Automated Fundus Montage via Offset Vector Triplet Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current fundus imaging techniques face challenges in producing accurate, consistent planar representations of the spherical fundus surface due to wide-angle lens distortions and the need for manual image stitching, which is time-consuming and prone to inaccuracies.
Innovation Solution
A device and method incorporating a local matching module for determining best offset vectors between overlapping images and a global matching module to align triplet images with zero offset vector sum, along with a projection and optimization module to create a consistent spherical montage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual image matching is used, then image alignment accuracy can be achieved, but time consumption increases significantly
Solution Approach 1:
The patent replaces manual mechanical image matching with an automated computer-based system that uses offset vector calculations and triplet matching algorithms to automatically align retinal images, eliminating the need for manual intervention while maintaining alignment accuracy
Solution Approach 2:
The system performs self-matching by automatically calculating offset vectors between overlapping images and identifying triplets with zero offset vector sum, enabling the montage process to complete itself without external manual operation
2Area of stationary object
If wide angle lens is used to capture larger fundus area, then field of view increases, but image resolution deteriorates
Solution Approach 1:
The patent divides the fundus imaging into multiple separate images taken with a moderate field of view lens, then automatically stitches these segmented images together using offset vector matching to create a comprehensive montage that maintains both resolution and coverage
Solution Approach 2:
The system transitions from attempting to capture the entire fundus in a single two-dimensional image to creating a composite montage by combining multiple two-dimensional images through automated geometric matching and stitching algorithms
3Area of stationary object
If wide angle lens is used, then larger fundus area is captured, but image distortion increases
Solution Approach 1:
The patent replaces optical distortion correction with a computational approach, using automated offset vector calculations and triplet matching algorithms to geometrically align and stitch multiple images, thereby eliminating wide-angle distortion without sacrificing field of view
4Productivity
If automatic local matching of image pairs is implemented, then processing speed improves, but global consistency is lost
Solution Approach 1:
The patent merges multiple local image pairs into global triplets by identifying sets of three images where the offset vectors sum to zero, thereby combining the speed benefits of automatic local matching with the global consistency required for accurate fundus montages
Data Source
AI summary
Some embodiments of the present invention may relate to a device and a method of enabling an automatic global matching of a plurality of images to provide a substantially consistent planar representation of a fundus. According to some embodiments of the invention, a device for enabling an automatic global matching of a plurality of images to provide a substantially consistent planar representation of a fundus may include a local matching module and a global matching module. The local matching module may be adapted to locally match a pair of overlapping images. As part of locally matching the images, the local matching module may be adapted to provide a best offset vector for the images based upon a matching of features from overlapping portions of the images. The global matching module may be adapted to globally match at least a triplet of locally matching pairs of images whose best offset vector sum is substantially zero.