Automatic Defect Image Identification via Complexity Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In high-density IC designs, accurately identifying defect versus reference images in a die-to-die inspection mode is challenging, leading to potential misidentification and increased manufacturing costs due to operator unavailability or inexperience.
Innovation Solution
A method is developed to automatically distinguish defect and reference images by aligning multiple images using alignment keys and circuit features, computing the complexity of overlapping areas through spatial frequency transforms or partitioning, and designating the more complex image as the defect image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual review by operator is used to identify reference image, then identification accuracy may be maintained, but manufacturing cost increases and fabrication line downtime increases
Solution Approach 1:
The system enables automatic self-identification of the reference image through computational analysis. The processor automatically compares complexity metrics of multiple images and designates the less complex image as the reference image, eliminating the need for manual operator review while maintaining identification accuracy.
Solution Approach 2:
The manual mechanical process of operator review is replaced with an automated computational system. The processor uses spatial frequency transforms and complexity calculations to automatically distinguish the reference image from defect images, substituting human judgment with algorithmic analysis.
2Measurement precision
If operator review is used for image identification, then accurate distinction between defect and reference images can be achieved, but time consumption increases
Solution Approach 1:
The system performs preliminary computational preparation by pre-calculating complexity metrics for all images before final comparison. Spatial frequency transforms and partitioning operations are executed in advance, so when identification is needed, the system can quickly compare pre-computed values rather than performing full analysis from scratch.
Solution Approach 2:
Time-consuming manual review is replaced with rapid automated computational comparison. The processor uses pre-computed complexity metrics to instantly identify the reference image, eliminating the time required for operator inspection while maintaining accuracy.
3Measurement precision
If complexity computation using spatial frequency transforms is performed, then accurate image distinction is achieved, but computational resources and processing time increase
Solution Approach 1:
The image analysis is segmented into distinct computational stages: spatial frequency transform, partitioning, and complexity calculation. Each stage processes only necessary portions of the data, and intermediate results are stored for reuse, reducing redundant computations and energy consumption.
Solution Approach 2:
The system performs partial complexity computation by focusing analysis only on regions of interest within images. Rather than computing full image complexity, the processor analyzes specific areas where defects are likely to occur, reducing overall computational load while maintaining identification accuracy.
Data Source
AI summary
Defect printability analysis in a mask or wafer requires the accurate identification of defect images and reference (i.e. defect-free) images, in particular for a die-to-die inspection mode. A method of automatically distinguishing a reference image from a defect image is provided. In this method, multiple images can be accessed and aligned. Then, a common area of the multiple images can be defined. At this point, a complexity of each of the images, as defined by the common area, can be computed. Advantageously, by comparing the complexity of the multiple images, the reference and defect images can be quickly and accurately designated. Specifically, the most complex image is designated the defect image because the defect image must describe the defect. Complexity can be computed using various techniques.