SEM Image Distortion Correction via Reference Grid Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for constructing panoramic images from mosaics of scanning electron microscopy images face challenges in distortion correction, requiring significant computational resources and reference measurements, which are not suitable for large numbers of distorted images and do not allow alignment of different panoramic images.
Innovation Solution
A method that processes elementary SEM images by determining patterns aligned on a reference grid, forming an average elementary grid, and transforming images to correct distortion, allowing for efficient correlation and construction of panoramic images with reduced computational resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional correlation-based methods are used to construct panoramic images from SEM images, then alignment of images can be achieved, but considerable computational resources are required
Solution Approach 1:
The patent segments the panoramic image construction process into two distinct stages: (1) distortion correction of individual SEM images using grid-based transformation, and (2) correlation-based alignment of corrected images. This segmentation allows the computationally intensive distortion correction to be performed independently on each image using efficient grid transformations, rather than requiring resource-intensive correlation calculations on all images simultaneously.
Solution Approach 2:
The patent applies distortion correction as a preliminary action before performing correlation-based alignment. By pre-correcting the distortion in each SEM image using grid transformations and reference grids, the subsequent correlation operation works on already-corrected images, reducing the computational complexity and resources needed for the final alignment step.
2Manufacturing precision
If reference measurements are used during image acquisition to correct distortion, then distortion correction can be achieved, but the method requires reference measurements to be carried out during acquisition
Solution Approach 1:
The patent creates a reference grid that represents the ideal undistorted pattern layout. Instead of requiring physical reference measurements during acquisition, the method uses this reference grid model to compare against the actual detected patterns in each SEM image, calculating the distortion transformation by comparing expected versus observed pattern positions.
Solution Approach 2:
The patent transforms the distortion correction approach from requiring direct physical reference measurements to using parameter-based grid transformations. By detecting pattern positions and comparing them against the reference grid parameters, the method calculates transformation parameters that correct the distortion, converting a measurement-intensive process into a parameter-based computational approach.
3Productivity
If conventional methods are used to construct panoramic images, then images can be assembled, but different panoramic images cannot be aligned with each other
Solution Approach 1:
The patent applies grid-based distortion correction to all SEM images across multiple panoramic views using a universal reference grid model. This reference grid serves as a common coordinate system that enables not only the construction of individual panoramic images but also the alignment of multiple panoramic images with each other, as they all reference the same grid structure.
Solution Approach 2:
The patent introduces a reference grid dimension as an intermediate coordinate system that transcends individual panoramic images. By transforming all images into this common grid reference frame, the method enables alignment across multiple panoramic views, adding a universal reference dimension that facilitates multi-image coordination.
Data Source
Figure 1~2C
Figure 3A~3C
Figure 4
AI summary
The invention concerns a method for processing a plurality of elementary images acquired on a zone of an integrated circuit to be imaged, comprising the following steps: -Providing the plurality of elementary images in mosaic form, -Determining patterns in each of the elementary images, -Determining, in each of the elementary images, a first line of alignment of the patterns and a second line of alignment of the patterns, -Forming an elementary grid for each elementary image, from the first and second lines, -Determining first and second mean grid pitches, -Forming an elementary reference grid having the first and second mean grid pitches, -Transforming the elementary grid comprising the positions of the patterns for each of the elementary images, such that it has the first and second mean grid pitches. This method advantageously allows SEM image distortion to be corrected.