Sub-pixel Vector Estimation via Quadratic Surface Fitting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing motion estimation techniques face challenges in achieving sub-pixel accuracy, particularly in reducing motion judder and flicker artifacts during frame rate conversion from film to video, as they rely on pixel resolution interpolation which is costly and provides limited resolution improvement.
Innovation Solution
An apparatus and method that generates refined sub-pixel vectors using quadratic approximations from vector correlation values and converged vector correlation values, defining a minimum vector position and determining correlation samples for a coarse correlation surface estimation, followed by correlation surface fitting to produce refined sub-pixel vectors with horizontal and vertical components.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If pixel resolution interpolation is used for motion estimation, then sub-pixel accuracy is improved, but computational cost increases significantly
Solution Approach 1:
The patent changes the mathematical parameters from pixel-level discrete values to continuous quadratic surface parameters. By fitting a quadratic surface to correlation values at integer pixel positions, the method derives sub-pixel motion vectors through continuous optimization rather than discrete interpolation, achieving higher precision with lower computational cost
Solution Approach 2:
The patent replaces the mechanical interpolation process (generating and storing sub-pixel resolution frame information) with a mathematical substitution approach. Instead of physically creating sub-pixel images, it uses quadratic approximation formulas to directly calculate sub-pixel motion vectors from integer pixel correlation values, eliminating the need for actual sub-pixel frame generation
2Speed
If frame rate conversion is performed to reduce flicker, then display refresh rate is improved, but motion judder artifacts are introduced
Solution Approach 1:
The patent employs feedback through quadratic surface fitting, where correlation values from multiple integer pixel positions are used to fit a continuous quadratic surface. This feedback mechanism allows the system to optimize motion vector precision by considering multiple reference points and adjusting the quadratic parameters to minimize error, thereby reducing motion artifacts while maintaining high refresh rates
Solution Approach 2:
The patent transitions from one-dimensional pixel sampling to two-dimensional quadratic surface approximation. By evaluating correlation values at multiple integer pixel positions and fitting a quadratic surface in the spatial domain, the method extracts motion information from additional spatial dimensions, improving motion estimation accuracy without requiring actual sub-pixel frame generation
3Manufacturing precision
If sub-pixel resolution frames are generated and stored, then resolution detail is improved, but memory requirements and processing complexity increase
Solution Approach 1:
The patent extracts only the essential motion information needed for sub-pixel accuracy without generating complete sub-pixel resolution frames. By using quadratic surface fitting on correlation values from integer pixel positions, it extracts motion vectors directly from the correlation surface, taking out only the necessary motion parameters while discarding the need for full sub-pixel frame generation and storage
Solution Approach 2:
The patent performs preliminary action by calculating correlation values at integer pixel positions first, then using these pre-computed values to fit the quadratic surface. This preliminary calculation of integer pixel correlations provides the foundation for subsequent sub-pixel vector derivation, avoiding the need to perform computationally intensive operations on full sub-pixel frames
Data Source
AI summary
To achieve the foregoing, and in accordance with the purpose of the present invention, an apparatus and method for generating refined sub-pixel vectors for motion estimation from vector correlation values and converged vector correlation values using quadratic approximations respectively is disclosed. The apparatus and method includes defining a minimum vector position value of a converged vector and then determining a predetermined number of vector correlation samples around the minimum vector position value. The predetermined number of vector correlation samples provide a coarse correlation surface estimation of the minimum vector position value. A correlation surface fitting of the predetermined number of vector correlation samples using a quadratic approximation of the coarse correlation surface estimation of the minimum value is then performed. The correlation surface fitting resulting in a refined sub-pixel minimum vector position with horizontal and vertical components.