a method for video decoding and encoding that involves determining a transform matrix for a current block, modifying it by discarding part of the eigenvectors, and performing inverse or transform processes using a trained transform matrix set based on block shape and intra prediction mode, along with techniques like LFNST, SBT, and SDT to optimize transform processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a full transform matrix with all eigenvectors is used, then transform accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent extracts only the most significant eigenvectors from the full transform matrix, discarding the less significant ones. This is achieved by computing the full transform matrix to obtain eigenvectors and eigenvalues, then selecting only the top k eigenvectors corresponding to the largest eigenvalues, thereby reducing computational complexity while maintaining transform accuracy through energy compaction principles.
Solution Approach 2:
The patent changes the parameter of transform matrix dimensionality by using a reduced set of eigenvectors instead of the complete basis. This parameter change allows the system to achieve lower computational complexity while preserving the essential information through the selected eigenvectors that capture the majority of signal energy.
2Device complexity
If transform matrix is selected based on block shape only, then device complexity is reduced, but coding efficiency decreases
Solution Approach 1:
The patent makes the transform matrix selection dynamic by considering multiple factors including block shape, intra prediction mode, and optionally neighboring block information. This dynamic selection approach allows the system to adapt the transform matrix to the specific characteristics of each block, thereby improving coding efficiency while managing complexity through structured decision-making frameworks.
Solution Approach 2:
The patent segments the transform matrix selection process into multiple decision stages based on different block characteristics. Instead of using a single static selection criterion, the system divides the selection process into steps that consider block shape first, then refine based on prediction mode and other factors, achieving a balance between complexity and efficiency.
3Productivity
If LFNST and KLT processes are applied, then compression performance is improved, but computational complexity increases
Solution Approach 1:
The patent applies partial action by selectively applying LFNST and KLT processes only when beneficial, rather than universally. The system determines whether to apply these complex transforms based on block characteristics, prediction modes, and compression needs, thereby achieving improved compression performance when needed while avoiding unnecessary computational complexity in cases where simpler transforms suffice.
Data Source
AI summary
This disclosure is related to video coding and compression. More specifically, this disclosure relates to methods and apparatus for transform training and coding. A method for video decoding is provided. The method includes: determining, by a decoder, a transform matrix for a current block, the transform matrix including a plurality of eigenvectors; obtaining, by the decoder, a modified transform matrix by discarding part of the plurality of eigenvectors; and performing, by the decoder, an inverse transform process on the current block by using the modified transform matrix.