Local Image Feature Matching with Epipolar Search Regions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing systems face challenges in accurately matching features between images captured from different camera viewpoints due to the inherent errors and uncertainties in feature detection and description methods, leading to a high likelihood of incorrect matches.
Innovation Solution
The method employs epipolar geometry to define a geometrically-constrained region in one image based on a feature in another image, comparing local descriptors within this region to identify a geometric best match and a global best match, and applies predefined thresholds to confirm the correct match, thereby reducing incorrect matches and increasing accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If feature detection algorithms are used to identify high contrast features, then features can be reliably identified and tracked between images, but matching errors still occur due to inherent errors and uncertainties in detection and description methods
Solution Approach 1:
The patent introduces an intermediary verification step that compares the geometric consistency of feature matches across multiple images. This intermediary check acts as a mediator between the feature detection process and the final matching result, filtering out incorrect matches by verifying whether they satisfy geometric constraints (epipolar geometry) before accepting them as valid matches.
Solution Approach 2:
The patent implements a feedback mechanism where the results of geometric verification are used to refine and improve the matching process. By checking whether candidate matches satisfy epipolar geometry constraints and using this information to confirm or reject matches, the system creates a closed-loop feedback system that continuously improves matching accuracy based on geometric consistency.
2Productivity
If local descriptors are used to represent feature characteristics, then feature comparison and matching can be performed efficiently, but incorrect matches still occur due to the high likelihood of errors in conventional methods
Solution Approach 1:
The patent performs preliminary geometric verification before finalizing feature matches. By checking whether candidate matches satisfy epipolar geometry constraints in advance, the system eliminates incorrect matches early in the process, preventing them from propagating through subsequent processing stages and improving overall matching reliability.
Solution Approach 2:
The patent segments the feature matching process into distinct stages: initial feature detection, local descriptor comparison, geometric verification, and final match confirmation. This segmentation allows each stage to focus on specific aspects of matching, with the geometric verification stage specifically dedicated to filtering out incorrect matches based on epipolar geometry constraints.
3Device complexity
If conventional feature matching methods are used without geometric constraints, then the matching process is simpler and faster, but the accuracy of matches decreases due to incorrect matches
Solution Approach 1:
The patent applies partial geometric constraints only to the verification stage rather than to the entire matching process. By using epipolar geometry checks selectively to verify candidate matches rather than to guide the entire matching process, the system adds minimal complexity while achieving significant improvements in matching reliability.
Data Source
AI summary
A method of matching features in first and second images captured from respective camera viewpoints related by an epipolar geometry. The coordinate system of the second image is transformed so as to map an epipolar line in the second image corresponding to a first feature in the first image, to be parallel to one of the coordinate axes of the coordinate system. The epipolar line defines a geometrically-constrained region in the second image in the transformed coordinate system corresponding to the first feature in the first image; measures of similarity between the first feature in the first image and features in the second image are determined; and a best match feature is identified from the measures of similarity between the first feature in the first image and the respective features in the second image.