Resampled L-Shaped Templates for Block-Wise Intra-Prediction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing video coding technologies face inefficiencies in intra-prediction modes, leading to high side information rate consumption and suboptimal prediction residuals, particularly in block-wise picture coding.

Innovation Solution

The implementation of L-shaped templates and neural networks for resampling intra-prediction signals, allowing for more efficient compression by determining intra-prediction modes based on neighboring samples and transforming prediction residuals.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If the number of intra-prediction modes is increased to improve prediction signal quality, then the prediction residual decreases, but the side information rate consumption increases

Engineering Contradiction:
Improveprediction signal qualityVSAvoidside information rate
Core Design Contradiction:
Measurement precisionVSLoss of information

Solution Approach 1:

The patent segments the prediction process by dividing blocks into different size categories (predetermined size vs. other sizes). For predetermined size blocks, the full L-shaped template is used, while for other sizes, the template is resampled. This segmentation allows the system to maintain high prediction quality for standard blocks while reducing complexity and side information for non-standard blocks.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent changes the parameter of template size by introducing resampling. Instead of using fixed-size templates for all blocks, the template size is adaptively adjusted through resampling operations (interpolation or decimation) based on the current block size, allowing the system to maintain prediction quality across varying block dimensions without increasing the number of prediction modes.

Inventive Principle:
Principle #35Parameter changes

2Measurement precision

If resampling operations are applied to adapt templates to different block sizes, then prediction accuracy is maintained, but computational complexity increases

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

Solution Approach 1:

The patent applies partial resampling by only processing the L-shaped template portion that is necessary for prediction, rather than resampling the entire block. The resampling is performed selectively on the neighboring samples that form the prediction template, reducing unnecessary computational operations while maintaining prediction accuracy.

Inventive Principle:
Principle #16Partial or excessive action

Solution Approach 2:

The patent uses copying by creating a resampled version of the L-shaped template from the original neighboring samples. Instead of developing entirely new prediction methods for different block sizes, the system copies and adapts the existing template structure through resampling operations, leveraging the original template design while adjusting it to fit different block dimensions.

Inventive Principle:
Principle #26Copying

Data Source

PatentEP3777141B1Intra-prediction mode concept for block-wise picture coding
Publication Date: 2025.09.17 FRAUNHOFER GESELLSCHAFT ZUR FORDERUNG DER ANGEWANDTEN FORSCHUNG EV
  • EP3777141B1 patent drawingFigure 1~2
  • EP3777141B1 patent drawingFigure 3~4
  • EP3777141B1 patent drawingFigure 5

AI summary

There are disclosed intra-prediction mode concepts for block-wise picture coding. In particular, there is disclosed an apparatus (14-1, 54-2) for block-wise decoding a picture (10) from a data stream (12) and/or encoding a picture (10) into a data stream (12), the apparatus supporting at least one intra-prediction mode according to which the intra-prediction signal for a block (136, 172) of a predetermined size of the picture is determined by applying a first template (130, 170) of samples which neighbours the current block onto a neural network (80). The apparatus may be configured, for a current block (18) differing from the predetermined size, to: resample (134, 166) a second template (60) of samples neighboring the current block (18), so as to conform with the first template (130, 170) so as to obtain a resampled template (130, 170); apply (138a, 170a, 44-1, 44-2) the resampled template (130, 170) of samples onto the neural network (80) so as to obtain a preliminary intra-prediction signal (138, 172, 176); and resample (140, 180) the preliminary intra-prediction signal (138, 172, 176) so as to conform with the current block (18) so as to obtain (140) the intra-prediction signal (142, 24-1, 24-2) for the current block (18).