TIMD Merge Mode Intra-Prediction Noise Filtering

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The existing video picture encoding and decoding technologies face challenges in optimizing the derivation of Template-based Intra Mode Derivation (TIMD) modes for predicting video picture blocks, particularly due to noise in block boundary samples.

Innovation Solution

A method is introduced to determine a TIMD merge mode by obtaining a merge list of TIMD merge mode candidates, each comprising internal prediction modes and associated information. The method selects the TIMD merge mode based on prediction costs, using Rate-Distortion optimization, and signals the selected index in the bitstream.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If TIMD mode is derived using template samples, then intra-prediction capability is provided, but prediction accuracy deteriorates due to noise in block boundary samples

Engineering Contradiction:
Improveintra-prediction capabilityVSAvoidprediction accuracy
Core Design Contradiction:
Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent extracts and removes the harmful noise components from template samples by identifying and excluding block boundary samples that contain quantization noise. This allows the system to retain the useful prediction information from template samples while eliminating the noise that degrades prediction accuracy.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent applies different processing to different regions of the template samples. Interior samples are used for prediction while block boundary samples are either excluded or processed differently to remove noise. This local differentiation allows the system to maintain prediction capability in clean regions while avoiding noise-affected regions.

Inventive Principle:
Principle #3Local quality

2Measurement precision

If multiple TIMD merge mode candidates are evaluated, 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 evaluates multiple TIMD merge mode candidates (more than the minimum required) to ensure high prediction accuracy, accepting the increased computational complexity as a trade-off for significantly improved prediction quality. This approach prioritizes accuracy over computational efficiency.

Inventive Principle:
Principle #16Partial or excessive action

Solution Approach 2:

The system uses rate-distortion optimization feedback to evaluate and select the best TIMD merge mode candidates. By measuring the actual prediction quality and using this feedback to guide candidate selection, the system efficiently identifies optimal modes while managing computational resources.

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP4561061A1Encoding/decoding video picture data
Publication Date: 2025.05.28 BEIJING XIAOMI MOBILE SOFTWARE CO LTD
  • EP4561061A1 patent drawingFigure 1~2
  • EP4561061A1 patent drawingFigure 3~4
  • EP4561061A1 patent drawingFigure 5

AI summary

The present disclosure relates a TIMD merge mode comprising at least one intra mode derivation candidate, i.e. at least one DIMD and/or TIMD mode, and associated information for intra-predicting a current block of a video picture VP. The at least one DIMD and/or TIMD mode candidates is collected from at least one previously (de)coded block which is predicted using either TIMD mode or DIMD mode or TIMD merge mode or DIMD merge mode. A merge list of TIMD merge mode candidates may be obtained (1510). Each TIMD merge mode candidate may contain at least one internal prediction mode of a DIMD or TIMD mode and their associated information. Each TIMD merge mode candidate is used for intra-prediction. The TIMD merge mode for intra-predicting the current block is determined (1520) by selecting one TIMD merge mode candidate of the merge list based on predicting costs obtained the TIMD merge mode candidates of the merge list.