Charged Particle Image Quality Matching for Stable Neural Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Charged particle microscopy image processing using Artificial Neural Networks (ANNs) and Convolutional Neural Networks (CNNs) is highly dependent on stable imaging conditions, making it challenging to maintain segmentation quality, especially in noisy or poorly focused images, and retraining networks is expensive and impractical in production environments.
Innovation Solution
A method that adjusts image quality parameters by deteriorating high-quality images and enhancing low-quality images to a moderate level, allowing them to be processed by ANNs/CNNs trained on medium-quality images, thereby reducing the need for retraining and ensuring consistent operation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If ANNs/CNNs are trained on high-quality images, then segmentation quality is improved, but the system becomes highly sensitive to imaging condition variations and requires frequent retraining
Solution Approach 1:
The patent transforms images by adjusting quality parameters (noise level, focus, contrast) to match the training conditions of the ANN/CNN. This parameter transformation allows the system to maintain consistent segmentation quality across varying imaging conditions without retraining the network, directly resolving the contradiction between achieving high segmentation quality and maintaining adaptability to different conditions
2Measurement precision
If the network is retrained to account for new imaging conditions, then segmentation quality under new conditions is improved, but computation time and cost increase significantly
Solution Approach 1:
The patent performs image transformation as a preliminary step before segmentation, converting images with varying quality parameters into a standardized format that matches the original training conditions. This preliminary action eliminates the need for time-consuming retraining operations while maintaining segmentation quality under new imaging conditions
3Productivity
If images with varying quality parameters are processed directly, then processing speed is maintained, but segmentation accuracy deteriorates due to noise and focus issues
Solution Approach 1:
The patent introduces an image transformation step as an intermediary process between image acquisition and segmentation. This intermediary transforms images with varying quality parameters into a standardized format, ensuring consistent segmentation accuracy while maintaining efficient processing throughput by avoiding repeated retraining operations
Data Source
AI summary
The invention relates to a method implemented by a data processing apparatus, comprising the steps of receiving an image; providing a set-point for a desired image quality parameter of said image; and processing said image using an image analysis technique for determining a current image quality parameter of said image. In the method, the current image quality parameter is compared with said desired set-point. Based on said comparison, a modified image is generated by using an image modification technique. The generating comprises a step of deteriorating said image in terms of said image quality parameter in case said current image quality parameter exceeds said set-point. The modified image is then output and may be further analysed.