Image processing system, image processing method, and display device

Abstract

The present invention provides a technology for acquiring parameter values according to which a contour line close to a contour line that a user anticipates and expects is extracted.　A controller 102 comprises: an image display unit 1021 that displays an observation image; a contour extraction unit 1022 that extracts first contour information for dividing the structure of a sample from the observation image on the basis of a preset first parameter value; a first contour information display unit 1023 that displays the first contour information extracted by using the first parameter value on the displayed observation image; a second contour information accepting unit 1024 that accepts, on the observation image, input of second contour information indicating the contour of the structure included in the observation image; and a parameter value acquisition unit 1025 that acquires a second parameter value, different from the first parameter value, for extracting third contour information for which the difference from the second contour information in terms of display is a minimum.

Description

Image processing system, image processing method, and display device

[0001] The present disclosure relates to techniques such as an image processing system, an image processing method, and a display device.

[0002] Many of the observation images taken using charged particle beam devices are used for shape and structure analysis of samples. In order to perform these numerical processing, image processing to extract the contours of structures is almost essential. Therefore, a method for accurately extracting contours is important.

[0003] Algorithms for extracting contours in observed images can handle a wide variety of images, so their operation can be controlled by the numerical values ​​of variables called parameters. For example, one typical parameter is a threshold. Contour extraction algorithms often include a process for selecting pixels that will become part of a contour from among pixels in the observed image that show large variations in brightness. The threshold is used in the pixel selection process. Even when extracting contours from the same image, changing the threshold value will change the extracted contour. Conversely, to accurately extract contours in an image, the parameters included in the algorithm, including the threshold, must be set to optimal values.

[0004] If the optimization of the contour extraction algorithm or the parameter values ​​associated with this algorithm is automatically performed by computer calculation, it is expected to be more accurate and efficient than manual optimization. To achieve this, in many cases, a method is adopted in which numerical data that serves as the basis for calculating the optimal parameter values ​​is received as an input separate from the observed image, and an algorithm that extracts the desired contour line is selected or parameter values ​​are calculated. Prior art techniques can be found in Patent Documents 1 and 2.

[0005] JP 2008-116207 A JP 2012-21832 A

[0006] In research and development and quality assurance in various industrial fields, transmission electron microscopes (TEM), scanning transmission electron microscopes (STEM), scanning electron microscopes (SEM), and other instruments are often used to capture numerous observation images, and shape analyses such as length and angle measurements are performed on all of the captured observation images. These shape analyses require the extraction of contours of structures within the observation images. Therefore, prior to shape analysis, users must optimize the settings of the contour extraction algorithm, particularly the parameter values, to suit the observation images.

[0007] However, for a given observation image, there are few cases where there is a clear relationship between parameter values ​​and the extracted contours, and no systematic method for adjusting parameter values ​​exists. Therefore, users must search for optimal parameter values ​​for the contour extraction algorithm by trial and error for each observation image. As mentioned above, when the number of observation images is large, the workload increases in proportion to the number of observation images, and this increased workload represents a significant cost for manufacturers. Furthermore, for semiconductor device manufacturers, for example, as the structures of semiconductor devices become increasingly miniaturized and complex, the shapes of the contours to be extracted also become more complex, making it more difficult to adjust optimal parameter values.

[0008] To address this issue, techniques for automatically optimizing contour extraction algorithms have been proposed, such as those described in Patent Documents 1 and 2. It should be noted that the method described in Patent Document 1 is applicable only when the sample manufacturing process is known. Therefore, it is difficult to apply it to naturally occurring samples, such as biological samples, which do not have a manufacturing process. Furthermore, the method described in Patent Document 2 automatically determines parameter values ​​that minimize the influence of noise in the observed image on contour extraction or the roughness of the contour, but does not guarantee the extraction of contours that the user expects, including rough contours. Therefore, there is a need for a method that can automatically obtain algorithm parameter values ​​that can extract contours that satisfy the user's expectations.

[0009] An object of the present disclosure is to provide a technology that can acquire parameter values ​​that extract a contour that is closer to the contour that the user envisions and expects.

[0010] A representative embodiment of the present disclosure has the following configuration. An image processing system according to one embodiment processes an observation image of a sample obtained by a charged particle beam device. The image processing system includes an image display unit that displays the observation image, a contour extraction unit that extracts first contour information relating to a structural boundary of the sample from the observation image based on a first set value of a preset parameter, a first contour information display unit that displays the first contour information extracted by the contour extraction unit on the observation image, a second contour information receiving unit that receives input of second contour information relating to a boundary visible in the observation image, and a parameter value acquisition unit that acquires a second set value different from the first set value so as to minimize a difference between the first contour information and the second contour information in the observation image.

[0011] An image processing method executed by an image processing system that processes an observation image of a sample obtained by a charged particle beam device of one embodiment, wherein the image processing system displays the observation image, extracts first contour information regarding a structural boundary of the sample from the observation image based on a first setting value of a preset parameter, displays the extracted first contour information on the observation image, accepts input of second contour information regarding a boundary visible in the observation image, and obtains a second setting value different from the first setting value so as to minimize a difference between the first contour information and the second contour information in the observation image.

[0012] In one embodiment, a display device is used in an image processing system that processes an observation image of a sample obtained by a charged particle beam device, the display device displays, on the observation image, first contour information relating to a structural boundary of the sample extracted from the observation image based on a first setting value of a preset parameter, accepts input of second contour information relating to a boundary visible in the observation image, obtains a second setting value different from the first setting value so as to minimize a difference between the first contour information and the second contour information in the observation image, and displays, on the observation image, third contour information extracted based on the obtained second setting value.

[0013] According to a representative embodiment of the present disclosure, it is possible to provide a technique for acquiring parameter values ​​that extract a contour line that is closer to the contour line that a user expects and desires. Problems, configurations, effects, etc. other than those described above will be described in the description of the embodiment.

[0014] 1 is a diagram illustrating an example of the configuration of a charged particle beam system according to a first embodiment; FIG. 2 is a diagram illustrating an example of functional blocks of a controller according to the first embodiment; FIG. 3 is a flowchart illustrating a processing flow of the functional blocks according to the first embodiment; FIG. 4 is a diagram illustrating an example of a sample cross-sectional image according to the first embodiment; FIG. 5 is a diagram illustrating an example of a user interface according to the first embodiment; FIG. 6 is a diagram illustrating an example of a reference line according to the first embodiment; FIG. 7 is a diagram illustrating an example of a line profile and a contour line according to the first embodiment; FIG. 8 is a diagram for explaining an example of acquisition of a horizontal line profile according to the first embodiment; FIG. 9 is a diagram illustrating an example of a horizontal line profile according to the first embodiment; FIG. 10 is a diagram illustrating an example of a radial line profile according to the first embodiment; FIG. 11 is a diagram illustrating an example of an initial contour calculated using initial setting values ​​of two parameters according to the first embodiment; FIG. 12 is a diagram illustrating an example of a range in which movement of a contour point is restricted according to the first embodiment; FIG. 13 is a diagram illustrating an example of a correct contour line according to the first embodiment; FIG. 14 is a diagram illustrating an example of a corrected contour line corrected based on the correct contour line according to the first embodiment; FIG. 15 is a diagram illustrating an example of a display during execution of an optimization algorithm according to the first embodiment; FIG. 16 is a diagram illustrating an example of a sample cross-sectional image according to the second embodiment; FIG. 17 is a diagram illustrating an example of contour points detected by the contour point detection algorithm according to the second embodiment; FIG. 18 is a diagram illustrating an example of a display of corrected contour points according to the second embodiment. 10 is a diagram showing an example of a display on the display unit while the optimization algorithm of the second embodiment is being executed. FIG. 11 is a diagram showing an example of a top view image of the second embodiment. FIG. 12 is a diagram showing an example of initial contour points and correct contour points extracted from a top view image of the second embodiment. FIG. 13 is a diagram showing an example of modified contour points that are closest to the correct contour points of the second embodiment. FIG. 14 is a diagram showing an example of a particle sample image obtained when observing and analyzing a particulate sample of the third embodiment. FIG. 15 is a diagram showing an example of a particle sample image after separation processing of the third embodiment. FIG. 16 is a diagram showing an example of a particle sample image after noise removal processing of the third embodiment. FIG. 17 is a diagram showing an example of a particle sample image after extraction processing of a "certain background region" of the third embodiment. FIG. 18 is a diagram showing an example of a particle sample image after extraction processing of a "certain particle region" of the third embodiment. FIG. 19 is a diagram showing an example of a particle sample image after image processing that does not require setting of parameter values ​​of the third embodiment.10A and 10B are diagrams illustrating an example of a particle sample image according to the third embodiment; FIG. 10A are diagrams illustrating an example of a particle sample image according to the third embodiment; FIG. 10B are diagrams illustrating an example of a display on a display unit during execution of an optimization algorithm according to the third embodiment;

[0015] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings. In the drawings, the same parts are generally designated by the same reference numerals, and repeated explanations will be omitted. In the drawings, the representation of components may not represent their actual positions, sizes, shapes, ranges, etc., in order to facilitate understanding of the invention.

[0016] For the purpose of explanation, when describing processing by a program, the program, functions, processing units, etc. may be described as the main components, but the main hardware components of these are the processor, or a controller, device, computer system, system, etc., configured with the processor, etc. A computer system executes processing according to a program read into memory using resources such as memory and communication interfaces as appropriate through the processor. This realizes predetermined functions, processing units, etc. The processor is configured, for example, with semiconductor devices such as a CPU / MPU or GPU. Processing is not limited to software program processing, and can also be implemented using dedicated circuits. Dedicated circuits such as FPGAs, ASICs, and CPLDs can be used.

[0017] The program may be pre-installed as data on the target computer system, or may be distributed as data to the target computer system from a program source. The program source may be a program distribution server on a communication network, or a non-transitory computer-readable storage medium, such as a memory card or disk. The program may be composed of multiple modules. The computer system may be composed of multiple devices. The computer system may be composed of a client-server system, a cloud computing system, an IoT system, etc. Various data and information may be composed of structures such as, for example, tables and lists, but are not limited to these. Expressions such as identification information, identifiers, IDs, names, and numbers are interchangeable.

[0018] In the following, as a technique of this embodiment, for example, a case where this technique is applied to contour extraction of a cross-sectional image of a sample such as a semiconductor device will be described (first embodiment). After that, as an applied example, it will be shown that this technique can also be applied to contour points, which are minute elements of a contour (second embodiment). Finally, it will be shown an example where this technique is applied to a case where a contour is a boundary line of an area in an observation image (third embodiment).

[0019] (First embodiment) A well-known method of observing a sample, for example, a semiconductor device, using a charged particle beam device is so-called top-view observation and measurement, in which the wafer surface is viewed from above using a scanning electron microscope, for example, a critical dimension scanning electron microscope (CD-SEM). On the other hand, semiconductor devices are three-dimensional structures, and observation of the wafer surface from the side, i.e., cross-sectional observation and measurement, also plays an important role in the device development and quality assurance processes. For this reason, the embodiment mainly focuses on image processing techniques for observing and measuring the cross sections of semiconductor devices.

[0020] <Configuration of Scanning Electron Microscope> FIG. 1 is a diagram showing an example of the configuration of a charged particle beam device (hereinafter referred to as "SEM") 100. The SEM 100 roughly comprises an SEM main body 101 and a controller 102 connected to the main body 101. The main body 101 further comprises an electron optical column (hereinafter referred to as "column") and a sample chamber provided below the column. The controller 102 is a system that controls imaging by the main body 101, etc. The controller 102 comprises an overall control unit 120, a signal processing unit 121, a storage unit 122, a communication interface 123, etc. An input device 124 and an output device 125 are externally connected to the controller 102.

[0021] The column of the main body 101 includes, as its components, an electron gun 111, a focusing lens 112, a deflection lens 113, and an objective lens 114. The electron gun 111 emits an electron beam b1, which is a charged particle beam. The focusing lens 112 focuses the electron beam b1. The deflection lens 113 deflects the trajectory of the electron beam b1. The objective lens 114 controls the focusing height of the electron beam b1. The sample chamber is a room in which samples such as wafers and coupons (broken pieces of wafers) are stored. The sample chamber includes a stage 115 and a detector 116. A sample 130 is placed on the stage 115.

[0022] The stage 115 is a sample stage on which a semiconductor device, which is the target sample 130, is placed. The stage 115 has the ability to move not only in the X and Y directions but also in the Z direction and to rotate about the XY, YX, or Z axes. Therefore, the stage 115 can move the captured field of view in the horizontal and vertical directions relative to the front-facing image, or in a rotational direction within the field of view. This allows the field of view for imaging to be set.

[0023] The detector 116 detects, as an electric signal, particles b2 such as secondary electrons and backscattered electrons generated from the sample 130 irradiated with the electron beam b1, and outputs a detection signal that is an electric signal.

[0024] The overall control unit 120 controls the operation of the controller 102 and the main body 101. The overall control unit 120 issues instructions such as drive control to each unit. Each unit, such as the overall control unit 120, can be implemented as a computer or a dedicated circuit. The signal processing unit 121 receives a detection signal from the detector 116, performs processing such as analog / digital conversion to generate an image signal, and stores the image data in the memory unit 122. The memory unit 122 can be implemented as a non-volatile storage device. The overall control unit 120 also stores image-related shooting information and the like in association with the memory unit 122. The communication interface 123 is a device that has implemented a communication interface for a communication network (not shown) or an external computer (not shown).

[0025] The overall control unit 120 outputs data such as images and photographing information stored in the storage unit 122 to the output device 125 in response to a request from the input device 124, for example. The overall control unit 120 may receive a request from an external computer via the communication interface 123, and transmit data such as images and photographing information stored in the storage unit 122 to the external computer in response to the received request.

[0026] <Functional Blocks> FIG. 2 is a diagram showing an example of functional blocks of the controller 102. As shown in FIG. 2, the controller 102 has an image display unit 1021, a contour extraction unit 1022, a first contour information display unit 1023, a second contour information receiving unit 1024, a parameter value acquisition unit 1025, and a third contour information display unit 1026. The overall control unit 120 reads out a program stored in the storage unit 122, for example, and executes the read program, thereby causing the controller 102 to realize the functions of the image display unit 1021, the contour extraction unit 1022, the first contour information display unit 1023, the second contour information receiving unit 1024, the parameter value acquisition unit 1025, and the third contour information display unit 1026. By realizing these functions, the controller 102 operates as an image processing system that performs image processing of contours in an observation image of the sample 130 obtained by the SEM 100. In this embodiment, the controller 102 will be described as being connected to the SEM 10, but the present invention is not limited to this. For example, the controller 102 may be configured independent of the SEM. In this case, the controller 102 may have an image reading function for reading an image acquired by an external SEM or the like, and may be configured to implement the functions shown in FIG. 2 for the image read by this image reading function.

[0027] The image display unit 1021 displays an observation image of a sample captured by a charged particle beam device. For example, the image display unit 1021 displays a sample cross-sectional image G10 (described later) of the sample 130 captured by the SEM 100.

[0028] The contour extraction unit 1022 extracts first contour information relating to the structural boundary of the sample from the observation image based on first set values ​​of preset parameters. The contour extraction unit 1022 includes set values ​​of multiple parameters. In this embodiment, the contour extraction unit 1022 extracts the first contour information from the observation image based on the set values ​​of multiple parameters. For example, as described below, the contour extraction unit 1022 extracts an initial contour 20 (first contour information) relating to the structural boundary of the sample 130 from the sample cross-sectional image G10 based on preset parameters, such as initial set values ​​(first set values) of smooth and threshold. There are multiple types of parameters, such as smooth and threshold.

[0029] The first contour information display unit 1023 displays on the observation image the first contour information extracted by the contour extraction unit 1022. For example, as will be described later, the first contour information display unit 1023 displays an initial contour line 20 extracted using parameter values, for example, initial setting values ​​of smooth and threshold, on the sample cross-sectional image G10 displayed on the image display unit 1021.

[0030] The second contour information receiving unit 1024 receives input of second contour information relating to a boundary visible in the observation image. For example, the second contour line receiving unit 1024 receives input of a correct contour line 21 relating to a boundary of a pillar 11 visible in a sample cross-sectional image G10 (described later) displayed on the image display unit 1021, as described later.

[0031] The parameter value acquisition unit 1025 acquires second setting values ​​that are different from the first setting values ​​so as to minimize the difference between the first contour information and the second contour information in the observed image. For example, as will be described later, the parameter value acquisition unit 1025 acquires setting values ​​(second setting values) of parameters that are different from the initial setting values ​​of the parameters so as to minimize the difference between the initial contour line 20 and the correct contour line 21.

[0032] The third contour information display unit 1026 displays, on the observed image, the third contour information extracted by the contour extraction unit 1022 based on the second setting value acquired by the parameter value acquisition unit 1025. For example, the third contour information display unit 1026 displays the corrected contour line 22 extracted by the contour extraction unit 1022 based on the parameter value acquired by the parameter value acquisition unit 1025, the smooth and threshold setting values, as will be described later.

[0033] <Flow of Image Processing> Fig. 3 is a flowchart showing the flow of processing of the functional blocks. As shown in Fig. 3, the image processing of this embodiment is performed in the following order: Step ST110: image display processing by the image display unit 1021 Step ST120: contour extraction processing by the contour extraction unit 1022 Step ST130: first contour information display processing by the first contour information display unit 1023 Step ST140: second contour information reception processing by the second contour information reception unit 1024 Step ST150: parameter value acquisition processing by the parameter value acquisition unit 1025 Step ST160: third contour information display processing by the third contour information display unit 1026

[0034] The processing executed by the controller 102, which is an image processing system, will be described in more detail below. The image display unit 1021 operates the SEM main body 101 to observe a cross section of a sample (semiconductor device) 130 and acquires a sample cross-sectional image with sufficient image quality for the purpose of analysis. Fig. 4 is a diagram showing an example of a sample cross-sectional image G10. In this embodiment, the sample cross-sectional image (observation image) G10 is acquired by the SEM 100, but it may also be acquired by a TEM or STEM.

[0035] As shown in Figure 4, the sample cross-sectional image G10 includes a protrusion structure (hereinafter referred to as a "pillar") 11 and a valley structure (hereinafter referred to as a "trench") 12 between two pillars. Note that in the sample cross-sectional image G10, a white area is provided as the background of the area where the symbols and lead lines are added, but the white area behind the symbols and lead lines is not an image. Hereinafter, with regard to the symbols and lead lines in the image display, a white area will be provided and the symbols and lead lines will be added.

[0036] The image file format of the sample cross-section image G10 is not limited to a specific image file format. However, the harmonic components contained in the image data are important in the contour extraction process of the contour extraction unit 1022. For this reason, the image file format should preferably be an uncompressed image file format such as Tagged Image File Format (TIFF format) or Microsoft Windows Bitmap Image (BMP format).

[0037] Next, the image display unit 1021 loads the sample cross-sectional image G10 shown in Fig. 4 into the shape analysis software. For this purpose, the shape analysis software has a user interface (UI) for loading the image format output by the SEM 100. Fig. 5 is a diagram showing an example of the user interface. In this embodiment, a graphical user interface (GUI: display device) 1251 is used as the user interface.

[0038] As shown in FIG. 5 , the GUI 1251 has buttons, including an import image button 1251A, a detect button 1251B, an edit button 1251C, and a setting adjustment button 1251D. The import image button 1251A is a button that instructs the shape analysis software to load the sample cross-sectional image G10 and display the observed image. The detect button 1251B is a button that instructs the software to extract structural units, in other words, the contours of structural boundaries, from the sample cross-sectional image G10. The edit button 1251C is a button that instructs the second contour information receiving unit 1024 to accept editing of the contours. The setting adjustment button 1251D is a button that instructs the parameter value acquisition unit 1025 to adjust the parameter value settings. The GUI 1251 also has a display unit 1251E. On the display unit 1251E, under the control of the image display unit 1021, for example, a sample cross-sectional image G10 that has been read into shape analysis software is displayed.

[0039] In this embodiment, the analysis of the sample cross-sectional image G10 is performed using the pillars 11 and trenches 12 shown in FIG. 4 as structural units (hereinafter referred to as "unit structures"). A contour extraction process using an algorithm specialized for contour extraction around these unit structures will be described below. The contour extraction process is executed by the contour extraction unit 1022.

[0040] The contour extraction unit 1022 sets a reference line in the horizontal direction of the image for the sample cross-sectional image G10 shown in Fig. 4. Fig. 6 is a diagram showing an example of the reference line 13. The contour extraction unit 1022 extracts the contour lines of the unit structures relative to the reference line 13, in an upper portion 14 above the reference line 13 for the pillars 11, and in a lower portion 15 below the reference line 13 for the trenches 12.

[0041] Next, the contour extraction unit 1022 sets reference points 16 near the tip ends of the pillars 11 or trenches 12 whose contours are to be extracted, as shown in Fig. 6. Any number of reference points 16 may be used. By setting the reference lines 13 and the reference points 16, the contour extraction unit 1022 can appropriately set line profiles 17 and 18 (described later), thereby enabling appropriate contour extraction of structures such as pillars 11 and trenches 12 from the sample cross-sectional image G10.

[0042] The following continues the description of the process of extracting the contour around the pillar 11. After setting the reference line 13 and the reference point 16, the contour extraction unit 1022 acquires a line profile of pixel values ​​at a certain width along the horizontal direction of the image if the area is below the reference point 16, or along the radial direction around the reference point 16 if the area is above the reference point 16.

[0043] 7 is a diagram showing an example of a line profile and a contour line. A line profile 17 in the horizontal direction of the image and a line profile 18 in the radial direction are shown in FIG. 7. For example, the contour extraction unit 1022 acquires horizontal line profiles 17 in one-pixel increments below the reference point 16, and acquires radial line profiles 18 in one-degree increments above the reference point 16. A contour line is formed by a plurality of contour points 19.

[0044] Here, a description will be given of the process in which the contour extraction unit 1022 acquires a line profile 17 on a horizontal line below the reference point 16. Fig. 8 is a diagram for explaining an example of acquiring a line profile 17 in the horizontal direction.

[0045] The (x, y) coordinate system in Fig. 8 corresponds to the pixel coordinates of the sample cross-sectional image G10. The pixel value at pixel coordinates (x, y) is represented as I(x, y). The horizontal line segment y = y ０ (x ｉ <x<x ｔ In this case, the outline extraction unit 1022 extracts a line profile 17 having a width of 2y in the direction perpendicular to the horizontal line segment in order to remove noise. ｓ The average pixel value is calculated within the pixel range and used as the profile value. ０ (x ｉ <x<x ｔ ) on the point (x, y ０ ) Profile value I Ｐ (x), then I Ｐ (x) is The number of pixels y determines the range of this average calculation. ｓ is required to be set as a parameter. ｓ is referred to as "smooth." Smooth is a parameter for removing noise from an image. By setting the smooth value, the contour extraction unit 1022 can remove desired noise from, for example, the sample cross-sectional image G10.

[0046] 9 is a diagram showing an example of a horizontal line profile 17. In this line profile 17, the contour extraction unit 1022 extracts the maximum profile value IP ＭＡＸ and the minimum value IP ＭＩＮ The minimum value is 0% and the maximum value is 100%, and the point having a profile value corresponding to T% (0<T<100) is detected as an edge point. ｅ When I Ｐ (x ｅ ) satisfies the following equation: As with smooth, the edge point x ｅ It is also necessary to set the parameter x ｅ The threshold value of the parameter is hereinafter referred to as "threshold". The threshold is a parameter that determines the threshold value for distinguishing between image regions. By setting the threshold value, the contour extraction unit 1022 can distinguish, for example, the region of the pillar 11 from the region other than the pillar 11 in the sample cross-sectional image G10. The contour extraction unit 1022 performs the above process on all horizontal line segments y=y below the reference point 16. ０ By repeating this, the edge point x of the line profile 17 below the reference line 13 is obtained. ｅ can be obtained.

[0047] The contour extraction unit 1022 extracts the edge point x above the reference point 16. ｅ The detection of the line profile 18 in the radial direction is performed in the same manner as the process below the reference point 16. Fig. 10 is a diagram showing an example of a line profile 18 in the radial direction.

[0048] As shown in FIG. 10, in the case of the area above the reference point 16, the contour extraction unit 1022 extracts the line profile 18 by taking the radial line segment (centered on the reference point 16) as the center and calculating 2θ ｓ Noise removal is performed by averaging pixel values ​​within the range of θ ｓ It may be itself, or θ ｓ It may be a constant multiple of .

[0049] Through the above process, the contour extraction unit 1022 can obtain contour points 19 of each of the line profiles 17, 18 above and below the reference point 16. The first contour information display unit 1023 can display a group of contour points 19 obtained in this manner, as shown in the above-mentioned Fig. 7. In Fig. 7, a plurality of contour points 19 are displayed so as to generally follow the shape of the pillar 11.

[0050] Two parameters (smooth and threshold) used in the contour extraction process are adjusted to optimal values ​​by applying the technology of this embodiment. Initial settings for smooth and threshold are determined in the shape analysis software. Therefore, the contour extraction unit 1022 first executes contour extraction using the initial settings for the two parameters, and extracts an initial contour line (first contour information) formed by multiple contour points 19.

[0051] Next, the first contour information display unit 1023 displays the extracted initial contour on the sample cross-sectional image G10 displayed on the display unit 1251E of the GUI 1251 shown in Fig. 5. Fig. 11 is a diagram showing an example of an initial contour 20 calculated using the initial setting values ​​of two parameters. The display in Fig. 11 is displayed, for example, when the Detect button 1251B is pressed.

[0052] As shown in FIG. 11 , if the initial contour line 20 displayed on the display unit 1251E differs from the shape of the pillar 11 visible to the user, in other words, if the initial contour line 20 differs from the contour line assumed or expected by the user, the user corrects the initial contour line 20 by operating the input device 124. The input device 124 is, for example, a mouse. The user can shape the displayed initial contour line 20 into a desired contour line by using the mouse pointer as a pen tip to push out the initial contour line 20 displayed on the display unit 1251E. In this example, because the initial contour line 20 deviates from the shape of the pillar 11, the user uses the mouse to shape the initial contour line 20 so that it matches the shape of the pillar 11, thereby creating a correct contour line 21 (second contour information) to be displayed in FIG. 13 (described later).

[0053] The shaping of the initial contour line 20 is achieved, for example, by moving each of the contour points 19 that make up the initial contour line 20. The movement of each contour point 19 is restricted within a certain range. FIG. 12 shows an example of the range within which the movement of a contour point is restricted. The display in FIG. 12 is displayed, for example, by pressing edit button 1251C and selecting a contour point 19. In FIG. 12, the double-headed arrows AW1 and AW2 indicate the range within which the movement of the contour point 19 is restricted. The double-headed arrow AW1 indicates the range within which the contour point 19 obtained from the horizontal line profile 17 can be moved. The double-headed arrow AW2 indicates the range within which the contour point 19 obtained from the radial line profile 18 can be moved.

[0054] The double-headed arrow AW1 extends horizontally from the reference point 16 with the contour point 191 at its center. The contour point 191 is one of multiple contour points 19 located below the reference point 16. One end of the double-headed arrow AW1 extends a distance L1 to a line extending vertically from the reference point 16. The other end of the double-headed arrow AW1 extends a distance L1 from the reference point 16 on the opposite side to the one end. The contour point 191, which is below the reference point 16 and to the left of the reference point 16 as viewed from the reference point 16, cannot be moved vertically, and cannot be moved horizontally from the position of the contour point 191 shown in FIG. 12 to the right of the reference point 16.

[0055] The double-headed arrow AW2 extends in the radial direction from the reference point 16, centered on the contour point 192. The contour point 192 is one of multiple contour points 19 located above the reference point 16. One end of the double-headed arrow AW2 extends a distance L2 to the reference point 16. The other end of the double-headed arrow AW2 extends a distance L2 from the reference point 16 in the opposite direction to the one end. The contour point 192 above the reference point 16 can only move along the straight line connecting the position shown in FIG. 12 and the contour point 192; it cannot be moved below the reference point 16 from the position of the contour point 192 shown in FIG. 12 .

[0056] The double-headed arrows AW1 and AW2 and the portions NG1 and NG2 allow the user to visually recognize the range in which the contour points 191 and 192 can be moved. Furthermore, by limiting the range in which the contour points 191 and 192 can be moved, it is possible to prevent the user from inputting incorrect contour points 191 and 192. Here, inputting incorrect contour points 191 and 192 refers to, for example, moving contour points 191 and 192 to positions that are clearly unrelated to the contour of a structure such as pillar 11. Because inputting incorrect movement positions can be prevented, the contour extraction unit 1022 can normally execute the contour extraction process.

[0057] The user uses a mouse to input a correct contour line within the limited range illustrated by the double-headed arrows AW1 and AW2 in FIG. 12 . The user modifies the initial contour line 20, for example, so that it follows the shape of the pillar 11. The correct contour line is stored, for example, as information indicating the positions of the multiple contour points 19 after each of the multiple contour points 19 has been moved. FIG. 13 is a diagram showing an example of the correct contour line 21. In FIG. 13 , the correct contour line 21 is displayed, following the shape of the pillar 11. The second contour information receiving unit 1024 stores the multiple contour points 19 that form the received correct contour line 21, for example, in a predetermined memory.

[0058] Next, when the user presses the setting adjustment button 1251D, the parameter value acquisition unit 1025 executes a parameter optimization algorithm. More specifically, the parameter value acquisition unit 1025 changes the smooth and threshold values ​​from their initial settings, repeats the process of calculating the loss, and finally acquires the smooth value and threshold setting values ​​that minimize the loss. Here, the smooth value and threshold value that minimize the loss are values ​​that, when the contour extraction unit 1022 extracts a contour, extracts a contour whose display difference is closest to the correct contour 21.

[0059] After the parameter values ​​are acquired by the parameter value acquisition unit 1025 in this manner, the contour extraction unit 1022 extracts the contour again, and the third contour information display unit 1026 displays the corrected contour. Fig. 14 is a diagram showing an example of a corrected contour 22 (third contour information) corrected based on the corrected contour 21. Fig. 14 displays the corrected contour 22 that follows the shape of the pillar 11.

[0060] The procedure for setting a series of parameter values ​​in the processing of the parameter value acquisition unit 1025 differs depending on the optimization algorithm used. The optimization algorithm may be, for example, a gradient method represented by the steepest descent method, a method using random numbers such as simulated annealing, or a statistical method such as Bayesian optimization. Here, the initial contour 20 calculated by the algorithm in the processing of the contour extraction unit 1022 described above is a point group made up of contour points 19 extracted on a specific line profile 17, as shown in FIG. 7. Therefore, the loss is the sum of squares of the distances between corresponding contour points on two contours: and the maximum distance between corresponding contour points: etc., where x ｉ (t), y ｉ (t) is the coordinate of the i-th contour point in the image at a certain time t during the execution of the optimization algorithm, and x ｉ Ｒ , y ｉ Ｒ is the image coordinate of the i-th contour point of the correct contour 21.

[0061] FIG. 15 is a diagram showing an example of the display on the display unit 1251E while the optimization algorithm is running. In FIG. 15, in addition to the sample cross-sectional image G10, a loss calculation display G15 that displays a loss calculation is displayed. The loss calculation display G15 displays the calculation results of the loss for the parameters. The loss calculation display G15 displays a graph showing the calculation results of the loss result. The graph showing the calculation results of the loss result does not have to be displayed. Furthermore, depending on the optimization algorithm, if not all losses are calculated, a smooth curve such as the solid line in FIG. 15 is not drawn, and the loss value calculated at each time t may be dotted as shown by A' in FIG. 15, etc.

[0062] As shown in FIG. 15, the parameter value acquisition unit 1025 calculates L shown in the above-mentioned formula (3) from the initial contour 20 (consisting of A and the like) and the correct contour 21 at a certain time t during the execution of the optimization algorithm. ２ , and L shown in formula (4) Ｍ The loss is calculated using the above. Display A is one of the contour points 19 that form the initial contour line 20. The loss is displayed, for example, in the loss calculation display G15 as a rectangular display A' on a graph showing the loss results.

[0063] The parameter value acquisition unit 1025 updates the parameter values ​​according to the optimization algorithm, and similarly calculates the loss for the contour (consisting of B and others shown in the figure) detected one step later at time t+1. Point B is one of the contour points 19 that form the correct contour 21. The loss is displayed, for example, as point B' in the loss calculation display G15.

[0064] After updating the parameter values, the parameter value acquisition unit 1025 repeats the process of calculating the loss. By repeating the process of calculating the loss a sufficient number of times, the parameter value acquisition unit 1025 can obtain the parameter value that minimizes the loss. For example, in the loss calculation display G25, the parameter value that minimizes the loss is displayed as point O'. The contour extracted by the contour extraction unit 1022 using the parameter value that minimizes the loss becomes the corrected contour 22 that is closer to the correct contour 21.

[0065] In cases other than those described above, automatic optimization of parameter values ​​is possible using a similar procedure in contour extraction on a sample cross-sectional image G10 obtained from a charged particle beam device such as SEM 100. However, for contour extraction algorithms other than the algorithm used in the contour extraction process described above, there may be cases where it is not possible to associate contour points between two given contours. In this case, L shown in equation (3) ２ , and L shown in formula (4) ＭTherefore, when the contour points 19 cannot be associated, the parameter value acquisition unit 1025 creates a binary array for each of the initial contour 20 and the correct contour 21, in which the pixel values ​​of the pixels that make up the contour in the sample cross-sectional image G10 are set to 1 and the other pixel values ​​are set to 0, and calculates the loss L by subtracting the product of the elements of the created array. Ｒ of Here, (x, y) are the coordinates in the sample cross-sectional image G10, W and H are the width and height of the image (both in number of pixels), and I Ｂ (x, y; t) is the element value corresponding to the (x, y) coordinates in the image of the binary array created from the initial contour 20 at a certain time t during the execution of the optimization algorithm, I Ｂ Ｒ (x, y) is an element value corresponding to the (x, y) coordinates in the image of the binary array created from the correct contour 21 .

[0066] According to this embodiment, the controller 102 can acquire parameter setting values ​​that minimize the difference between the corrected contour 21 assumed and expected by the user, in other words, that extract a corrected contour 22 that is closest to the corrected contour 21. This technique for acquiring parameter setting values ​​can be applied to any type of observation image and any shape of contours captured by a charged particle beam device such as the SEM 100. The controller 102 can also display the corrected contour 22 extracted using the acquired parameter values ​​on the display unit 1251E. This allows the user to visually confirm how close the corrected contour 22 is to the corrected contour 21 assumed and expected.

[0067] Second Embodiment In the contour extraction process for the sample cross-sectional image G10 of the sample 130 described in the first embodiment, there are cases where it is not necessary to extract all contours. In the second embodiment, processing by the image processing system for such cases will be described.

[0068] <Example of Application in Point-to-Point Measurement of Sample Cross-Section Image> Fig. 16 is a diagram showing an example of a sample cross-section image (observation image) G20. As shown in Fig. 16, when measuring the distance AW3 from a contour point 193 of one pillar 11 to a contour point 194 of another pillar 11, and when the positions of the two contour points to be measured are approximately determined, the contour extraction unit 1022 does not need to extract all contour points. Instead, it is sufficient for the contour extraction unit 1022 to extract one contour point 193, 194 for each pillar 11 within a specified small region in the sample cross-section image G20.

[0069] In this case, the simplest algorithm for the contour extraction unit 1022 to extract contour points 19 is to apply smoothing filter processing to the pixels of the small region. Here, smoothing filter processing is a filter processing for blurring and smoothing an image. By applying smoothing filter processing to the small region, the contour extraction unit 1022 can smooth the image of the small region. After this filter processing, the contour extraction unit 1022 detects the pixel with the highest pixel value as a contour point 193 (or contour point 194). The image I obtained by applying smoothing filter processing to the image I(x, y) is Ｓ (x, y) is That is, I Ｓ (x, y) is the original image I Ｓ For each pixel (x, y), this is an image obtained by averaging pixel values ​​in a pixel region of area K × K centered on the pixel. K used here is a positive odd number and is a parameter called kernel size. This parameter value can be adjusted by the parameter value acquisition unit 1025 using the same optimization algorithm as in the first embodiment.

[0070] FIG. 17 is a diagram showing an example of contour points detected by the contour point detection algorithm. FIG. 17 shows initial contour points 231 and 232 (first contour information) for each pillar 11. The user can modify the positions of the initial contour points 231 and 232 extracted by the contour point detection algorithm by operating the mouse as described in the first embodiment. In this case, the range within which the initial contour points 231 and 232 can be moved may be limited.

[0071] The initial contour points 231, 232, the positions of which have been corrected by the user, are designated as correct contour points 241, 242 (second contour information). The parameter value acquisition unit 1025 minimizes the sum of squares of the distances between the initial contour points 231, 232 and the correct contour points 241, 242 for all small regions as a loss, thereby obtaining optimal parameter values, i.e., kernel size values, for detecting correct contour points that minimize the correct contour points 241, 242. Based on the parameter settings thus obtained, the contour extraction unit 1022 extracts contours, and the third contour information display unit 1026 displays the correct contour points, which are the extraction results, on the display unit 1251E.

[0072] 18 is a diagram showing an example of the display of corrected contour points 251 and 252 (third contour information). In FIG. 18, the corrected contour points 251 and 252 are displayed near the correct contour points 241 and 241.

[0073] FIG. 19 is a diagram showing an example of the display on the display unit 1251E while the optimization algorithm is being executed. In FIG. 19, in addition to the sample cross-sectional image G20, a loss calculation display G25 that displays a loss calculation is displayed. The loss calculation display G25 displays the calculation results of the loss for the parameters. The loss calculation display G25 displays a graph showing the calculation results of the loss result. The parameter value that minimizes the loss is displayed as a point O' on the graph. The processing order of this optimization algorithm is the same as that explained using FIG. 15 in the first embodiment. The graph showing the calculation results of the loss result does not have to be displayed, and may be dotted, as explained using FIG. 15.

[0074] As described above, according to this embodiment, when measuring the distance from contour point 193 of one pillar 11 to contour point 194 of another pillar 11, and when the positions of the two contour points to be measured are substantially determined, controller 102 can obtain parameter values ​​with the least loss without extracting all contour points. This reduces the image processing load on controller 102.

[0075] <Image Processing in Top-View Image> The extraction process of corrected contour points 251, 252 described above is effective not only for the sample cross-sectional image G20, but also for a top-view image of the sample 130. FIG. 20 is a diagram showing an example of a top-view image (observation image) G30 of the sample 130. The sample 130 is a semiconductor device. FIG. 21 is a diagram showing an example of initial contour points 261, 262 (first contour information) and correct contour points 271, 272 (second contour information) extracted from the top-view image G30. FIG. 22 is a diagram showing an example of corrected contour points 281, 282 (third contour information) that minimize the display difference from the correct contour points.

[0076] 20 , which is provided by the processing of image display unit 1021, contour extraction unit 1022 also specifies the positions of two points to be measured, detects the pixel with the highest brightness after smoothing filter processing, and extracts initial contour points 261 and 262. First contour information display unit 1023 displays initial contour points 261 and 262 in top-view image G30, as shown in FIG.

[0077] The user operates a mouse or the like to correct the positions of initial contour points 261, 262 and input correct contour points 271, 272. Second contour information accepting unit 1024 then accepts the positions of correct contour points 271, 272. Next, parameter value acquiring unit 1025 executes an optimization algorithm to acquire parameter settings that minimize the difference between the initial contour points 261, 262 and correct contour points 271, 272. Based on these acquired settings, contour extraction unit 1022 extracts corrected contour points 281, 282. Third contour information display unit 1026 displays corrected contour points 281, 282 in top-view image G30, as shown in FIG. 22 . This allows the user to visually recognize corrected contour points 281, 282 that are closest to correct contour points 271, 272.

[0078] Third Embodiment In the third embodiment, image processing by the controller 102 when expected values ​​of geometric feature amounts of contour lines are given in advance will be described.

[0079] <Particle Sample Image> Figure 23 is a diagram showing an example of a particle sample image (observation image) G40 obtained when observing and analyzing a particulate sample using SEM 100. A region extraction algorithm can be used to extract the boundaries, i.e., contours, of individual particles in particle sample image G40 of such sample 130. As shown in Figure 24, in particle sample image G40, multiple circular particles are displayed in pixel area PA1.

[0080] <Particle Region Extraction Processing> In this embodiment, prior to the outline extraction processing described above, the outline extraction unit 1022 performs particle region extraction processing on the particle sample image to be analyzed. Here, we consider an image in which the particle region is brighter than the background region. In many cases, the particle region extraction processing for a given image involves the following image processing: Step ST1: Rough separation processing of particle regions (white) and background regions (black) by binarization processing Step ST2: Noise removal processing by opening processing / closing processing Step ST3: Extraction processing of "reliable background region" by expansion processing Step ST4: Extraction processing of "reliable particle region" by distance transformation and binarization processing

[0081] In this embodiment, the contour extraction unit 1022 performs image processing of particle images using steps ST1 to ST4. Step ST4 is followed by extraction of "areas where it is unclear whether they are particles or background," labeling of each area, and watershed processing. However, these image processes do not require parameter value setting, and therefore will not be described in detail in this embodiment.

[0082] Fig. 24 is a diagram showing an example of a particle sample image G40 after the separation process of step ST1. Fig. 25 is a diagram showing an example of a particle sample image G40 after the noise removal process of step ST2. Fig. 26 is a diagram showing an example of a particle sample image G40 after the extraction process of step ST3. Fig. 27 is a diagram showing an example of a particle sample image G40 after the extraction process of step ST4.

[0083] In each of the processes of steps ST1 to ST4 described above, the parameter setting values ​​that the contour extraction unit 1022 should set when executing image processing of the particle sample image G40 are as follows: In the separation process of step ST1, this is the threshold for the binarization process (parameter P1); In the noise removal process of step ST2, this is the kernel size and number of repetitions for the opening process / closing process (parameter P2); In the extraction process of step ST3, this is the kernel size and number of repetitions for the expansion process (parameter P3); and In the extraction process of step ST4, this is the threshold for the distance transformation / binarization process (parameter P4). The relationship between each image process and the setting values ​​will be explained below.

[0084] <Binarization process> A threshold I is applied to an image I(x, y). Ｔ Image I after binarization Ｂ (x, y) is where I ＭＡＸ is the maximum pixel value in the bit depth of the original image I(x, y). For example, if the bit depth is 8, then I ＭＡＸ =2 ８ -1=255. By the binarization process, pixel areas that look like particle areas are colored white (I Ｂ (x, y) = I ＭＡＸ), and the pixel area that seems to be the background area is black (I Ｂ (x, y) = 0), and the image shown in Figure 24 is obtained. As shown in Figure 24, the particle sample image G40 contains a pixel area PA2 that appears to be a particle area, as well as multiple black spot noises N1 and white spot noises N2. The black spot noise N1 is a black spot-like noise in a white area. The white spot noise N2 is a black spot-like noise in a black area.

[0085] <Expansion / Erosion Processing> The particle sample image G40 obtained by the binarization processing shown in Fig. 24 contains sesame-like noise N. The opening processing / closing processing is performed to remove this noise N. The opening processing / closing processing is a combination of expansion / erosion processing, so it will be explained first.

[0086] Both the expansion and contraction processes are performed on a binary image. First, the expansion process is performed on the binary image I. Ｂ For (x, y), the kernel size is K Ｄ , image I after dilation processing with one iteration （Ｄ，１）（ｘ，ｙ） teeth, That is, the original image I Ｂ For each pixel in (x, y), the area K centered at the pixel Ｄ ×K Ｄ If there is at least one pixel value I ＭＡＸ If there is a (white) pixel, I(x, y) = I ＭＡＸ The process of expanding is the process of Ｄ is a positive odd number. By this process, as shown in FIG. 24, black dot noise N1 (black) in the white area is replaced with white and removed. Hereinafter, for simplicity of notation, That is, for example, the binarized image I Ｂ For (x, y), the kernel size is K Ｄ , image I after dilation processing is repeated n times （Ｄ，ｎ）（ｘ，ｙ） of, where o represents the composition of the process and has the following meaning:

[0087] The erosion process is the inverse operation of the expansion process.Ｂ For (x, y), the kernel size is K Ｅ , image I after erosion processing with one iteration （Ｅ，１）（ｘ，ｙ） teeth, That is, the original image I Ｂ For each pixel in (x, y), the area K centered at the pixel Ｅ ×K Ｅ If there is even one pixel with a pixel value of 0 (black) in the pixel area of ​​I(x, y) then the process of setting I(x, y) to 0 is the contraction process. Ｅ Is K Ｄ 24, is a positive odd number. By this process, as shown in FIG. 24, the black dot noise N2 (white) in the black area is replaced with black and removed.

[0088] <Opening / Closing Process> Expansion processing can be used to remove black spot noise N1 from the white area, but performing expansion processing also expands the white area itself. Therefore, by performing contraction processing after expansion processing, the contour extraction unit 1022 can remove black spot noise N1 from the white area without changing the white area. This process of expansion followed by contraction is called opening processing. In other words, when the binary image (I Ｂ For (x, y), the kernel size is K Ｏ , image I after opening processing with one repetition （Ｏ，１）（ｘ，ｙ） teeth, It is calculated by: Ｏ is a positive odd number.

[0089] The closing process is the reverse process of the opening process. In the closing process, the contour extraction unit 1022 can remove white dot noise from the black area without changing the black area. That is, the binarized image I Ｂ For (x, y), the kernel size is K Ｃ , image I after the closing process is repeated once （Ｃ，１）（ｘ，ｙ） teeth, It is calculated by: Ｃ is a positive odd number. The contour extraction unit 1022 may repeat both the opening process and the closing process.

[0090] In the analysis of particle images, in order to find a reliable background region, a kernel size K Ｏ (K Ｃ ), the number of repetitions N Ｏ (N Ｃ ) to remove noise, and then a kernel size K Ｄ , the number of repetitions N Ｄ This makes the larger areas that contain what appear to be particle regions white, and the remaining black areas can be considered "certain background areas."

[0091] <Distance Transform> A certain particle region is considered to be a pixel region near the center of the "likely particle region (white)" PA3 shown in Fig. 25, that is, a pixel region that is sufficiently far from the "certain background region" (black). In order to identify such a region, an image transformation process called distance transformation is performed.

[0092] Distance transformation is a process performed on a binary image. Ｂ A distance map D(x, y;I) obtained by performing a distance transformation on (x, y) Ｂ ))teeth, That is, the original image I Ｂ For each pixel in (x, y), the nearest black pixel (I Ｂ (x', y') = 0) is defined as D(x, y; Ｂ When the distance map is visualized, the white area PA4 that appears to be a particle area is displayed with a gradation of brightness as shown in FIG. 26, and the brighter the area, the farther it is from the "certain background area" (black). The brighter the area, the more likely it is to be a "certain particle area."

[0093] The distance map image shown in Fig. 26 is binarized using a threshold specified by the user to obtain the binarized image shown in Fig. 27. The binarization process using the threshold here is the same as that of the above-mentioned equation (7), but the parameter threshold I ＴThe specific value of varies, and the user specifies and sets an initial value appropriate for the image of the particle region. This specification is performed, for example, using the input device 124. The white particle region PA5 in the particle sample image G40 shown in Figure 27 is a region that is determined to be sufficiently far from the "certain background region," and is therefore considered to be the "certain particle region."

[0094] The above is a description of the image processing involved in particle analysis. After these image processing steps that require the setting of parameter values, the contour extraction unit 1022 performs image processing that does not require the setting of parameter values, thereby obtaining particle regions and their contours as shown in Figure 28. Figure 28 is a diagram showing an example of a particle sample image G40 after image processing that does not require the setting of parameter values. A plurality of particle regions PA6 are displayed in the particle sample image G40 shown in Figure 28.

[0095] When performing the image processing of steps ST1 to ST4 described above, if the actual size of the particles (such as the average diameter or area) is known in advance, the user can input that value into the software as a geometric feature. When the average diameter of the particles is input as a geometric feature, the contour extraction unit 1022 determines the initial values ​​of parameters P1 to P4 using an optimization algorithm for each of the parameters P1 to P4 so as to minimize the difference between the diameter value input by the user and the average diameter of the particle region determined by the processing of steps ST1 to ST4 and the subsequent processing.

[0096] The contour extraction unit 1022 performs steps ST1 to ST4 and subsequent image processing using the initial values ​​of the parameters P1 to P4 thus determined or the initial settings of the parameters P1 to P4 that the software has as constants. Next, the first contour information display unit 1023 displays the particle region and initial contour line obtained based on the results of the image processing on the display unit 1251E of the GUI 1251.

[0097] Next, the user modifies the initial contour displayed on the display unit 1251E and inputs the correct contour by operating the mouse as described in the first embodiment. When the second contour information receiving unit 1024 receives the input correct contour, the second contour information receiving unit 1024 stores the correct contour in, for example, the memory of the overall control unit 120.

[0098] Figure 29 is a diagram showing an example of a particle sample image G40. Figure 29 shows a top-view image of one particle. In Figure 29, a plurality of initial contour points 29 and a plurality of correct contour points 30 are displayed for one particle. The plurality of initial contour points 29 are displayed by the first contour information display unit 1023. The plurality of correct contour points 30 are contour points that the user assumes and expects and inputs by operating a mouse or the like. The positions of the plurality of correct contour points 30 are each received by the second contour information receiving unit 1024.

[0099] By using an optimization algorithm that operates the parameter values ​​of the parameters P1 to P4, the parameter value acquisition unit 1025 can obtain the values ​​of the parameters P1 to P4 so as to minimize the difference between the initial contour point 29 and the correct contour point 30.

[0100] Fig. 30 is a diagram showing an example of a particle sample image G40. In Fig. 30, for one particle, multiple corrected contour points 31 are displayed on the particle sample image G40. The corrected contour points 31 are contour points that minimize the difference between the initial contour points 29 and the correct contour points 30.

[0101] FIG. 31 is a diagram showing an example of the display of the display unit 1251E while the optimization algorithm is running. In FIG. 31, in addition to the particle sample image G40, a loss calculation display G45 that displays a loss calculation is displayed. The loss calculation display G45 displays the calculation results of the loss for the parameters. The loss calculation display G45 displays a graph showing the calculation results of the loss result. The parameter value that minimizes the loss is displayed as point O' on the graph. The processing order of this optimization algorithm is the same as that explained using FIG. 15 in the first embodiment. The graph showing the calculation results of the loss result does not have to be displayed, and may be dotted, as explained using FIG. 15.

[0102] The process of setting the initial values ​​of the parameters P1 to P4 described above can also be applied to the contour extraction process performed by the contour extraction unit 1022 for the sample cross-sectional image G10, as described in the first embodiment. For example, if the expected value of the periodic width (referred to as the pitch) of the horizontally arranged pillars 11 shown in FIG. 4 is known in advance, the contour extraction unit 1022 can obtain the pitch calculated from the initial contour extracted using that expected value. The contour extraction unit 1022 can then obtain initial settings for the smoothing and threshold parameters used in the contour extraction process so that the displayed difference between the pitch of contour points extracted using the initial settings and the pitch obtained from the expected values ​​is minimized. This allows the contour extraction unit 1022 to extract an initial contour that more accurately follows the structure of the pillars 11. After the first contour information display unit 1023 displays the initial contour line, the second contour information receiving unit 1024 receives the correct contour line, and the parameter value acquisition unit 1025 optimizes the parameter values ​​P1 to P4, etc., are the same as the processing described in the first embodiment.

[0103] Although the embodiments of the present disclosure have been specifically described above, they are not limited to the above-described embodiments and various modifications are possible without departing from the spirit of the present disclosure. In each embodiment, components can be added, deleted, or replaced, except for essential components. Unless otherwise specified, each component may be singular or plural. Combinations of the embodiments and their variations are also possible. Some or all of the above-described configurations, functions, processing units, etc. may be realized by hardware, such as an integrated circuit design, or by software in which a processor interprets and executes a program. Data and information such as programs, tables, and files that realize each function can be stored in a storage device such as a memory, hard disk, or SSD, or in a storage medium such as an IC card, SD card, or DVD.

[0104] 11...protrusion structure (pillar), 12...valley structure (trench), 13...reference line, 16...reference point, 17, 18...line profile, 19...contour point, 20...initial contour line, 21...correct contour line, 22...corrected contour line, 29, 231, 232, 261, 262...initial contour point, 30, 241, 242, 271, 272...correct contour point, 31, 251, 252, 281, 282...corrected contour point, 100...charged particle beam device (SEM), 101...main body, 102...controller, 111...electron gun, 112...focusing lens, 113...deflection lens, 114...objective lens, 114...and objective lens, 115...stage, 116...detector, 130 ...Sample, 1021...Image display section, 1022...Contour extraction section, 1023...First contour information display section, 1024...Second contour line reception section, 1024...Second contour information reception section, 1025...Parameter value acquisition section, 1026...Third contour information display section, 1251...Graphical user interface, 1251A...Import image button, 1251B...Detect button, 1251C...Edit button, 1251D...Setting adjustment button, 1251D...Setting adjustment button by user, AW1, AW2...Double arrows, G10, G20...Sample cross-section image, G15, G25, G45...Loss calculation display, G30...Top view image, G40...Particle sample image

Claims

1. An image processing system that processes an observation image of a sample obtained by a charged particle beam device, comprising: an image display unit that displays the observation image; a contour extraction unit that extracts first contour information relating to a structural boundary of the sample from the observation image based on a first setting value of a preset parameter; a first contour information display unit that displays the first contour information extracted by the contour extraction unit on the observation image; a second contour information receiving unit that receives input of second contour information relating to a boundary visible in the observation image; and a parameter value acquisition unit that acquires a second setting value different from the first setting value so as to minimize the difference between the first contour information and the second contour information in the observation image.

2. An image processing system according to claim 1, further comprising a third contour information display unit that displays, on the observed image, third contour information extracted by the contour extraction unit based on the second setting value acquired by the parameter value acquisition unit.

3. An image processing system according to claim 1, wherein there are a plurality of types of said parameters, and one of said plurality of parameters is a parameter for removing noise from an image.

4. An image processing system according to claim 1, wherein there are a plurality of types of said parameters, and one of said plurality of parameters is a parameter that determines a threshold for distinguishing between regions of an image.

5. An image processing system as described in claim 1, wherein the contour extraction unit receives input of reference lines and reference points for defining the structural boundary based on the observation image displayed on the image display unit, and extracts the first contour information from the observation image based on the reference lines and reference points and the first setting value.

6. An image processing system according to claim 1, wherein the first contour information is moved on the observation image displayed on the image display unit, the second contour information is shaped by moving the first contour information, and a range within which the first contour information can be moved on the observation image is displayed on the observation image.

7. An image processing system according to claim 1, wherein there are a plurality of types of said parameters, and one of said plurality of parameters is a kernel size of a smoothing filter process that smoothes an image of a selected region of said observed image.

8. An image processing system according to claim 1, wherein there are a plurality of types of parameters, and one of the plurality of parameters is a kernel size of an opening process that removes black spot noise from a white region of the observed image without changing the white region.

9. An image processing system according to claim 1, wherein there are a plurality of types of parameters, and one of the plurality of parameters is a kernel size of a closing process that removes white spot noise from black areas of the observed image without changing the black areas.

10. An image processing method executed by an image processing system that processes an observation image of a sample obtained by a charged particle beam device, wherein the image processing system displays the observation image, extracts first contour information relating to a structural boundary of the sample from the observation image based on a first setting value of a preset parameter, displays the extracted first contour information on the observation image, accepts input of second contour information relating to a boundary visible in the observation image, and obtains a second setting value different from the first setting value so as to minimize the difference between the first contour information and the second contour information in the observation image.

11. A display device used in an image processing system that processes an observation image of a sample obtained by a charged particle beam device, the display device: displays first contour information relating to a structural boundary of the sample extracted from the observation image based on a first setting value of a preset parameter on the observation image; accepts input of second contour information relating to a boundary visible in the observation image; acquires a second setting value different from the first setting value so as to minimize the difference between the first contour information and the second contour information in the observation image; and displays third contour information extracted based on the acquired second setting value on the observation image.

Images