Video Block Prediction Using IBC Padding Under Memory Limits

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing video coding technologies face challenges in efficiently managing motion vector prediction, particularly in scenarios where reference picture memory is limited, leading to suboptimal compression efficiency and increased computational requirements.

Innovation Solution

The proposed method involves intra block copy (IBC) mode for video coding, where padded values of a reference block are determined based on a block vector, and samples are reconstructed using these padded values, with specific padding strategies for adjacent blocks to optimize memory usage and enhance compression efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If intra prediction is performed using only neighboring blocks within the same picture, then processing complexity is reduced, but prediction accuracy deteriorates due to lack of temporal information

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

Solution Approach 1:

The patent segments the prediction process into two independent stages: temporal prediction using motion vectors from reference pictures, and spatial prediction using neighboring blocks. This segmentation allows each stage to specialize in its strength while keeping individual processing steps manageable, resolving the contradiction between accuracy and complexity.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent merges temporal prediction and spatial prediction into a unified prediction process where both prediction types are generated and then combined through weighting. This combining approach leverages the complementary strengths of temporal and spatial predictions to achieve higher accuracy without requiring one complex method to replace the other.

Inventive Principle:
Principle #5Merging (Combining)

2Measurement precision

If inter prediction uses reference pictures from multiple frames, then prediction accuracy is improved, but memory requirements and processing time increase

Engineering Contradiction:
Improveprediction accuracyVSAvoidmemory usage
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent applies local quality by differentiating between temporal and spatial prediction based on the characteristics of the current block and available reference data. The weighting process dynamically adjusts the contribution of temporal versus spatial predictions locally for each block, optimizing memory usage by only accessing reference pictures when temporal prediction is beneficial.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent uses partial action by selectively applying inter prediction only when beneficial, rather than universally applying it to all blocks. The weighting mechanism allows the system to use a subset of available reference pictures and prediction methods, avoiding the need to process all possible reference frames and thus reducing memory requirements.

Inventive Principle:
Principle #16Partial or excessive action

3Productivity

If prediction residues are transformed using complex transform kernels, then compression efficiency is improved, but computational complexity increases

Engineering Contradiction:
Improvecompression efficiencyVSAvoidcomputational complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent applies parameter changes by using adaptive quantization parameters that are adjusted based on the frequency characteristics of the prediction residue. Instead of using a fixed complex transform kernel, the system modifies the transform parameters dynamically to match the local signal characteristics, achieving efficient compression with simpler computational operations.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3906686B1Method and apparatus for video coding
Publication Date: 2026.05.06 TENCENT AMERICA LLC
  • EP3906686B1 patent drawingFigure 1
  • EP3906686B1 patent drawingFigure 2
  • EP3906686B1 patent drawingFigure 3

AI summary

Aspects of the disclosure provide method and apparatus for video coding. In the method, prediction information of a current block in a current coding tree unit (CTU) from a coded video bitstream is decoded. The prediction information indicates an intra block copy (IBC) mode. Padded values of a reference block are determined based on a block vector that points to the reference block. The padded values of the reference block are copied from a reference sample line. At least a sample of the current block is reconstructed based on the padded values of the reference block.