Motion Estimation Using Motion Blur Directional Weighting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing motion estimation algorithms face challenges in accurately determining motion vectors, particularly in areas with small objects, lighting changes, and multiple motions, leading to erroneous vectors and ambiguities, which can result in motion blur and artefacts.
Innovation Solution
The use of motion blur information to calculate a directional weighting term that penalizes inconsistent candidate vectors, improving vector selection and convergence to true motion, by incorporating edge detection and adjusting the maximum allowed vector size in problematic areas.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If conventional motion estimation algorithms are used, then computational complexity is reduced by dividing the image into blocks, but accuracy of motion vector determination deteriorates in areas with small objects, lighting changes, and multiple motions
Solution Approach 1:
The patent applies local quality by analyzing motion blur characteristics at the pixel level rather than treating entire blocks uniformly. Each pixel's motion blur state is evaluated to determine appropriate weighting factors, allowing accurate motion vector estimation in complex regions (small objects, lighting changes, multiple motions) while maintaining efficiency in simple regions.
Solution Approach 2:
The patent changes the evaluation parameters from simple intensity-based metrics (SAD, MSE) to a composite metric that incorporates motion blur information. By introducing motion blur detection as a new parameter and using it to calculate directional weighting factors, the system achieves more accurate motion vector determination without significantly increasing computational complexity.
2Device complexity
If error functions based on intensity differences (SAD, MSE) are used to evaluate motion vectors, then computation is simplified, but reliability deteriorates due to ambiguities in small object areas and lighting changes
Solution Approach 1:
The patent introduces motion blur information as an intermediary factor that mediates between the intensity-based error function and the final motion vector selection. This intermediary provides additional reliability cues without requiring complete replacement of the existing error function, thus maintaining computational simplicity while improving reliability in ambiguous regions.
Solution Approach 2:
The patent implements feedback by using motion blur detection results to adjust the weighting factors in the error function evaluation. The motion blur information feeds back into the motion vector selection process, allowing iterative refinement of the best match without requiring multiple separate processing stages.
3Measurement precision
If motion blur information is incorporated into motion estimation, then accuracy of motion vector estimation is improved, but device complexity increases due to additional edge detection and weighting calculations
Solution Approach 1:
The patent merges the motion blur detection process with the existing motion estimation algorithm. Rather than implementing separate stages for blur detection and motion vector calculation, the system combines these operations by integrating motion blur information directly into the error function evaluation, thereby improving accuracy without proportionally increasing complexity.
Solution Approach 2:
The patent applies partial action by calculating motion blur weights only for pixels where motion blur is detected, rather than processing all pixels uniformly. This selective approach improves accuracy in critical regions while minimizing the additional computational overhead in regions without motion blur.
4Reliability
If multiple candidate vectors are evaluated to handle ambiguities, then reliability is improved, but loss of time increases due to additional iterations and evaluations
Solution Approach 1:
The patent applies preliminary action by pre-calculating motion blur weights for all pixels before the main motion vector estimation process. This preliminary computation allows the error function to immediately benefit from motion blur information during candidate vector evaluation, reducing the need for iterative refinements and decreasing overall processing time while maintaining high reliability.
Data Source
AI summary
A method and apparatus are provided for estimating motion and sequence of video images. A plurality of fields of a sequence of video images are stored, then, for each of plurality of blocks of pixels in each field, a test is performed on a set of candidate motion vectors to determine which motion vector gives the best fit in determining the motion between adjacent fields for that block. The testing step includes determining the amount of motion blur present in the image and modifying the testing of candidate motion vectors in dependence on this motion blur.