Matrix-Based Intra Prediction with Binary Coefficients

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing video coding technologies face challenges in achieving high compression ratios with minimal sacrifice in picture quality, particularly in the context of limited network resources and increasing video quality demands.

Innovation Solution

Implement matrix-based intra prediction (MIP) using restricted sets of MIP coefficients with binary representation, allowing for a simplified implementation by replacing multiplication operations with left shifts and additions, thereby reducing computational complexity.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If matrix-based intra prediction with full-precision coefficients is used, then prediction accuracy is improved, but computational complexity increases

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

Solution Approach 1:

The patent changes the parameter representation of MIP coefficients from full-precision fixed-point values to restricted binary forms (powers of 2). This allows the encoder to select from a limited set of coefficient values that can be represented efficiently in binary, reducing the computational burden of matrix-vector multiplication while maintaining acceptable prediction accuracy through adaptive selection.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent employs inexpensive approximation by replacing expensive multiplication operations with cheaper left-shift operations. The restricted binary coefficient set enables this substitution, as multiplying by powers of 2 is equivalent to bit-shifting, which is computationally trivial compared to general multiplication. This creates a computationally lightweight prediction method suitable for real-time encoding.

Inventive Principle:
Principle #27Cheap short-living objects (Disposable)

2Productivity

If conventional intra prediction methods are used, then implementation is simple, but compression ratio is insufficient

Engineering Contradiction:
Improvecompression ratioVSAvoidimplementation complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent segments the MIP coefficient space into discrete, restricted levels (powers of 2). Instead of using continuous full-precision coefficients, the method divides the coefficient domain into manageable discrete values, making the implementation tractable while preserving the essential functionality of MIP for improving compression efficiency.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces an intermediary representation layer between the original full-precision coefficients and the actual computation. By mapping coefficients to a restricted binary set, this intermediary representation simplifies the computational operations required for MIP, bridging the gap between theoretical prediction accuracy and practical implementability.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP3970365B1Matrix-based intra prediction for still picture and video coding
Publication Date: 2025.08.06 HUAWEI TECH CO LTD
  • EP3970365B1 patent drawingFigure 1A
  • EP3970365B1 patent drawingFigure 1B
  • EP3970365B1 patent drawingFigure 2

AI summary

The present invention relates to a method and apparatus for intra prediction of a block using a matrix-based intra prediction, MIP. In accordance with the present invention, possible values of MIP coefficients are restricted to those which can be represented by a binary representation with a limited number of non-zero bits. This may enable reducing the number of computationally expensive multiplication operations for MIP.