Image Correspondence Filtering with Multi-Attribute Quality Evaluation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image evaluation methods for driver assistance systems face challenges in efficiently selecting reliable correspondences between images with high accuracy, often resulting in excessive data volumes and potential accuracy losses during processing.
Innovation Solution
A method that provides quality measures as attributes for characterizing correspondences, allowing for conditional selection and reduction of data volume without accuracy loss, using techniques such as tessellation, quality measure combination, and temporal extrapolation to represent correspondences as images or matrices, and employing lookup tables for efficient processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If all correspondences are processed without selection, then complete data is available for evaluation, but data volume becomes excessively large and processing efficiency decreases
Solution Approach 1:
The patent applies preliminary action by computing quality measures for all correspondences before the actual selection process. These quality measures (such as match quality, uniqueness, and geometric consistency) are calculated in advance to enable efficient filtering. This preliminary evaluation allows the system to identify and retain only high-quality correspondences while discarding poor matches, thereby reducing data volume without compromising the reliability of the final evaluation results.
Solution Approach 2:
The patent implements local quality by assigning different quality measures to different correspondences based on their individual characteristics. Instead of treating all correspondences uniformly, the system evaluates each correspondence locally using multiple quality metrics (match quality, uniqueness, geometric consistency) and selects correspondences based on their specific quality profiles. This allows the system to maintain high reliability by preserving locally optimal correspondences while reducing overall data volume.
2Measurement precision
If multiple quality measures are combined for evaluation, then selectivity and accuracy of correspondence selection improve, but computational complexity increases
Solution Approach 1:
The patent applies segmentation by dividing the correspondence evaluation process into distinct stages, each handling specific quality measures. The system segments quality assessment into separate components (match quality evaluation, uniqueness verification, geometric consistency checking) and processes them in a structured sequence. This segmentation allows the system to maintain high measurement precision through multiple quality measures while managing computational complexity by organizing the evaluation into modular, manageable stages.
Solution Approach 2:
The patent implements parameter changes by dynamically adjusting the weighting and thresholds of different quality measures based on the specific application context. The system can modify evaluation parameters such as quality thresholds, weighting factors for different measures, and selection criteria to optimize the balance between accuracy and complexity. This flexibility allows the system to adapt to different requirements without permanently increasing structural complexity.
3Productivity
If correspondences are filtered selectively, then data volume is reduced for further processing, but risk of losing accurate correspondences increases
Solution Approach 1:
The patent implements feedback by using the results of quality measure evaluations to guide the selection process. The system computes quality measures for all correspondences, uses these results to identify high-quality matches, and retains them for further processing. This feedback mechanism ensures that correspondences are selected based on objective quality criteria rather than arbitrary filtering, thereby maintaining reliability while achieving productivity gains through reduced data volume.
Solution Approach 2:
The patent applies preliminary action by performing comprehensive quality assessments before the selection decision is made. All quality measures (match quality, uniqueness, geometric consistency) are calculated in advance, allowing the system to make informed selection decisions. This preliminary evaluation ensures that no potentially accurate correspondences are lost during filtering, as the selection is based on thorough pre-computed quality metrics rather than simplistic filtering criteria.
Data Source
AI summary
A method for evaluating images and in particular for evaluating correspondences of images. The method includes (i) providing correspondences between given first and second images, (ii) providing a quality measure or a plurality of quality measures as attributes for characterizing a particular correspondence, (iii) evaluating and conditionally selecting the correspondences, (iv) providing selected correspondences as an evaluation result, the evaluation of correspondences being based on a combination of attributes and the selection of correspondences being based on a result of the evaluation.