Linear Intra Prediction for High-Resolution Image Encoding

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image compression technologies face challenges in efficiently predicting pixel values within a current picture, particularly in high-resolution and high-quality images, necessitating improved intra prediction methods and apparatus for enhanced compression efficiency.

Innovation Solution

An image decoding method and apparatus that derive a prediction sample based on a linear model parameter using neighboring reference samples, allowing for adaptive selection of linear model-based prediction modes and transform kernels to optimize intra prediction performance.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If conventional intra prediction methods are used, then device complexity is reduced, but manufacturing precision (prediction accuracy) deteriorates for high-resolution images

Engineering Contradiction:
Improveprediction accuracyVSAvoiddevice complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent applies parameter changes by transitioning from conventional fixed prediction methods to adaptive linear model parameters. The encoder and decoder derive linear model parameters (such as gradient values and offset values) from reference samples, allowing the prediction model to adapt to different image content characteristics. This enables higher prediction accuracy for high-resolution images while maintaining manageable device complexity through standardized parameter derivation processes.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent implements dynamics by making the prediction model adaptive rather than static. The linear model parameters are dynamically derived based on the actual reference samples in each coding block, allowing the prediction to respond to local image characteristics. This dynamic adaptation improves prediction precision without requiring overly complex device architecture, as the adaptivity is achieved through algorithmic parameter derivation.

Inventive Principle:
Principle #15Dynamics

2Manufacturing precision

If linear model parameters are derived from multiple reference samples, then prediction accuracy improves, but loss of information increases due to additional signaling

Engineering Contradiction:
Improveprediction accuracyVSAvoidsignaling overhead
Core Design Contradiction:
Manufacturing precisionVSLoss of information

Solution Approach 1:

The patent applies the extraction principle by selectively deriving only the necessary linear model parameters from reference samples that are already available in the decoding process. Instead of transmitting additional reference sample data, the method extracts the essential prediction parameters (gradient and offset) directly from the reconstructed reference samples at the decoder side. This reduces signaling overhead while maintaining prediction accuracy.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent implements self-service by enabling the decoder to autonomously derive the linear model parameters from the reference samples without requiring explicit transmission of these parameters from the encoder. The decoder uses the same reference samples and derivation process as the encoder, allowing it to self-generate the prediction parameters and eliminating the need for additional signaling bits.

Inventive Principle:
Principle #25Self-service

3Manufacturing precision

If adaptive linear model parameters are used, then prediction accuracy improves, but processing time increases

Engineering Contradiction:
Improveprediction accuracyVSAvoidprocessing time
Core Design Contradiction:
Manufacturing precisionVSLoss of time

Solution Approach 1:

The patent applies preliminary action by preparing and storing the reference samples during the reconstruction process before they are needed for linear model parameter derivation. The reference samples are made available in advance, allowing the linear model parameters to be derived efficiently without requiring additional data fetching or complex computations during the prediction stage. This preprocessing reduces processing time while maintaining prediction accuracy.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP4679832A1Image encoding/decoding method and device, and recording medium storing bitstream
Publication Date: 2026.01.14 LG ELECTRONICS INC
  • EP4679832A1 patent drawingFigure 1
  • EP4679832A1 patent drawingFigure 2
  • EP4679832A1 patent drawingFigure 3

AI summary

An image decoding/encoding method and device according to the present disclosure can derive prediction samples of the current block on the basis of linear model parameters of the current block. Here, the linear model parameters may be derived on the basis of one or more reference samples in a neighboring region of the current block. The prediction samples of the current block may be derived by applying the linear model parameters to the coordinates of the prediction samples in the current block and/or reconstructed samples near the current block.