Image Processing for Semiconductor Simulation Edge Noise
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In semiconductor process simulations, tuning image processing parameters for edge detection and smoothing is challenging due to complex influencing factors, leading to noise in edge information and increased calculation time, especially with concave or convex shapes.
Innovation Solution
An image processing method that detects edge information, identifies lines, divides the image into areas based on line locations, calculates similarity between adjacent areas, detects boundaries between dissimilar areas, and connects line segments to generate a connected shape, reducing noise and improving simulation accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If conventional edge detection and smoothing parameters are used for image processing, then the process can be completed, but noise is included in edge information and manufacturing precision deteriorates
Solution Approach 1:
The patent segments the image processing into distinct stages: edge detection to obtain edge information, line identification from edge information, area division based on line locations, similarity calculation between adjacent areas, boundary detection between dissimilar areas, and shape generation by connecting line segments. This segmentation allows each stage to be optimized independently, improving edge information accuracy while managing noise effectively.
Solution Approach 2:
The patent performs preliminary actions by first identifying lines from edge information and dividing the image into areas before detecting boundaries. This preliminary organization of data structures and preprocessing of edge information enables more accurate boundary detection and reduces the impact of noise on final shape extraction.
2Productivity
If conventional image processing parameters are used, then the process can be completed, but tuning parameters is difficult and time-consuming
Solution Approach 1:
The patent changes the approach from manually tuning edge detection and smoothing parameters to an automated process that identifies lines, divides areas, calculates similarities, and detects boundaries based on dissimilarity thresholds. This parameter transformation converts a complex manual tuning problem into a more systematic automated process, improving productivity while reducing the perceived complexity for users.
Solution Approach 2:
The system performs self-service by automatically calculating similarities between adjacent areas and autonomously detecting boundaries based on dissimilarity, without requiring manual parameter adjustment. The process adapts to different images and scenarios autonomously, significantly improving parameter tuning efficiency.
3Loss of time
If conventional edge detection is used, then edge information can be obtained, but concave or convex shapes increase calculation time
Solution Approach 1:
The patent segments complex shapes with concave or convex features into simpler geometric primitives (lines and line segments) through systematic identification and connection. By dividing the image into areas and detecting boundaries between dissimilar areas, the method represents complex shapes as sequences of connected line segments, which reduces calculation time for subsequent shape simulation while preserving essential geometric features.
Solution Approach 2:
The patent approximates complex curved or irregular shapes with straight line segments connecting key boundary points. This linearization approach replaces complex concave or convex curves with simpler polygonal representations, significantly reducing the computational burden of shape simulation while maintaining adequate geometric fidelity for process simulation purposes.
Data Source
AI summary
An image processing method includes the steps of detecting edge information from an input image, identifying a plurality of lines from the edge information, dividing the input image into a plurality of areas based on the relative locations of the plurality of identified lines, calculating a similarity between adjacent areas of the plurality of divided areas, detecting boundaries between the adjacent areas as line segments partitioning the adjacent areas based on a degree of dissimilarity of the adjacent areas, wherein each of the line segments is at least a portion of the plurality of lines, and connecting the line segments forming the boundaries, and generating a connected shape using the boundaries.