Non-linear rasterized contour filters for IC lithography
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image-processing techniques for simulating contour lines in different stages of integrated circuit design, such as mask-making and substrate patterning, face challenges in accurately converting contour information between pixelized images of varying resolutions due to infinite bandwidth and non-linear contouring, leading to slow processing and introduction of artifacts.
Innovation Solution
A conversion module that divides pixelized images into regions and applies pre-determined nonlinear contributions based on pixel values relative to a threshold, allowing for faster and more precise contour conversion without requiring vector-based representations, using a lookup table for efficient access to precomputed contributions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If vector-based contour construction and rasterization is used, then contour accuracy is improved, but processing time increases significantly
Solution Approach 1:
The patent segments the image into multiple regions based on contour proximity and divides the contour conversion process into discrete steps: identifying contour pixels, assigning region labels, and applying resolution-specific contribution tables. This segmentation allows the complex vector-based contour reconstruction to be broken down into manageable pixel-level operations that can be processed efficiently in parallel.
Solution Approach 2:
The patent pre-computes contribution tables for different image resolutions before the actual contour conversion process. These tables contain pre-determined pixel contribution values that are specific to each resolution level, allowing the conversion module to simply look up and apply the appropriate contributions during processing, eliminating the need for real-time complex calculations.
2Productivity
If linear filtering is applied to the image, then processing speed is improved, but contour accuracy deteriorates due to non-linear contouring requirements
Solution Approach 1:
The patent changes the fundamental parameter of how contour information is represented and processed. Instead of using continuous linear filtering, the system uses discrete contribution tables that map region characteristics to specific pixel contributions. This parameter change allows the system to achieve non-linear contour accuracy while maintaining efficient processing speeds through lookup operations rather than complex mathematical computations.
3Device complexity
If threshold-based contour extraction is used, then simplicity is improved, but artifacts are introduced such as splitting, merging, and hole formation
Solution Approach 1:
The patent introduces an intermediary region labeling system that acts as a mediator between the simple threshold-based contour extraction and the final accurate contour representation. By assigning region labels to pixels based on their spatial relationship with contours and using these labels to select appropriate contribution tables, the system eliminates artifacts while maintaining the simplicity of threshold-based initial extraction.
Data Source
AI summary
A system includes a conversion module that preserves the shape of a contour when converting an image to a different resolution. The conversion module receives a first image and divides the first image into regions of pixel values. For each region, a contribution of the region to the pixel values in the second image is determined. The contribution is selected from a set of pre-determined contributions that are a nonlinear function of the values in the region, and the selection is made based at least in part on the values in the region. The contributions are accumulated together to generate a second image. The conversion module may be, for example, part of a design flow for an integrated circuit that connects a mask simulation stage with an optical simulation stage.