Optical Flow Prediction Refinement for Unified BDOF and PROF
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies, such as VVC, face inefficiencies in motion compensation due to non-unified designs between bi-directional optical flow (BDOF) and prediction refinement with optical flow (PROF) for affine mode, leading to challenges in hardware implementation and precision mismatch in gradient calculations.
Innovation Solution
Harmonize the designs of BDOF and PROF by aligning precision of motion vector differences and gradients, and integrating them into a unified pipeline for efficient hardware implementation, ensuring consistent bit-depth handling and shared processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If separate designs are used for BDOF and PROF, then flexibility in handling different video modes is improved, but device complexity and hardware implementation difficulty increase
Solution Approach 1:
The patent merges BDOF and PROF into a unified design where both functions share common computational components including gradient calculation units, motion vector difference processing, and optical flow computation. This consolidation reduces hardware complexity while maintaining the ability to handle both affine and non-affine video modes through a single integrated processing pipeline.
Solution Approach 2:
The unified BDOF/PROF design implements a multi-functional processing unit that can operate in different modes depending on the video block type. The same hardware structure performs both BDOF for non-affine blocks and PROF for affine blocks, achieving versatility without requiring separate dedicated hardware for each function.
2Productivity
If different precision levels are used for BDOF and PROF gradient calculations, then computational efficiency is improved, but manufacturing precision and calculation accuracy deteriorate
Solution Approach 1:
The patent applies different precision levels to different stages and components within the gradient calculation process. Early stages use lower precision for initial computations while critical stages such as final motion refinement use higher precision, optimizing the balance between computational efficiency and calculation accuracy for each specific operation.
Solution Approach 2:
The unified design dynamically adjusts precision parameters based on the operating mode (BDOF or PROF) and the specific computation stage. Bit-depth and precision settings are changed adaptively according to the requirements of each function, ensuring sufficient accuracy while maintaining computational efficiency across different video coding scenarios.
Data Source
AI summary
A method of video coding for PROF: first SPS level control flag and second SPS level control flag are determined. When the first SPS level control flag indicates BDOF is enabled for the current video sequence, BDOF for a current video block is not applied when the current video block is encoded in affine mode; or when the first SPS level control flag indicates BDOF is enabled for the current video sequence, BDOF is applied to derive motion refinements of the current video block based on first prediction samples and second prediction samples when the current video block is not encoded in affine mode. When the second SPS level control flag indicates PROF is enabled for the current video sequence, PROF is applied to derive motion refinements of the video block based on first prediction samples and second prediction samples when the video block is encoded in affine mode.