Red-Eye Detection Using Reference Image Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current digital image processing methods for red-eye detection in cameras are inefficient and prone to false detection, especially in low-quality images or limited computational devices, due to the complexity of image analysis and the need for optimal image characteristics.
Innovation Solution
A method that utilizes reference images to improve red-eye detection by analyzing candidate red regions in main images against corresponding regions in preview or reference images, applying a chain of filters including pixel locators, shape analyzers, and falsing analyzers, and performing corrective actions such as contrast normalization and image sharpening.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If complex image analysis algorithms are used for red-eye detection, then detection accuracy may improve, but computational efficiency deteriorates and processing time increases
Solution Approach 1:
The patent divides the image analysis process into multiple filtering stages (pixel locator, shape analyzer, falsing analyzer, pixel modifier) that process the image sequentially. Each stage refines the detection results and eliminates false candidates, achieving high accuracy while maintaining computational efficiency through progressive filtering rather than applying one complex algorithm to the entire image.
2Reliability
If comprehensive image analysis is performed on the main image, then red-eye detection reliability improves, but the complexity of the detection system increases
Solution Approach 1:
The patent captures a reference image before the main image that serves as preliminary data for red-eye detection. This reference image contains information about potential red-eye regions without the flash interference, allowing the system to pre-identify candidate regions and focus subsequent analysis only on those areas, thereby improving reliability without proportionally increasing overall system complexity.
3Adaptability or versatility
If red-eye detection is performed on low-quality or non-optimally acquired images, then the system's adaptability improves, but false detection increases
Solution Approach 1:
The patent introduces a reference image as an intermediary element that mediates between the main image and the detection algorithm. The reference image, captured without flash, serves as a reliable baseline that helps distinguish true red-eye regions from false candidates in the main image, even when the main image quality is poor or suboptimal, thereby reducing false positives while maintaining adaptability.
Data Source
AI summary
A method for red-eye detection in an acquired digital image acquiring one or more preview or other reference images without a flash. Any red regions that exist within the one or more reference images are determined. A main image is acquired with a flash of approximately a same scene as the one or more reference images. The main image is analyzed to determine any candidate red eye defect regions that exist within the main image. Any red regions determined to exist within the one or more reference images are compared with any candidate red eye defect regions determined to exist within the main image. Any candidate red eye defect regions within the main image corresponding to red regions determined also to exist within the one or more reference images are removed as candidate red eye defect regions.