Wafer Image Contour Extraction via Selection-Based Auto Tuning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing contour extraction tools for integrated circuit images require extensive time and resources for auto-tuning due to the analysis of large sets of wafer images, which is onerous and inefficient.
Innovation Solution
A computing system partitions wafer images into blocks, classifies them based on image characteristics, assigns scores, selects representative images, and uses these to set parameters for contour extraction, thereby reducing the number of images needed for training and speeding up the auto-tuning process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine-learning algorithms are used to analyze large sets of wafer images for auto-tuning, then contour extraction accuracy is improved, but processing time and computational resources increase significantly
Solution Approach 1:
The patent divides the large set of wafer images into multiple subsets and uses a two-stage selection process: first selecting representative images from each subset based on image characteristics, then using these representative images for machine-learning-based parameter tuning. This segmentation reduces the total number of images processed while maintaining accuracy.
Solution Approach 2:
The patent extracts and selects only the most representative wafer images from the complete dataset for use in training the machine-learning algorithm. By taking out only the essential representative images rather than processing all images, the system reduces computational load while preserving the accuracy needed for effective contour extraction.
2Measurement precision
If machine-learning algorithms analyze large sets of wafer images for parameter tuning, then contour extraction accuracy is improved, but computational resources and processing complexity increase
Solution Approach 1:
The patent segments the image analysis process into distinct stages: image characteristic analysis, representative image selection, and machine-learning parameter tuning. This segmentation simplifies the overall processing complexity by breaking down the complex task into manageable steps that can be executed more efficiently.
Solution Approach 2:
The patent extracts only the essential information needed for parameter tuning from the wafer images by selecting representative images. This extraction approach reduces the volume of data that requires processing, thereby decreasing computational resource requirements and processing complexity while maintaining the accuracy needed for effective contour extraction.
Data Source
AI summary
A computing system can obtain wafer images of integrated circuitry having physical structures and classify each of the wafer images based on image characteristics of the wafer images. The computing system can partition each of the wafer images into a plurality of blocks, analyze each of the blocks to determine which of the blocks correspond to a background portion or a contour portion of the wafer images, assign an image score to each wafer image based on the analysis of each of the blocks, and classify the wafer images based on the image scores assigned to the wafer images. The computing system can set parameters for contour extraction using at least one of the wafer images selected from each of the classifications of the wafer images, and extract contours corresponding to the physical structures of the integrated circuitry from the wafer images based, at least in part, on the parameters.