Automated Eye Fundus Image Registration Using Anchor Block-Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for creating panoramic images of the eye fundus are time-consuming, limited in reproducibility, and lack robustness due to manual interaction and sensitivity to imaging quality, making it difficult to accurately combine and transform partial images into comprehensive compositions.
Innovation Solution
An automated method involving pre-positioning, anchor image determination, block-matching algorithm application, and quadratic parameter calculation to geometrically and photometrically align and superimpose images, with a field mask to filter out non-relevant areas and enhance image contrast, ensuring accurate and efficient image registration.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If manual marking of corresponding points and semiautomatic transformation methods are used, then geometric accuracy of image alignment can be achieved, but the process becomes time-consuming and requires significant technician interaction
Solution Approach 1:
The system performs automatic landmark detection and image registration without requiring technician intervention. The computer automatically identifies corresponding points across images and calculates transformation parameters, making the system self-sufficient and eliminating manual marking steps.
Solution Approach 2:
The patent replaces manual mechanical operations (cutting, shifting, rotating, and gluing prints) with automated computational methods. Digital images are transformed and superimposed using computer-calculated transformation parameters, substituting physical manual processes with automated digital operations.
2Productivity
If automatic landmark detection methods are used to reduce computational effort, then processing speed improves, but robustness decreases due to sensitivity to imaging quality and framing
Solution Approach 1:
The system performs pre-positioning of images before detailed landmark detection and transformation. This preliminary arrangement establishes a rough geometric framework that guides subsequent automatic landmark detection, making the process more robust to variations in imaging quality and framing.
Solution Approach 2:
The patent divides the image processing into distinct stages: pre-positioning, landmark detection, transformation calculation, and superimposition. This segmentation allows each step to be optimized independently, with pre-positioning providing a stable foundation that improves the reliability of subsequent automatic landmark detection.
3Manufacturing precision
If blood vessel reconstruction methods are used to determine transformations, then accurate anatomical landmarks are obtained, but the computational effort becomes extraordinary
Solution Approach 1:
Instead of performing complete blood vessel reconstruction, the system uses pre-positioning and simplified landmark detection to obtain sufficient accuracy for clinical purposes. This partial action approach achieves adequate precision without the extraordinary computational burden of full vessel tree reconstruction.
Solution Approach 2:
The pre-positioning step establishes a rough geometric alignment before landmark detection, reducing the computational complexity required for accurate transformation. This preliminary arrangement allows the system to achieve good results without performing computationally intensive blood vessel reconstruction.
4Device complexity
If imaging parameters are saved and panoramic images are created analytically without image content, then processing is simplified, but adaptability to actual image variations is reduced
Solution Approach 1:
The system uses automatically detected landmarks as intermediaries between the raw images and the final transformation. These landmarks serve as mediators that adapt the transformation process to actual image content and quality, bridging the gap between simplified parameter-based methods and content-adaptive processing.
Solution Approach 2:
The automatic landmark detection provides feedback about the actual image content and quality, allowing the transformation process to adapt to variations in imaging conditions. This feedback mechanism enables the system to maintain both simplicity and adaptability by adjusting transformations based on detected features.
Data Source
AI summary
A method for creating or calculating panoramic images of the eye fundus particularly from images of a fundus camera. In the method a pre-positioning process is carried out in which a first variable is determined for geometrically associating the images with each other. An anchor image is determined as a reference for the first variables for the geometric association; areas that are associated with each other are determined in the anchor image and the other images by a block-matching algorithm; transformation parameters for a geometric transformation between the anchor image and the other images are determined from the mutual position of the associated areas; and the other images are transformed onto the anchor image by transformation parameters and are superimposed onto the anchor image and among each other.