NZ714911B2 - Method for automatic correction of astigmatism - Google Patents

Description

- 1 – METHOD FOR AUTOMATIC CORRECTION OF ASTIGMATISM Technical Field The present invention relates to a method for automatic correction of astigmatism such as astigmatism in lenses of electron-microscopes. More particularly, the present invention includes an automatic image-based objective lens astigmatism correction for electron copes.

Background and y of the Invention High quality image acquisition in electron microscopes requires careful alignment of the electron beam and precise ng for optimized image contrast and fine detail. In the past, physical features and characteristics of the electron-microscope have been important and used to perform the alignment. The electron beam is aligned by using electromagnetic devices. A misaligned on beam results in artifacts (ripples), blurriness in the image, and loss of ation on fine details.

An important e of the present invention is that the method automatically corrects for lens astigmatism during the alignment process by using only image data and without relying on complicated and cumbersome features of the microscope itself. T he method of the present invention - 2 – provides a solution to the above-outlined problems or provides the public or industry with a useful .

More particularly, the method is for automatic astigmatism correction in one direction through a set of lenses. Of , the present invention is not limited to correcting in only one direction because the correction can also be done in many directions simultaneously such as both the x- and y-directions. A first image of a view is provided at a first stigmator setting of a lens. Preferably, the image is under-focused. Based on the first image at the first stigmator setting, a calculating device calculates a first Fourier spectrum image. The distribution and directions of intensities in the Fourier spectrum image are determined by calculating a first vector corresponding to the main direction and extent of the bright pixels, and a second vector being perpendicular to the first vector and corresponding to the extent in that direction. The first vector is compared with the second . The set of lenses is d from the first stigmator setting to a second stigmator setting to provide a second under-focused image of the view. The second image at the second stigmator setting being of the same view as the first image of the view at the first tor setting. Based on the second image, the - 3 – second corresponding Fourier spectrum image is calculated at the second stigmator setting. The distribution and directions of intensities in the second Fourier um image is ined by calculating a third vector and a fourth vector. The third vector is compared with the fourth vector. When the first vector is more similar to the second vector than the third vector is to the fourth vector the first image at the first stigmator setting is selected as being more round than the second image. When the third vector is more similar to the fourth vector than the first vector is to the second vector then the second image at the second stigmator setting is selected as being more round than the first image. The stigmator settings providing the Fourier um with the most round Fourier spectrum is what is strived and searched for.

The method further includes the step of calculating eighted moments of a circular Fourier spectrum image as a means of measuring the direction and extent of the intensity distribution.

In another embodiment, a first ratio of eigenvectors of the first Fourier um image is compared with a second ratio of eigen-vectors of the second Fourier spectrum image. - 4 – The image with the lowest ratio is ed.

The x-stigmator and mator settings are d to the stigmator settings that correspond to the image with the lowest ratio.

The x-stigmator and the y-stigmator settings can also be simultaneously changed.

In yet another embodiment, the stigmator setting that minimizes an elongation value of a selected Fourier spectrum image is searched for.

The first and the second images are set to an under-focus or an over-focus.

Brief Description of Drawings Fig. 1 is a schematic view of examples of images acquired with different objective lens x- and y-stigmator settings; Fig. 2 is a schematic view of the corresponding Fourier spectra of the images of Fig. 1; Fig. 3 is a schematic view of an image seen with some micrometers under-focus through a lens with astigmatism; Fig. 4 is a schematic view of a log Fourier um of the image in Fig 3; Fig. 5 is a cut-out circular image of the view of Fig. 4; - 5 – Fig. 6 is the view of Fig. 5 including eigenvectors of the gray-weighted moments of the gray-values inside the circle n the length of the vectors corresponds to the sizes of the eigen-vectors; Fig. 7 is the same view as Fig. 4 but with no or very little under-focus; and Fig. 8 is a schematic view of an information flow of the method steps of the present invention.

Detailed ption Figs. 1-2 are schematic illustrations 100 of astigmatic images of objects with roughly circular shapes (see Fig. 1) and corresponding Fourier spectra images 120 (see Fig. 2). It does not matter what the image shows but the depicted objects in Fig. 1 are virus-like particles.

Astigmatism in the ystem causes ripple like artifacts and loss of detail in images, corresponding to the Fourier spectrum images to become elongated in a ion that correlates to the direction of the misalignment. Fig. 1 shows examples of images with astigmatism in different directions. More particularly, Fig. 1 shows sample images acquired at ent x- and y-stigmator settings. It shows 49 versions of the same view in a transmission electron - 6 – microscope with different x- and y-stigmator settings.

Because the settings are different, the images are blurry in ent ways and have waviness in different directions.

One object of the t invention is to determine which version is the best image (corresponding to the Fourier spectrum image with the roundest distribution of the pixels).

This is difficult to do by merely using manual and visual methods which are most often used today. A feature of the present invention is that the method automatically determines which image is roundest by only using image data. This is then used to automatically adjust the lenses in the microscope so the image becomes stigmatic.

Fig. 2 shows the corresponding Fourier a images 120. The transformed images are ted in directions that correlate to the direction of the misalignment of the lenses originating from the stigmator settings. In general, the Fourier transformation makes it easier to see the astigmatism effects, such as roundness and tion, of the images. The 49 versions in Fig. 1 are in Fig. 2 shown as 49 new ns (as white elongated shapes/figures) after the Fourier transformation. The image 122 at the center has very little astigmatism (shown as a round image in the Fourier spectrum). The image at the center - 7 – has an x-stigmator setting of 0 and a y-stigmator setting of 0. It is to be understood that even if it would correspond to other x- and y- stigmator values, it would be the solution because the Fourier spectrum has the most circular pixel distribution at these settings. The further away from the center of Fig. 2, the more ripples and blur exist in the images and the more elongated the pixel butions appear in the Fourier spectrum images. The computer thus s the stigmator settings in a step-by-step n and determines r the image is more round or not before the stigmator setting is changed. As described in detail below, when the image is determined to get rounder as a result of an adjustment of the stigmator setting in the x- or y- direction then the computer preferably changes the microscope’s stigmator g again in same direction. When the images do not get r, the computer may stop the search or possibly change the stigmator setting in the opposite ion and then compare the roundness.

Fig. 3 shows an image 124 that is seen with some under-focus through a lens-system with astigmatism. As explained below, the use of under-focus is an important feature of the method of the present invention to avoid creating Fourier images with complex shapes. It is also - 8 – possible to use an over-focus of the lens. Another important feature of the present ion is that the astigmatism is corrected by searching for a minimal elongation (or optimal circuitry) of the shapes in the Fourier spectra. One important goal is thus to select x- and mator settings that create the most circular distribution of the pixels in the Fourier um i.e. where there is no or very little atism.

The astigmatism correction may be done in at least two different ways. One way is to minimize the elongation in one astigmatism direction at a time, corresponding to one set of electro-magnetic lenses i.e. in one direction such as by only changing the x-stigmator setting before changes of the y-stigmator g is investigated. Another w ay is to calculate the direction of the elongation that corresponds to the ratio of x- and y-astigmatism, and, at the same time, optimize both the x- and y-stigmator settings. The elongation of the images in the r spectra of Fig. 2 is measured by using gray-weighted moments. This can also be used in the one-dimensional situation i.e. adjusting the setting in only one direction. The gray-weighted moments are a way to calculate the direction of the bright pixels in the image i.e. the bright intensity distribution. - 9 – Below is a step-by-step description of the first approach. The x-astigmatism may be corrected by: 1) Acquire an image and calculate the corresponding r spectrum. Fig. 3 is an example of a suitable image 124 that happens to depict virus-like particles at an underfocus of the lens setting. Fig. 4 is an image 126 that shows the corresponding r spectrum. The lighter or lightgrey elliptical portion 128 that extends through the center of the image from the upper left to the lower right tes the astigmatism of the lens system. The portion 128 has been marked with a dashed line to make it easier to see the elliptical portion. The center has also been marked. The size and length of the elliptical n partly depends on the magnification and content in the image. In general, when the light-grey portion is round or rounder then the settings of the lens system are correct to minimize the astigmatism. 2) Cut out a circular image 129 that, for example, has diameter that is half the image side. It is not necessary to use a circular image as long as the grey- weighted moments are calculated of the portion of the image that is of interest. It is however preferred/recommended in order to have the same number of pixels buting to the distribution in all directions. The size of the cut-out - 10 – should match the magnification provided by the lens. Fig. 5 illustrates the cut-out image 129 of the image 126 shown in Fig. 4. A circular image is preferred so that the radius is the same from the center of the image to the periphery and so that there is the same amount of distance and information in all directions from the center. The lower frequencies in and very close to the center of the Fourier spectrum, and the high frequencies very far from the center, carry information less ant as far as indicating atism of the system of lenses. 3) Calculate the distribution of the intensities of the Fourier spectrum image by, for example, using grayweighted moments (such as orders 1 and 2) of the circular image. It is also possible to use an approach that relies on first making the Fourier spectrum image binary (extracting the elliptical/circular bright shape) and then e.g., do principal ent analysis to calculate and determine the bution and pal directions of the binary shape.

It is also possible to use methods that rely on radial density profiles. Preferably, a mathematical method is used to calculate the distribution and direction of the intensity in the image. - 11 – A suitable formula for calculating gray-weighted moments is: �� ���� = ∑ ∑ �� �� �� �� �� (�� , �� )���� , where x and y are the pixel positions inates) in the image, and I(x,y) is the intensity (gray-value) at image position x,y. The order of the moment is i+j, so there are two moments of order 1, �� 01 and �� 10, and 3 moments of order 2, �� 20, �� 02, and �� 11.

The higher the intensity the lighter the image becomes. The lighter the image pixel is and the further away from the center, the more weight it is given when it comes to calculating the gray-weighted moments used to determining astigmatism. In other words, the weight of each pixel is based on the whiteness of each pixel and its distance to the center wherein bright pixels further away from the center are given more weight. Note though that pixels very far from the center (well outside the bright elongated shape in the r spectrum image), do not contain information related to atism.

Fig. 6 is a view 130 that is the same as view 129 of Fig. 5 but it includes eigen-vectors 132, 134 of the grayweighted s of the gray-values inside the circle wherein - 12 – the length of the vectors correspond to the sizes of the eigen-vectors. The gray-weighted moments may be used to calculate the intensity distribution around the center point 138. In other words, the gray-weighted moments are preferably used to be the directions and how stretched out the light-colored portion 128 extends in the ions relative to the center 138. It is then possible to determine how elliptical the light-color portion 128 is and how nonround it is. The non-roundness of the image indicates the level of astigmatism of the lens and the direction of the image contains information about how to correct for the astigmatism. The moments are thus used to derive the main direction of intensities in the image as well as how much more distributed the intensities are along the main direction in relation to the onal direction. The ratio of the distribution in the second direction and the distribution in the main direction is a measure of how ted the bution is. It is to be understood that the ratio is merely an rative example and that other comparison parameters such as differences between the vectors may be used as long as the goal is to determine a change of the stigmator settings that leads to a rounder figure in the Fourier spectrum image. - 13 – 4) ine the elongation that ponds to the ratio of eigenvector 132 and 134 of the normalized moment covariance matrix = [m20/m00, m11/m00; m11/moo, m02/m00]. ector 134 shows the longest tion (i.e. how the long the distribution of whitish pixels is) and extends along the direction of the light colored elliptical shape. Eigen vector 132 shows the second longest elongation that is perpendicular to eigenvector 134 in the two-dimensional image. The direction and length of eigenvector 134 is the most interesting. When the eigenvector 134 has a length that is identical or close to eigenvector 132 then the stigmator setting have been correctly set to compensate for the astigmatism of the lens system i.e. the ratio between the eigenvectors is 1 or close to 1. The further away from 1 the ratio is the more astigmatism of the lens.

) Repeat steps 1-4 above in an iterative optimization algorithm that searches for the stigmator settings that minimizes the elongation value. The adjustments are thus made to make the ratio as close to one as practically possible.

The same correction for y-astigmatism may be done as for the correction for x-astigmatism described above by keeping the x-stigmator at the hitherto determined best - 14 – setting and repeat the procedure for the y-stigmator as was done for the x-stigmator described above. In other words, the correction for y-astigmatism may be done by keeping the x-stigmator at the best setting, as determined above, and repeat steps 1-5 above only for the mator. Instead of ly ining which image is the roundest among all the images along one row in the x-direction of Fig. 1, the method of the present invention thus does this determination automatically by analyzing the r spectrum for each image and determines the amount of elongation and selects the image with the smallest elongation, fter the stigmators of the microscope are set to the corresponding values. In other words, once the roundest image has been found along one row in the x-direction, as shown in Fig. 2, and the corresponding x-stigmator value set in the microscope, then the method may determine the roundest image in the ydirection along the column of the roundest image found in the row in the x-direction.

In order to speed up the correction procedure and to make it more robust, the Fourier spectra can be downsampled and filtered with a smoothing filter (e.g. mean, median or Gauss) prior to calculating the moments.

One important feature of the present invention is - 15 – that the correction is preferably done at a vely large under-focus (or over-focus). When the image is very close to or in focus, the atism effect in the Fourier um is not so easy to single out and measure and it is ult to use those spectra to correct for the astigmatism.

Different directions are then in focus at slightly different under (or over) focus due to the astigmatism. This, together with information from sharp image details, generates strong disturbing signals in the Fourier spectrum with t to determining the direction and amount of astigmatism in the lens system. In other words, by setting the sample so that it is not in or close to focus the fault of the astigmatism is not interfered so much by other factors. Fig. 7 shows an image 140 that has as much astigmatism as the example in Fig. 6 but the corresponding image is in focus. By placing the sample at an under-focus the problem of astigmatism is first solved without any disturbance of any problem associated with having the correct focus of the lens. By using an underfocus , it is easier to determine whether the roundness of the Fourier spectrum improves or not when the stigmator settings of the lens system are d. Image 140 clearly shows how difficult it is to see the direction and extent of the astigmatism because there is no pronounced elongation. In - 16 – addition, the Fourier spectra also depend upon the image t. In a very ly event, the content may produce intensity in the r spectrum that makes the method unsuccessful the first time. This can be solved by repeating the optimization procedure in another place of the sample image to make sure the same answer is obtained.

In operation, certain steps of the method of the present invention are ed in Fig. 8. In a providing step 200, an under-focused image 202 (such as the first image 96 in Fig. 1) is provided at a first value of an x-stigmator setting and at a first value of a y-stigmator setting. In a calculation step 203, a corresponding Fourier spectrum image 204 (such as image 312 in Fig. 2) is calculated by a calculating device such as a computer. In a determination step 206, eighted moments 207 of the image 204 are calculated by the computer to determine the distribution and direction of the pixels in the image. In a comparison step 208, a ratio 209 or difference between eigen-vectors are compared by the computer to determine the roundness of the image 204. In a setting step 210, the er changes the setting in the microscope of the x-stigmator (or the ystigmator ) 211 from the first value to a second value wherein the second value preferably is a sequential increase or - 17 – decrease compared to the first value such as going from e.g., -0.12 to -0.08, as shown in Fig. 1 so that a new image at the second value of the x-stigmator is displayed. The corresponding Fourier spectrum image is then calculated. The computer may be programmed so that it continues to change the x-stigmator in the same direction as long as the vector ratios improve and if the ratios se then the xstigmator setting is changed in the opposite direction instead. It is also possible to use relatively large changes of the stigmator setting in the beginning and then reduce the changes of the x-stigmator setting as the ratios improve i.e. get closer to 1 or it is determined that the likely best setting of the x-stigmator has been passed. For example, if the ratio is gradually reduced and then starts to increase the size of change of the stigmator g is reduced and the direction is d. Preferably, the y-stigmator is kept at the first value such as -0.12 which corresponds to the first row in Figs. 1-2. Figs. 1-2 show preferred examples values for the x-stigmator and for the mator.

The stigmator values may differ from system to system. Other values and intervals may also be used. The computer starts the same ion and does the calculation step 203 and determination step 206 with the x-stigmator set to the second - 18 – value. In the comparison step 208, a second eigen-vector ratio at the second value is compared to the first eigenvector ratio at the first value of the x-stigmator setting.

If the first ratio is closer to one than the second ratio, then the computer stops the search in the x-direction. If the second ratio is closer to one than the first ratio then the computer, in the setting step 210, changes the value setting of the x-stigmator from the second value to a third value such as -0.04. This iteration continues with fourth, fifth etc. values until the computer determines that the next xvalue does not result in a rounder image i.e. when the new eigen-vector ratio is not closer to one compared to the vector ratio associated with the us x-stigmator value.

When the best x-stigmator value has been found then the computer, in a setting step 212, s the y-stigmator 213 setting from a first value, such as -0.12 as shown in Figs. 1-2, to a second value, such as -0.08. The computer does the same iteration as that was done for the x-stigmator setting until the best y-stigmator setting has been found.

However, the mator is fixed at the value as determined in the first iteration so the setting step 210 is skipped as illustrated by the dashed line 214 in Fig. 8.

Instead of merely adjusting the tor setting - 19 – in one ion at a time (such as the x-stigmator) it is also possible to adjust the stigmator settings in both the xand y-directions at the same time. This is illustrated in Fig. 2 wherein image 300 may be an image selected as a starting image. The system then investigates the eigen - vector ratio of not only image 302 (by ing the stigmator setting only in the x-direction) but also investigates the ratios as a result of stigmator settings of the images surrounding image 300 so that the eigen-vector ratios of images 304, 306, 308, 310, 312, 314 and 316 are determined. The system then selects the x- and y-stigmator settings that result in the lowest vector ratio, as explained above, which most likely is closer to the center ed to the ng image 300, as shown in Fig. 2. These steps are iterated, as outlined in Fig. 8, until the lowest ratio is found and the image is determined to show the most round image. This is the preferred x- and y-stigmator settings that should be used to best compensate for the astigmatism of the lens system. As best shown in Fig. 6, the size of the two eigen-vectors 132, 134 together with the angle alpha of the elliptical image 130 may thus be used to indicate which direction is likely going to be the best. As indicated earlier, it is also possible to simply continue in the same - 20 – direction, without calculating the ratios in all directions, as long as the ratios e i.e. get closer to 1. When the ratios start to increase again, the computer may then start to investigate all the directions, as described above.

While the present invention has been described in accordance with preferred compositions and embodiments, it is to be tood that certain substitutions and alterations may be made thereto without departing from the spirit and scope of the following claims. - 21 –

Claims (6)

We claim:

1. A method for automatic astigmatism tion of a lens system of an on microscope, comprising: 5 providing a first image of a view at an under-focus or over-focus at a first stigmator setting of a lens, based on the first image at the first stigmator setting, calculating a first r spectrum image, describing a distribution of intensities in the 10 first image by calculating, based on the first Fourier spectrum image, a first vector, representing a main direction of ities in said image which main direction is the direction with the longest elongation with bright pixels, the first vector having a length representing the elongation of 15 the bright pixels portion of the first image in said main direction and a second vector representing an orthogonal direction of intensities in said image, that is orthogonal to said main direction, and an extent of bright pixels in the orthogonal direction the second vector having a length 20 representing the tion of the bright pixels portion of the first image in the orthogonal direction, calculating a first ratio as a ratio between the length of the first vector and the length of the second - 22 – vector, changing the lens from the first stigmator setting to a second stigmator setting to provide a second image of the view at the under-focus or over-focus, the second image 5 at the second stigmator setting being of the same view as the first image of the view at the first stigmator setting, based on the second image and the second stigmator g, calculating a second Fourier spectrum image describing a distribution of intensities in the second image 10 by calculating, based on the second r spectrum image, a third vector representing a main direction of intensities in said image which main ion is the direction with the longest elongation with bright pixels, the third vector having a length representing the elongation of the bright 15 pixels portion of the second image in said main direction; and a fourth vector representing the orthogonal direction of intensities in said image and an extent of bright pixels in the orthogonal direction the fourth vector having a length enting the elongation of the bright pixels portion of 20 the second image in the orthogonal ion, calculating a second ratio as a ratio between the length of the third vector and the length of the fourth vector, when the first ratio is closer to 1 than the second - 23 – ratio selecting the first image of the view at the first stigmator setting, and when the second ratio is closer to 1 than the first ratio selecting the second image of the view at the second stigmator setting.

2. The method of claim 1 wherein the method further comprises the step of selecting an image having a lowest ratio. 10

3. The method of claim 1 or claim 2 wherein the method further comprises the step of changing x-stigmator and ystigmator settings.

4. The method of claim 3 wherein the method further 15 comprises the step of simultaneously changing the x-stigmator and the y-stigmator g.

5. The method of any one of claims 1 to 4 wherein the method further comprises searching for a stigmator g 20 that minimizes an elongation value of the intensity distribution of a ed Fourier spectrum image. - 24 –

6. A method for automatic astigmatism correction of a lens system as before described with reference to the drawings. WO 14177 ‘ ‘3. ii} .'r5w» W0