Multi-Resolution Motion Vector Calculation for Hand-Movement Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for calculating motion vectors on standard computers face challenges in achieving high accuracy while minimizing computational load, which is crucial for correcting hand-movement blur and generating high-quality images.
Innovation Solution
A multi-resolution motion vector calculation method that involves defining energy functions, calculating reliability and correlation, and smoothing motion vectors across different resolution levels, allowing for accurate motion vector estimation with reduced computational load.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If motion vectors are calculated with high accuracy using conventional methods, then the reliability of motion vectors is improved, but the computational load on the computer increases significantly
Solution Approach 1:
The patent divides the image data into multiple resolution levels (e.g., original resolution, half resolution, quarter resolution). Motion vectors are calculated separately at each resolution level, starting from the lowest resolution and progressively refining to higher resolutions. This segmentation reduces the computational complexity at each stage while maintaining overall accuracy through the hierarchical refinement process.
Solution Approach 2:
The patent performs preliminary motion vector calculation at lower resolution levels before calculating at higher resolutions. The motion vectors obtained from lower resolution levels serve as initial estimates or constraints for the higher resolution calculations, reducing the search space and computational requirements at each subsequent stage.
2Measurement precision
If feature points are identified and polygons are constructed to calculate motion vectors, then the accuracy of motion vectors is maintained, but the data amount and processing complexity increase
Solution Approach 1:
Instead of processing all pixels or constructing complex polygons across the entire image, the patent segments the problem by calculating motion vectors at multiple resolution levels. This reduces the number of data points that need to be processed at each stage while preserving the essential motion information through the hierarchical structure.
Solution Approach 2:
The patent adds the resolution level dimension to the motion vector calculation process. By introducing this additional dimension (different resolution levels), the method reduces the two-dimensional pixel space at each stage, thereby reducing data amount while maintaining accuracy through progressive refinement from coarse to fine resolutions.
Data Source
AI summary
A motion vector calculation method includes: processing for calculating multi-resolution data up to a predetermined resolution level L from a plurality of inputted image data; and processing for estimating motion vectors per resolution executing a matching step for temporarily calculating motion vectors minimizing an energy function within a predetermined range of the input motion vectors and a smoothing step for finally calculating the motion vectors by averaging the surrounding data. By using the motion vectors of the resolution level L finally calculated as input motion vectors of a further higher resolution level L-1, the processing for estimating motion vectors per resolution is successively repeated on image data of a high resolution, thereby calculating the motion vectors.