Eye Torsion Detection via Direction Histogram Shift
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for determining eye torsion from images are complex, require high processing power, and struggle with image regions of different sizes and alignments, complicating precise torsion detection.
Innovation Solution
The method involves capturing at least two eye images, generating direction histograms of gradient or tangential features, and comparing these to determine the torsion angle from the shift or displacement, utilizing features like blood vessels which have scaling and translation invariant properties, allowing for quick and accurate measurements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If eye models are constructed and adapted to current images using reference images, then position and rotation coordinates can be determined, but the process becomes complex and requires high processing power and computation time
Solution Approach 1:
The patent extracts only the essential feature points (cornea center, iris center, pupil center) from the eye image, rather than constructing a complete eye model with all anatomical structures. This selective extraction of critical features simplifies the processing while maintaining measurement accuracy for torsion determination.
Solution Approach 2:
The patent segments the eye image into distinct feature regions (cornea, iris, pupil) and processes each independently to identify characteristic points. This segmentation approach reduces overall complexity by breaking down the complex eye model construction into manageable, independent feature detection tasks.
2Measurement precision
If complete eye models with multiple features are used for torsion determination, then measurement accuracy improves, but processing time and computational power requirements increase
Solution Approach 1:
The patent extracts only the three essential feature points (cornea center, iris center, pupil center) needed for torsion calculation, eliminating the need to process all eye model features. This extraction approach maintains measurement accuracy while dramatically reducing computation time by focusing only on critical features.
Solution Approach 2:
The patent uses a partial approach by selecting only the minimum necessary features (three center points) rather than processing the complete eye model. This partial action is sufficient for torsion determination and avoids the excessive computational burden of analyzing all possible eye features.
3Adaptability or versatility
If image regions of different sizes and alignments are used for torsion detection, then measurement versatility improves, but detecting corresponding regions becomes more difficult
Solution Approach 1:
The patent changes the approach from comparing entire image regions to comparing specific feature point coordinates. By transforming the problem into parameter-based comparison (x, y coordinates of center points) rather than region-based comparison, the system can easily handle images of different sizes and alignments without difficulty in detecting corresponding regions.
Data Source
AI summary
The invention relates to a device for determining eye torsion, comprising a camera and an image processing unit, which is designed to carry out a method for determining eye torsion. In the method, at least two images of an eye are recorded and image data are produced from said images, and at least one characteristic feature of the eye is identified from the image data, for example a blood vessel. Direction histograms of the feature in both images are produced and compared with each other. The angle of the torsion of the eye is determined from the shift of the directions in the direction histograms.