Matrix Intra Prediction Mode Selection for Efficient Image Coding

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The increasing demand for high-resolution, high-quality image/video data, particularly in applications like virtual reality and augmented reality, necessitates a more efficient compression technology to reduce transmission and storage costs while maintaining visual quality.

Innovation Solution

Adaptive selection of matrix-based intra prediction (MIP) methods based on block type, allowing for efficient intra prediction by selectively applying MIP processes and reducing the number of modes for blocks with low prediction efficiency, thereby enhancing overall image/video compression efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If MIP process is applied to all block types, then prediction accuracy is improved, but transmission data and processing complexity increase

Engineering Contradiction:
Improveprediction accuracyVSAvoidtransmission data
Core Design Contradiction:
Measurement precisionVSLoss of information

Solution Approach 1:

The patent applies different MIP mode counts to different block types based on their prediction efficiency characteristics. Specifically, it uses a first MIP mode count for blocks with high prediction efficiency and a second MIP mode count for blocks with low prediction efficiency, rather than uniformly applying the same number of modes to all blocks. This local differentiation optimizes the balance between prediction accuracy and transmission data.

Inventive Principle:
Principle #3Local quality

2Measurement precision

If MIP process is applied to all block types, then prediction accuracy is improved, but device complexity increases

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

Solution Approach 1:

The patent implements local quality by configuring different MIP mode counts for different block types. The decoding apparatus determines the appropriate MIP mode count based on the block type, applying a first count for high-efficiency blocks and a second count for low-efficiency blocks. This approach reduces overall processing complexity while maintaining prediction accuracy where it matters most.

Inventive Principle:
Principle #3Local quality

3Loss of information

If MIP modes are reduced for low efficiency blocks, then transmission data is reduced, but prediction accuracy may deteriorate

Engineering Contradiction:
Improvetransmission dataVSAvoidprediction accuracy
Core Design Contradiction:
Loss of informationVSMeasurement precision

Solution Approach 1:

The patent changes the parameter of MIP mode count based on block type characteristics. By identifying which blocks have low prediction efficiency and applying a reduced MIP mode count specifically to those blocks, the system optimizes transmission data without significantly compromising overall prediction accuracy, since the reduced modes are applied only where they would be least effective.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentUS20260012579A1Matrix intra prediction-based image coding apparatus and method
Publication Date: 2026.01.08 LG ELECTRONICS INC
  • US20260012579A1 patent drawing
  • US20260012579A1 patent drawing
  • US20260012579A1 patent drawing

AI summary

An image decoding method includes obtaining image information including prediction mode information from a bitstream, obtaining matrix-based intra prediction (MIP) flag information related to whether an MIP is applied to a current block based on the prediction mode information, deriving an MIP mode for the current block based on the MIP flag information, deriving an MIP matrix for the current block based on the MIP mode, generating MIP samples for the current block based on the MIP matrix, deriving reconstructed samples for the current block based on the MIP samples, wherein for the current block of which size is equal to 64×4, the MIP flag information for the current block is obtained from the bitstream, and wherein the MIP matrix is derived further based on a width and a height of the current block.