Interpolated Matrix Intra Prediction for Low-Storage Video Coding

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The existing video compression technologies, such as the Versatile Video Coding (WC) standard, face limitations in utilizing the Matrix Intra Prediction (MIP2) mode due to excessive storage requirements, restricting its widespread application and potential performance improvement.

Innovation Solution

Introduce a new prediction mode, MIP_mix, which interpolates between two adjacent MIP2 modes and combines them with a regular angular prediction, allowing for a weighted sum to generate a mixed prediction, thereby expanding the usage of MIP2 modes without increasing complexity.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If MIP2 mode is used for intra prediction, then coding performance is improved, but storage requirements increase excessively

Engineering Contradiction:
Improvecoding performanceVSAvoidstorage requirements
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent segments the MIP2 mode application by dividing the prediction process into two parts: using regular intra prediction modes for blocks that don't require MIP2, and selectively applying MIP2 mode only for specific block types or configurations. This segmentation allows the system to achieve coding performance improvement where needed while avoiding excessive storage requirements for blocks that don't benefit from MIP2.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent changes the parameter of mode selection by introducing a flag or condition that determines whether to use regular intra prediction mode or MIP2 mode for each block. This parameter change allows dynamic adaptation based on block characteristics, achieving better coding performance for suitable blocks while maintaining lower storage requirements for others.

Inventive Principle:
Principle #35Parameter changes

2Measurement precision

If more MIP2 modes are implemented, then prediction accuracy is improved, but device complexity increases

Engineering Contradiction:
Improveprediction accuracyVSAvoidoperational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent applies partial action by implementing only a subset of MIP2 modes rather than all possible modes. The system selectively enables MIP2 mode for specific block types, sizes, or configurations where it provides the most benefit, rather than universally applying all MIP2 modes. This approach improves prediction accuracy for relevant cases while keeping device complexity manageable.

Inventive Principle:
Principle #16Partial or excessive action

Solution Approach 2:

The patent introduces dynamic mode selection where the prediction mode (regular intra or MIP2) is determined based on block characteristics such as size, shape, or content properties. This dynamic approach allows the system to adaptively choose the appropriate mode for each block, improving overall prediction accuracy without requiring all MIP2 modes to be active simultaneously, thus controlling device complexity.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP4676033A1Method and apparatus for encoding/decoding with an interpolated matrix based intra prediction mode
Publication Date: 2026.01.07 INTERDIGITAL CE PATENT HOLDINGS SAS
  • EP4676033A1 patent drawingFigure 1
  • EP4676033A1 patent drawingFigure 2
  • EP4676033A1 patent drawingFigure 3

AI summary

In various implementations, method and devices are disclosed that encode or decode a block of a picture with an interpolated matrix based intra prediction mode. For instance, the method comprises determining that a block is coded using a regular intra prediction mode, a regular intra prediction mode being an angular directional mode; responsive to determining that the regular intra prediction mode used to code the block belongs to a list of intra prediction mode candidates comprising at least two matrix-based intra prediction modes, obtaining a predicted block from an interpolation of a first intra matrix based intra prediction mode and of a second matrix based intra prediction mode.