Matrix-Based Intra Prediction for Lower-Complexity Transform Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face challenges in efficiently handling high-resolution video data due to increasing bandwidth demands and the need for improved coding efficiency, particularly in standards like HEVC and VVC, where intra prediction methods can be optimized for better performance.
Innovation Solution
Implementing matrix-based intra prediction (MIP) modes that utilize boundary downsampling, matrix vector multiplication, and optional upsampling operations to generate a most probable mode (MPM) list, allowing for more efficient video coding by reducing computational complexity and memory requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If matrix-based intra prediction mode is used, then coding efficiency is improved, but device complexity increases due to additional operations
Solution Approach 1:
The prediction process is segmented into distinct operational stages: boundary downsampling, matrix vector multiplication, and optional upsampling. Each stage processes specific portions of the video block boundary, allowing the complex prediction to be broken down into manageable steps that can be implemented efficiently in hardware or software.
Solution Approach 2:
Boundary downsampling is performed as a preliminary action before the matrix vector multiplication. By pre-processing the boundary samples to reduce their quantity, the subsequent matrix operation operates on smaller data, reducing computational complexity while maintaining prediction accuracy.
2Power
If boundary downsampling is applied, then computational load is reduced, but prediction precision may be affected
Solution Approach 1:
The boundary downsampling operation changes the parameter of boundary sample quantity from full resolution to reduced resolution. By carefully designing the downsampling process to retain essential boundary information and using appropriate upsampling interpolation methods when needed, the system maintains prediction precision while reducing computational load.
Solution Approach 2:
The boundary downsampling creates a reduced representation (copy) of the boundary samples that captures the essential prediction information. This compressed boundary representation is then used in the matrix vector multiplication, achieving lower computational complexity while preserving the necessary predictive accuracy through the copying of key boundary characteristics.
3Ease of operation
If MPM list is generated using simplified procedure, then ease of operation is improved, but coding accuracy may be compromised
Solution Approach 1:
The MPM list generation procedure is designed to be universal and multi-functional. The same simplified procedure used for non-MIP modes can also be applied to MIP modes, allowing the system to generate MPM lists efficiently without requiring separate complex procedures. This unified approach maintains coding accuracy while improving ease of operation through procedure reuse.
Solution Approach 2:
The MPM list generation process is self-service in that it automatically adapts to the current video block's prediction mode. When MIP mode is detected, the system automatically adjusts the MPM list generation to incorporate the appropriate boundary samples and matrix parameters, eliminating the need for manual intervention or complex conditional logic while maintaining coding accuracy.
Data Source
AI summary
Devices, systems and methods for digital video coding, which includes matrix-based intra prediction methods for video coding, are described. In a representative aspect, a method for video processing includes performing a conversion between a current video block of a video and a bitstream representation of the current video block according to a rule, where the rule specifies a relationship between applicability of a matrix based intra prediction (MIP) mode or a transform mode during the conversion, where the MIP mode includes determining a prediction block of the current video block by performing, on previously coded samples of the video, a boundary downsampling operation, followed by a matrix vector multiplication operation, and selectively followed by an upsampling operation, and where the transform mode specifies use of a transform operation for the determining the prediction block for the current video block.