Non-Referential Low-Contrast Defect Detection Using Binary Slicing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing inspection techniques struggle to accurately detect low-contrast defects in semiconductor devices, such as watermarks, due to minimal contrast differences between defect and non-defect areas, making them difficult to identify during manufacturing.
Innovation Solution
A method involving binary slicing, segmentation based on statistical contrast properties, and a supervised machine learning network is used to enhance image contrast and detect defects without relying on reference images, utilizing preprocessing, binarization, and a multi-layered machine learning model to identify defects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If reference image-based inspection techniques are used, then defect detection capability is improved, but system complexity and difficulty in operation increase
Solution Approach 1:
The patent extracts and removes the dependency on reference images from the inspection system. By using unsupervised learning algorithms that analyze only the current device image, the system eliminates the need for storing, managing, and comparing against reference images, thereby reducing system complexity while maintaining defect detection capability
Solution Approach 2:
The patent creates statistical representations (copies) of normal device structures through clustering algorithms. Instead of using actual reference images, the system generates statistical models that represent typical device features, enabling defect detection through comparison with these statistical copies rather than physical reference images
2Measurement precision
If high-resolution imaging is used to detect low-contrast defects, then measurement precision is improved, but processing time and computational resources increase
Solution Approach 1:
The patent segments the image processing task into multiple stages: initial binarization to separate defect regions from background, contrast enhancement specifically applied to segmented regions, and focused analysis only on areas with potential defects. This segmentation approach maintains high detection accuracy while reducing overall processing time by avoiding full-image high-resolution processing
Solution Approach 2:
The patent applies contrast enhancement and detailed analysis only to specific regions identified as potential defect areas, rather than processing the entire image at maximum resolution. This partial action approach concentrates computational resources on critical regions, improving detection accuracy for low-contrast defects while reducing total processing time
3Measurement precision
If full image processing is applied, then defect detection accuracy is improved, but computational resources and processing time increase
Solution Approach 1:
The patent divides the image into multiple binary slices at different threshold levels, then processes each slice independently through clustering algorithms. This segmentation enables parallel processing of different image regions and threshold levels, improving computational efficiency while maintaining comprehensive defect detection coverage
Solution Approach 2:
The patent applies different processing strategies to different regions of the image based on their local characteristics. Regions with high contrast variations receive different treatment compared to uniform regions, allowing the system to optimize computational resources by applying intensive processing only where needed while maintaining overall detection accuracy
Data Source
AI summary
Defect detection techniques can be used for inspecting semiconductor devices, such as CMOS image sensors, during the manufacturing process. The defects can include common defects, such as scratches, dirt, etc., as well as low-contrast defects, such as watermarks. The detection technique may use a supervised machine learning network.