CN-122023539-A - Rapid calculation method for image drift amount of electron microscope based on Fourier transform

CN122023539ACN 122023539 ACN122023539 ACN 122023539ACN-122023539-A

Abstract

The application discloses a fast calculation method of electron microscope image drift amount based on Fourier transform, comprising the steps of obtaining two electron microscope images with the same size, and respectively naming the images as And Image is processed And Respectively performing Fourier transform to obtain images And And carrying out Fourier transform on the phase difference, and determining the drift amount of the image according to the distance between diffraction points in the frequency domain. The method effectively solves the problems of high calculation complexity and low processing speed in the prior art by a Fourier transform method, and can obviously improve the calculation efficiency while ensuring the accuracy.

Inventors

ZHANG DALIANG
ZHAO GANG
WEI NINI
QU MINGYANG

Assignees

重庆大学

Dates

Publication Date: 20260512
Application Date: 20251230

Claims (9)

1. The fast calculation method of the image drift amount of the electron microscope based on the Fourier transform is characterized by comprising the following steps of: Two electron microscope images with the same size are acquired, and the images are respectively named as And ; Image is formed And Respectively performing Fourier transform to obtain images And A phase difference in the frequency domain; and carrying out Fourier transformation on the phase difference, and determining the drift amount of the image according to the distance between diffraction points in the frequency domain.
2. The fast fourier transform-based image drift amount calculation method as defined in claim 1, wherein the image is obtained by combining the image with a reference image And Respectively performing Fourier transform to obtain images And The phase difference of the frequency domain specifically comprises the following steps: Step 1, setting Translation to the right Downward translation Obtaining a translated image: ; Step 2, pair And Fourier transforming to obtain respectively Frequency domain signal of (2) And Frequency domain signal of (2) : Wherein, the And (3) with The frequency coordinates are represented as such, Representing the units of an imaginary number, The width of the image is represented and, The height of the image is indicated and, And Representing the frequency domain signal; and 3, obtaining according to the translation characteristic of Fourier transformation: ; step 4, connecting And Dividing the frequency domain rate signal to obtain a complex matrix: Wherein, the Representing a complex matrix comprising And Amplitude ratio and phase difference information in the frequency domain; step 5, through complex matrix Calculating a phase difference: Namely: Wherein, the An operator representing the complex phase of the complex signal, Representation of Is used for the phase of the (c) signal, Representation of Is a phase of (a) of (b).
3. The fast fourier transform-based method for calculating an image drift amount of an electron microscope according to claim 2, wherein the fourier transforming the phase difference map, determining the image drift amount according to the distance between diffraction points in the frequency domain, comprises the steps of: fourier transforming the phase difference map: Wherein, the Representing the spatial frequency of the transverse stripes, Spatial frequencies representing longitudinal fringes (periodic fringes in positive space correspond to diffraction points in the spatial frequency domain); from the above formula, the periodic fringes can be represented as a group of sharp diffraction points in the frequency domain after the phase difference is subjected to Fourier change, and the group of diffraction points corresponds to the lateral drift amount of the image in half of the distance of the group of diffraction points in the transverse axis direction Half the distance in the longitudinal axis corresponds to the longitudinal drift 。
4. The fast fourier transform-based image drift amount calculation method as defined in claim 2, wherein the obtained image And After the phase difference of the frequency domain, the drift amount of the image can be determined by calculating the fringe number and fringe spacing (namely fringe period) of the phase difference image, and the specific steps are as follows: Step 1, setting The phase difference formula of the fringe period in the direction is: ; Is provided with The phase difference of the fringe period in the direction is completed by one And (3) circulation: Wherein, the Representation of Stripe period in direction; Namely: ; Step 2, calculating Number of stripes in direction (frequency range is ): Wherein, the Representing the number of stripes; From the above formula, in the direction Period of upper stripe And the lateral drift amount Inversely proportional, number of stripes And the lateral drift amount Proportional to, and the same way can be obtained in the direction Period of upper stripe And the longitudinal drift amount Inversely proportional, number of stripes And the longitudinal drift amount Proportional to the ratio.
5. The method for rapidly calculating the image drift amount of the electron microscope based on the Fourier transform according to claim 1, the fringe pattern can also be calculated by the following method, fourier transformation is carried out on the fringe pattern, and the drift amount of the image is determined according to the distance between diffraction points in the frequency domain: step 1, two electron microscope images with the same size are obtained, and the images are respectively named as And ; Step 2, image is processed And an image Pixel-by-pixel addition: Namely: ; And step 3, carrying out Fourier transform on the added result to obtain a corresponding frequency domain signal: Wherein, the Representing the frequency domain signal; Is a fringe pattern; And 4, carrying out Fourier transformation on the fringe pattern, and determining the drift amount according to the generated group of diffraction points, wherein the fringe pattern is similar to the generated phase difference pattern.
6. The method for rapidly calculating the drift amount of the image of the electron microscope based on the Fourier transform according to claim 1, wherein the drift amount is determined by means of statistics of the abscissa of the peak position of the image, while preventing the interference of periodicity and noise.
7. An electron microscope image drift amount rapid calculation device based on fourier transform, which is characterized by comprising: A data acquisition unit for acquiring two electron microscope images with the same size, respectively named as And ; A data processing unit for processing the image And Respectively performing Fourier transform to obtain images And A phase difference in the frequency domain; And the data calculation unit is used for carrying out Fourier transformation on the phase difference and determining the drift amount of the image according to the distance between diffraction points in the frequency domain.
8. An electronic device comprising a processor, a memory and a computer program stored on the memory and executable on the processor, the processor implementing the fast fourier transform-based electron microscope image drift amount calculation method according to any one of claims 1 to 7 when the program is executed.
9. A computer-readable storage medium, wherein a computer program is stored on the computer-readable storage medium, which when executed by a processor implements the fourier transform-based rapid calculation method of an image drift amount of an electron microscope according to any one of claims 1 to 7.

Description

Rapid calculation method for image drift amount of electron microscope based on Fourier transform Technical Field The application relates to the field of signal and image processing, in particular to a fast calculation method for image drift amount of an electron microscope based on Fourier transform. Background Electron microscopes have become an important research tool in the fields of material science, biotechnology, and nanoscience due to their high resolution and accurate imaging capabilities. However, image drift phenomena are a common and critical problem during the practical use of electron microscopes. Image drift refers to the displacement of an image acquired by an electron microscope due to mechanical vibration, environmental interference, thermal drift, or the influence of an electron beam on a sample. This drift can have a significant impact on the accurate analysis and quantitative investigation of the image, especially in time series imaging. To solve the image drift problem, accurately calculating the drift amount of the image is a key step. The calculation of the drift amount is not only a foundation for realizing drift correction, but also has important significance for applications such as dynamic process observation, superposition imaging, high-precision three-dimensional reconstruction and the like. However, due to the characteristics of electron microscope images, the existing drift amount calculation method has the following technical problems: the high noise interference is that the electron microscope image, especially the image under low dosage imaging, has lower signal to noise ratio, and the traditional drift amount calculating method has insufficient precision under the high noise environment and is easy to be influenced by random noise. Multiscale drift-in electron microscope imaging, drift may involve not only simple translation, but also rotation, scaling and even nonlinear deformations, and the existing methods have poor adaptability to complex drift. The real-time requirement is that along with the popularization of dynamic observation and in-situ experiments, the real-time calculation requirement of the image drift amount is continuously increased, the calculated amount of the existing algorithm is larger, and the real-time processing requirement is difficult to meet. Currently, the method for calculating the image drift amount mainly depends on feature point matching, cross correlation and phase cross correlation. (1) Feature point matching is a commonly used way in image drift correction, and is particularly suitable for the situation that geometric transformations such as translation, rotation, scaling and the like exist between two images. The following is a specific step of calculating the image drift amount by the method: a. Feature point detection, namely, extracting key feature points in two images by using a feature point detection algorithm (such as SIFT, SURF, ORB and the like). These feature points are typically local extreme points in the image, with significant image features (such as corner points, edges, or other significant image features). B. Feature descriptor calculation a descriptor (e.g. feature vector of SIFT) is calculated for each detected feature point for matching corresponding points in different images. C. feature point matching, namely calculating similarity of feature point descriptors between two images by using a Brute-Force matching (Brute-Force) or FLANN matching (Fast Library for Approximate Nearest Neighbors) algorithm. The most similar match is found by calculating the distance of the descriptor of each feature point from all descriptors in the other image. D. The key point of eliminating the mismatching point by using the RANSAC algorithm is the division of the interior point and the exterior point. First, a minimum subset is randomly extracted from the set of matching points, and a drift model (e.g., a translation, affine transformation, or homography matrix) is fitted. Then, all matching points are substituted into the model, and the error (e.g., the distance of the point from the fitted model) for each point is calculated. And classifying points with errors smaller than the threshold value as inner points according to the preset error threshold value, and taking the points with larger errors as outer points. Through multiple iterations, the model with the largest number of inner points is selected as a final result, and all outer points which do not accord with the model are removed, so that reliable matching point pairs are reserved for subsequent correction operation E. the amount of drift is calculated by estimating the amount of translation using the center point (centroid) of the matching point. The centroid of the feature point (i.e., the average of all feature point coordinates) in each image is first calculated. The amount of drift is then calculated by the distance between the centroids of the two images as follows: Where