Motion Vector Difference Range Constraints in Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding standards, such as VVC, face challenges in accurately representing motion vector differences (MVD) due to bitstream constraints that may result in inaccurate MVD values, especially when non-quarter-pel MVD resolutions are used, leading to potential inaccuracies in motion vector reconstruction.
Innovation Solution
The proposed solution involves determining a range of MVD values based on the maximum allowed motion vector resolution, precision, or properties of the video region, and constraining these values within specific ranges to ensure accurate representation and reconstruction, such as the range [−2M, 2M−1] where M=17, or adapting the range to allowable MVD precision and motion vector precision of a codec.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If bitstream constraints are applied to MVD values, then device complexity is reduced, but measurement precision of MVD deteriorates
Solution Approach 1:
The patent changes the parameter representation by introducing scaled MVD values (e.g., multiplying by 4 or 16) to align with quarter-pel or 1/16-pel precision grids. This parameter transformation allows the MVD to satisfy bitstream constraints while maintaining accurate motion representation at sub-integer precision levels.
2Reliability
If MVD values are constrained to fixed ranges, then reliability of bitstream processing is improved, but manufacturing precision of motion vector reconstruction deteriorates
Solution Approach 1:
The patent applies preliminary scaling to MVD values before they are constrained and encoded into the bitstream. By pre-scaling the MVD to match the precision grid (quarter-pel or 1/16-pel), the subsequent constraint operations preserve the scaled precision while ensuring bitstream compliance, thus maintaining reconstruction accuracy.
3Measurement precision
If maximum motion vector resolution is increased, then measurement precision of motion is improved, but loss of information in bitstream increases
Solution Approach 1:
The patent transforms the MVD parameter by scaling it to match the motion precision requirements (e.g., multiplying by 4 for quarter-pel or by 16 for 1/16-pel precision). This parameter change allows high-precision motion representation while using efficient entropy coding for the scaled values, minimizing bitstream information loss.
Data Source
AI summary
A method of video processing comprises determining, for a conversion between a first block of video and a bitstream representation of the first block, a range of motion vector difference (MVD) component associated with the first block, wherein the range of MVD component is [−2M, 2M−1], where M=17; constraining value of the MVD component to be in the range of MVD component; and performing the conversion based on the constrained MVD component.