Sample Observation Image Estimation With Regional Loss Functions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing sample observation methods struggle to output images with high visibility for defects and circuit patterns due to the difficulty in learning different image qualities for various regions using a single loss function.
Innovation Solution
The method involves acquiring first and second learning images, dividing them into regions, and using region-specific loss functions to estimate images with varying qualities, employing techniques like deep neural networks and region division based on luminance gradients, layer determination, and defect detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single loss function is used to learn the correspondence between captured images and high-quality images, then the learning process is simple and efficient, but the ability to estimate images with different image qualities for different regions is compromised
Solution Approach 1:
The patent divides the image into multiple regions based on luminance gradients and layer determination. Each region is then processed with region-specific loss functions that are tailored to the characteristics of that region. This segmentation allows different image quality estimations for different regions while maintaining a unified learning framework, thus resolving the contradiction between learning simplicity and regional adaptability.
Solution Approach 2:
The patent applies local quality by assigning different loss functions to different regions of the image. Each region receives a loss function that is optimized for its specific characteristics (e.g., high luminance regions vs. low luminance regions, different layers). This allows the system to estimate images with appropriate quality for each region while maintaining overall learning efficiency through a modular approach.
2Manufacturing precision
If different loss functions are applied to different regions, then image estimation performance and regional adaptability are improved, but the complexity of the learning process increases
Solution Approach 1:
The patent segments the image into regions based on luminance gradients and layer information, then applies specialized loss functions to each segment. This segmentation strategy improves image estimation performance by tailoring the loss function to regional characteristics while maintaining manageable complexity through automated region detection and classification algorithms.
Solution Approach 2:
The patent changes the loss function parameters dynamically based on region characteristics. Instead of using a single fixed loss function, the system adjusts loss function parameters (such as weighting factors, threshold values) according to the luminance and layer properties of each region. This parameter adaptation improves estimation performance while keeping the overall learning process systematic and reproducible.
Data Source
AI summary
A method and device for observing a sample, includes acquiring a first learning image and a second learning image corresponding to the first learning image., Estimation processing parameters for an estimation engine for estimating the second learning image from the first learning image are learned using the first learning image and the second learning image, an estimated image estimated from the first learning image and the second learning image corresponding to the first learning image are divided into areas Ri (i=1 to N, and N is the number of areas) during learning of the estimation processing parameters, and the loss of pixel groups Pi from the second learning image and pixels groups Qi from the estimated image which are included in each of the areas Ri is learned using a loss function Fi for performing evaluation by means of a predetermined standard.