Prediction Filter Regularization for IntraTMP and IBC Blocks

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing video coding systems face challenges in efficiently compressing digital video signals, particularly in adapting prediction filters based on prediction modes and block sizes, leading to suboptimal encoding and decoding performance.

Innovation Solution

Implementing regularization in prediction filters by generating matrices based on prediction modes and block sizes, and adaptively employing regularization based on template cost comparisons to optimize filtering parameters.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If prediction filters are adapted based on prediction modes and block sizes, then encoding and decoding performance is improved, but device complexity increases

Engineering Contradiction:
Improveencoding and decoding performanceVSAvoidfilter adaptation complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent applies different filtering operations and regularization matrices to different prediction modes (IntraTMP vs. IBC) and different block size ranges. This local differentiation optimizes performance for each specific case without requiring a completely complex adaptive system for all scenarios, resolving the contradiction by making the system complexity proportional to the actual performance needs of each prediction case.

Inventive Principle:
Principle #3Local quality

2Measurement precision

If regularization matrices are generated based on prediction modes and block sizes, then filtering precision is improved, but computational resources increase

Engineering Contradiction:
Improvefiltering precisionVSAvoidcomputational resources
Core Design Contradiction:
Measurement precisionVSPower

Solution Approach 1:

The patent changes the regularization parameter based on prediction mode and block size characteristics. By selecting from predefined regularization matrices or adjusting the regularization strength parameter according to these conditions, the system achieves higher filtering precision when needed while avoiding unnecessary computational overhead in cases where strong regularization is not required, thus resolving the contradiction between precision and computational resources.

Inventive Principle:
Principle #35Parameter changes

3Reliability

If adaptive regularization is employed based on template cost comparisons, then prediction accuracy is improved, but processing time increases

Engineering Contradiction:
Improveprediction accuracyVSAvoidprocessing time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent implements adaptive regularization by comparing template costs, but applies the full regularization operation only when the comparison indicates it is beneficial. In cases where the template cost comparison shows that regularization would not provide significant improvement, the system skips or reduces the regularization step. This partial application of regularization maintains prediction accuracy where needed while reducing processing time in cases where it would be wasteful, resolving the contradiction between accuracy and processing time.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP4676050A1Regularization for prediction filters
Publication Date: 2026.01.07 INTERDIGITAL CE PATENT HOLDINGS SAS
  • EP4676050A1 patent drawingFigure 1A
  • EP4676050A1 patent drawingFigure 1B
  • EP4676050A1 patent drawingFigure 1C

AI summary

Systems, methods, and instrumentalities are disclosed for performing regularization in prediction filters. A device, which may be an encoder device and/or a decoder device, may determine for a picture block that a prediction mode is one of IntraTMP or IBC. The device may determine, using the prediction mode, a prediction block associated with the picture block. The device may generate, based on a current template and a reference template, a first matrix associated with pixel values in the reference template and/or a second matrix associated with vectors associated with the current template. The device may generate a regularization matrix. The device may be configured to determine, using the at least one matrix associated with the filtering function and the regularization matrix, filtering parameters. The device may filter the prediction block using the determined filtering parameters and may decode and/or encode the picture block using the filtered prediction block.