Precision-Aligned Optical-Flow Prediction Refinement for Affine Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies, such as VVC, face inefficiencies in motion compensation due to limitations in block-based motion refinement methods, particularly with bi-directional optical flow (BDOF) and prediction refinement with optical flow (PROF) for affine mode, leading to suboptimal coding efficiency and hardware implementation challenges.
Innovation Solution
Harmonizing the designs of BDOF and PROF by aligning the precision of motion vector differences and gradient calculations to match the intermediate prediction sample precision, and integrating them into a unified pipeline for improved hardware compatibility and coding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If block-based motion refinement methods (BDOF and PROF) are used in existing video coding standards, then motion compensation is performed, but coding efficiency is suboptimal and hardware implementation is complex
Solution Approach 1:
The patent merges BDOF and PROF into a unified pipeline by harmonizing their precision requirements. Motion vector differences and gradient calculations are aligned to match intermediate prediction sample precision, allowing both methods to operate within a single coherent hardware architecture rather than requiring separate processing paths.
Solution Approach 2:
The unified pipeline design enables a single hardware structure to perform multiple functions - both BDOF and PROF operations can be executed using the same precision-aligned computational units, reducing overall hardware complexity while maintaining support for both motion refinement methods.
2Adaptability or versatility
If BDOF and PROF use different precision for motion vector differences and gradient calculations, then each method can be optimized independently, but integration into a unified pipeline is difficult
Solution Approach 1:
The patent changes the precision parameters of motion vector differences and gradient calculations to align with intermediate prediction sample precision. This parameter harmonization allows BDOF and PROF to be integrated into a unified pipeline without requiring separate precision management, reducing integration complexity while preserving method-specific optimizations.
Data Source
AI summary
Methods, devices, and non-transitory computer-readable storage mediums are provided. The method may include an encoder obtaining a video block that is coded based on an affine mode, obtaining a first reference picture and a second reference picture associated with the video block, obtaining first and second horizontal and vertical gradient values based on first prediction samples and second prediction samples, obtaining first and second horizontal and vertical motion refinements based on control point motion vectors (CPMVs), obtaining first and second prediction refinements based on the first and second horizontal and vertical gradient values, and the first and second horizontal and vertical motion refinements, obtaining first and second refined samples based on the first prediction samples, the second prediction samples, and the first and second prediction refinements, and obtaining final prediction samples of the video block based on the first and second refined samples and prediction parameters.