History-Based Motion Vector Prediction Using Affine Control Points
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding techniques face challenges in efficiently compressing high-quality video data, leading to increased demands on communication networks and storage devices, as they struggle to balance bit rate reduction with maintaining video quality.
Innovation Solution
The implementation of history-based motion vector prediction using affine motion information, which estimates a translational motion vector for a block based on control points and populates a history-based motion vector predictor table, expanding the type of motion information included in the table and improving prediction efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If video coding techniques use traditional motion prediction methods, then implementation complexity is low, but video compression efficiency and quality are insufficient
Solution Approach 1:
The motion model is segmented into multiple control points (e.g., top-left, top-right, bottom-left control points) that independently define motion parameters for different regions of the block. This allows the system to capture complex motion patterns within each block while maintaining manageable computational complexity through localized parameter estimation.
Solution Approach 2:
The patent extends traditional translational motion models by adding affine transformation dimensions, introducing rotation and scaling parameters alongside translation. This dimensional expansion enables more accurate representation of complex motion patterns such as object rotation and perspective changes, thereby improving compression efficiency for high-quality video.
2Loss of substance
If video data is compressed with higher bit rate reduction, then storage and transmission demands decrease, but video quality degradation increases
Solution Approach 1:
The patent changes the parameters used in motion compensation from simple translational vectors to affine transformation parameters including rotation and scaling. By estimating control point positions and deriving affine parameters (a11, a12, a21, a22, b1, b2) from these control points, the system achieves more accurate motion compensation that maintains video quality even at reduced bit rates.
Solution Approach 2:
The patent replaces traditional block-matching mechanical search methods with analytical affine transformation-based motion estimation. Instead of exhaustively searching reference frames for matching blocks, the system uses control point geometry and affine mathematics to directly compute motion parameters, achieving both computational efficiency and high prediction accuracy.
3Measurement precision
If traditional motion vector prediction is used, then processing speed is high, but prediction accuracy for complex motion patterns is low
Solution Approach 1:
The patent performs preliminary estimation of control point positions and affine parameters before actual motion compensation. By pre-calculating control point coordinates and deriving affine transformation parameters in advance, the system prepares accurate motion models that can be efficiently applied during encoding/decoding, balancing prediction accuracy with processing speed.
Solution Approach 2:
The system uses self-service by deriving affine motion parameters from the video data's own geometric structure (control points within blocks) rather than relying on external reference block matching. This self-contained approach using internal block geometry enables accurate motion estimation for complex patterns while maintaining processing efficiency.
Data Source
AI summary
Systems, methods, and computer-readable media are provided for updating history-based motion vector tables. In some examples, a method can include obtaining one or more blocks of video data; determining a first motion vector derived from a first control point of a block of the one or more blocks, the block being coded using an affine motion mode; determining a second motion vector derived from a second control point of the block; based on the first motion vector and the second motion vector, estimating a third motion vector for a predetermined location within the block; and populating a history-based motion vector predictor (HMVP) table with the third motion vector.