Matrix-Based Intra Prediction with Boundary Downsampling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face challenges in efficiently compressing and decompressing digital video due to increasing bandwidth demands and the need for improved coding efficiency, particularly in higher resolution formats.
Innovation Solution
Implementing matrix-based intra prediction methods, including affine linear weighted intra prediction (ALWIP) and matrix vector multiplication operations, to enhance video coding efficiency by reducing computational complexity and improving compression performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If traditional intra prediction methods are used, then the coding process is simpler, but the compression performance is insufficient for high resolution video
Solution Approach 1:
The patent segments the intra prediction process into distinct stages: boundary sample extraction, downsampling to create a reduced set of boundary samples, matrix-based prediction calculation, and selective upsampling. This segmentation allows each stage to be optimized independently, improving compression performance while managing computational complexity through efficient processing at each segment.
Solution Approach 2:
The patent changes the parameters of the prediction process by introducing matrix-based calculations with configurable dimensions, adjustable boundary sample counts, and flexible upsampling decisions based on block size and mode. These parameter changes enable adaptive optimization between compression performance and computational complexity for different video resolutions and content types.
2Productivity
If matrix-based intra prediction with boundary downsampling is implemented, then compression efficiency improves, but processing complexity increases
Solution Approach 1:
The patent performs preliminary downsampling of boundary samples before the main prediction calculation, creating a reduced set of boundary samples that captures essential information while reducing the computational burden of subsequent matrix operations. This preliminary action prepares the data in an optimized state for efficient processing.
Solution Approach 2:
The patent applies partial upsampling only when necessary based on block size and prediction mode, rather than always upsampling the full prediction block. This selective approach achieves sufficient reconstruction quality for most cases while avoiding unnecessary computational complexity in scenarios where full resolution reconstruction is not required.
3Ease of manufacture
If uniform MPM list generation procedure is used for both MIP and non-MIP modes, then implementation is simpler, but coding efficiency is reduced
Solution Approach 1:
The patent creates a universal MPM list generation procedure that serves both MIP and non-MIP modes, allowing the same code path and data structures to handle different prediction modes. This universal approach maintains implementation simplicity while achieving good coding efficiency through mode-adaptive selection from the generated MPM list rather than mode-specific procedures.
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 using a matrix based intra prediction (MIP) mode in which a prediction block of the current video block is determined 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, where the performing the conversion includes deriving, according to a rule, boundary samples by applying a left bit shift operation or a right bit shift operation on a sum of at least one reference boundary sample, and where the rule determines whether to apply the left bit shift or the right bit shift operation.