SE/BSE Image Synthesis Preserving Surface and Shadow Information
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image synthesis methods for semiconductor circuit patterns lose surface and shadow information when combining SE and BSE images, leading to excessive enhancement of shadows, reduction of contour sharpness, or occurrence of artifacts.
Innovation Solution
An image processing method using machine learning to train a model with multiple images under different conditions, estimating structural and material features, and calculating shadow and gradation data to generate a synthesized image without losing surface or shadow information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If SE and BSE images are combined using a given mixing ratio to generate a synthesized image, then the image synthesis process is simple and fast, but surface information and shadow information are lost in the synthesized image
Solution Approach 1:
The patent segments the image synthesis process into two independent stages: first generating a shadow map from structural features, then combining it with a brightness map from material features. This segmentation preserves both shadow information (from BSE) and surface information (from SE) without mutual interference, resolving the information loss problem while maintaining computational efficiency.
Solution Approach 2:
The patent introduces a shadow map as an intermediary element that carries shadow information separately from the brightness map. This intermediary allows shadow information to be preserved and applied selectively during image synthesis, preventing the information loss that occurs in direct mixing ratio methods.
2Device complexity
If image quality improvement processing is performed uniformly across the entire image, then processing is simple, but excessive enhancement of shadows, reduction of contour sharpness, or occurrence of artifacts occurs
Solution Approach 1:
The patent applies local quality by estimating structural features and material features at each pixel location independently, then generating shadow maps and brightness maps with locally optimized characteristics. This allows different regions of the image to have different enhancement characteristics, preventing uniform over-enhancement and preserving contour sharpness while maintaining processing simplicity.
3Measurement precision
If a trained neural network is used to convert low quality images into high quality images, then image quality is improved, but the method performs processing uniformly across the entire image causing excessive enhancement of shadows
Solution Approach 1:
The patent segments the neural network processing into separate structural feature estimation and material feature estimation pathways. The structural features are used exclusively for shadow map generation, while material features are used for brightness map generation. This segmentation prevents the uniform processing problem and preserves shadow information by dedicating specific processing streams to specific information types.
Data Source
AI summary
To implement image synthesis without loss of surface information of SE images and shadow information of BSE images, the present disclosure proposes image processing techniques include applying data of a first quality image (low quality image) to a trained model, estimating a structural feature and a material feature of a second quality image (high quality image) corresponding to the first quality image, calculating at least one shadow datum based on the structural feature and a synthesis parameter and calculating at least one gradation datum based on the material feature and a synthesis parameter, generating a synthesized image from the at least one shadow datum and the at least one gradation datum, and outputting the synthesized image as a prediction result of the second quality image (see FIG. 8).