Bi-prediction Weighting for Video Encoding Precision
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The bi-prediction of inter coded blocks in video compression schemes faces challenges such as high computational complexity and suboptimal motion vector propagation, particularly with triangular partitions, which affect compression efficiency and precision.
Innovation Solution
A method for weighted bi-prediction is introduced, where weights depend on the sample position within the block, allowing for improved precision and efficiency by using position-dependent weights for averaging uni-directional predictors, and optimizing motion vector storage and signaling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional bi-prediction with fixed weights is used, then implementation is simpler, but prediction precision is suboptimal
Solution Approach 1:
The patent applies local quality by using position-dependent weights in bi-prediction, where different weights are applied to different spatial locations within a prediction block. This allows the prediction to adapt locally to varying motion characteristics across the block, improving precision without requiring completely different prediction methods for each region.
Solution Approach 2:
The patent introduces dynamic weighting where the weights for combining reference block predictions vary based on the position of samples within the current block. This dynamic approach allows the prediction to adapt to local motion variations, improving accuracy while maintaining a relatively simple implementation through systematic weight calculation formulas.
2Reliability
If triangular partitions are used, then motion vector propagation is improved, but computational complexity increases
Solution Approach 1:
The patent uses triangular partitioning to divide prediction blocks into smaller triangular regions, allowing for more accurate motion vector propagation within each triangle. This segmentation improves reliability by better capturing local motion patterns while the systematic approach to handling the partitions keeps computational complexity manageable.
3Measurement precision
If position-dependent weights are used, then prediction accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent changes the parameters of bi-prediction by introducing position-dependent weights that vary systematically across the prediction block. This parameter change improves prediction accuracy by adapting to local variations, while the use of systematic weight calculation formulas maintains computational efficiency through predictable and repeatable operations.
Data Source
AI summary
Different implementations are described, particularly implementations for video encoding and decoding using motion compensation with bi-prediction are presented. The encoding method comprises for a picture, obtaining a first predictor for a block of the picture using a first reference picture; obtaining a second predictor for said block of the picture using a second reference picture; using the first predictor and the second predictor for forming a third predictor for the block in bi-prediction inter prediction, wherein the third predictor is obtained as a weighted average of the first predictor and the second predictor; and wherein a weight used in the weighed prediction depend on the position of the sample in the block. Ohers embodiments are presented for implementing block triangle partition prediction, for implementing block partition prediction using multiple patterns and for corresponding motion compensation in decoding method.