Optical Flow Refinement Video Coding Grid Structure
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding technologies face challenges in efficiently compressing video data, particularly in reducing redundancy and managing motion vectors, which affects bandwidth and storage requirements, and existing methods for motion vector prediction are limited in accuracy and complexity.
Innovation Solution
The implementation of optical flow refinement coding tools, such as bi-directional optical flow (BDOF) and affine prediction refinement with optical flow (PROF), which use a grid structure for motion compensation and gradient calculations to refine predictions and reduce deblocking filter strength, enabling more accurate motion vector refinement and prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If motion vector prediction is used to reduce data requirements, then compression efficiency is improved, but prediction accuracy deteriorates due to rounding errors and limited predictor options
Solution Approach 1:
The current block is divided into multiple sub-blocks, and motion vector prediction is performed separately for each sub-block. This segmentation allows for more localized and accurate motion compensation, reducing the impact of rounding errors and improving overall prediction accuracy while maintaining compression efficiency.
Solution Approach 2:
Gradient calculations are introduced as an intermediary mechanism to refine motion vector prediction. By computing gradients of prediction signals and using them to adjust motion vectors, the system achieves higher prediction accuracy without directly increasing the amount of transmitted data, thus resolving the contradiction between compression efficiency and accuracy.
2Measurement precision
If optical flow refinement with gradient calculations is implemented, then motion prediction accuracy is improved, but computational complexity increases
Solution Approach 1:
Gradient calculations and optical flow refinement are applied selectively to different sub-blocks based on their specific characteristics. By focusing computational resources on areas where they provide the most benefit and using simplified models where appropriate, the system improves prediction accuracy while controlling overall computational complexity.
Solution Approach 2:
The patent applies optical flow refinement and gradient calculations partially - only to certain sub-blocks and only when beneficial - rather than uniformly across the entire block. This partial application reduces computational complexity while still achieving improved prediction accuracy in critical areas.
3Measurement precision
If deblocking filter strength is reduced for internal edges, then prediction accuracy is improved, but block artifacts may increase
Solution Approach 1:
The deblocking filter strength is made dynamic and adaptive rather than fixed. The filter automatically adjusts its strength based on local image characteristics, block boundaries, and prediction accuracy requirements. This dynamic adjustment allows the system to maintain high prediction accuracy while minimizing block artifacts through context-dependent filtering.
Solution Approach 2:
The patent changes the parameter of deblocking filter strength from a fixed value to a variable that is adjusted based on local conditions. By modifying this parameter dynamically according to the specific characteristics of each block and its boundaries, the system achieves both high prediction accuracy and artifact suppression.
Data Source
AI summary
Aspects of the disclosure provide methods and apparatuses for video encoding/decoding. In some examples, an apparatus for video decoding includes receiving circuitry and processing circuitry. For example, the processing circuitry decodes prediction information of a current block in a current picture from a coded video bitstream. The prediction information is indicative of a prediction mode that uses a first optical flow refinement coding tool. The first optical flow refinement coding tool uses a grid structure that is used additionally by at least a second optical flow refinement coding tool during operations of optical flow refinement. Then, the processing circuitry calculates, using the first optical flow refinement coding tool, refinements to predictions of samples of the current block based on the grid structure, and reconstructs the samples of the current block based on the refinements and the predictions of the samples.