Block Vector Candidate Clustering for Low-Complexity Video Coding

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing video coding systems fail to efficiently handle redundancy and complexity in block vector selection and reference enlargement, particularly in the context of video decoding devices, leading to inefficiencies in the context of decoder side intra mode derivation and template-based intra mode derivation, and spatial geometric partitioning modes, and spatial geometric partitioning modes, and auto-relocated block vector prediction.

Innovation Solution

The video decoding device is configured to determine a current video block, and to select a reference block, and to select a current video block, and to determine whether to add a second candidate BV to the BV candidate list, and to decode the current video block based on the BV candidate list, and to determine whether to add a third candidate BV to the BV candidate list, and to decode the current video block based on the enlarged reference block.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If multiple candidate BVs are added to the BV candidate list, then the prediction accuracy is improved, but the computational complexity and processing time increase

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

Solution Approach 1:

The patent extracts and removes redundant candidate BVs from the BV candidate list by comparing reference positions and identifying duplicates. This reduces the number of candidates that need to be processed while maintaining the essential prediction options, thereby reducing computational complexity without significantly compromising prediction accuracy.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent applies different processing strategies to different candidates based on their specific characteristics. By evaluating each candidate's reference position against existing candidates and applying selective removal based on local redundancy analysis, the system optimizes the candidate list in a targeted manner rather than applying uniform processing to all candidates.

Inventive Principle:
Principle #3Local quality

2Adaptability or versatility

If the reference block is enlarged to include more candidates, then the prediction options are increased, but the processing overhead and memory requirements increase

Engineering Contradiction:
Improveprediction optionsVSAvoidprocessing overhead
Core Design Contradiction:
Adaptability or versatilityVSQuantity of substance

Solution Approach 1:

The patent extracts redundant candidates from the enlarged reference block by comparing reference positions and identifying duplicates. This removal process reduces the effective number of candidates that need to be stored and processed, thereby reducing memory requirements and processing overhead while maintaining the benefit of having accessed a larger search area.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent merges multiple candidates that point to the same or overlapping reference positions into a single representative candidate. This consolidation reduces the total number of candidates that need to be processed while preserving the prediction information, thereby reducing processing overhead and memory usage.

Inventive Principle:
Principle #5Merging (Combining)

3Productivity

If redundant candidate BVs are removed based on distance threshold, then the processing efficiency is improved, but the risk of removing potentially useful candidates increases

Engineering Contradiction:
Improveprocessing efficiencyVSAvoidcandidate selection reliability
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The patent applies the distance threshold criterion locally to each candidate comparison, evaluating the specific spatial relationship between reference positions. By making decisions based on local geometric relationships rather than global assumptions, the system maintains reliability in candidate selection while achieving processing efficiency through systematic redundancy removal.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent dynamically adjusts the candidate selection process by evaluating candidates in sequence and making adaptive decisions based on the current state of the BV candidate list. This dynamic approach allows the system to maintain reliability by considering the context of already-selected candidates when evaluating subsequent candidates, rather than applying static removal rules.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP4668735A1Clustering of block vector merge candidates
Publication Date: 2025.12.24 INTERDIGITAL CE PATENT HOLDINGS SAS
  • EP4668735A1 patent drawingFigure 1A
  • EP4668735A1 patent drawingFigure 1B
  • EP4668735A1 patent drawingFigure 1C

AI summary

Disclosed herein are systems, methods, and instrumentalities associated with the selection of block vector (BV) candidates. A video coding device (e.g., a video encoder or a decoder) as described herein may be configured to obtain a current video block, and determine a BV candidate list associated with the current video block, wherein the BV candidate list may include at least a first candidate BV that points to a first reference position associated with the current video block. The video coding device may determine whether to add a second candidate BV to the BV candidate list, wherein the second candidate BV may point to a second reference position associated with the current video block and wherein the video coding device may decide not to add the second candidate BV to the BV candidate list if a distance between the first reference position and the second reference position is smaller than a predetermined value. The video coding device may then code the current video block based on the BV candidate list.