Method and non-transitory computer-readable medium for intelligent measurement of microscope images

Abstract

The intelligent metering method and device disclosed in this document process images for automatic metering of required features. An example method at least includes: extracting a region of interest from the image, the region containing one or more boundaries between different segments; enhancing the extracted region of interest based on one or more filters; generating a multi-scale data set of the region of interest based on the enhanced region of interest; initializing a model of the region of interest; optimizing a plurality of active contours within the enhanced region of interest based on the model of the region of interest and further based on the multi-scale data set, the optimized plurality of active contours identifying the one or more boundaries within the region of interest; and performing metering on the region of interest based on the identified boundaries.

Description

[0001] Divisional Application

[0002] This application is a divisional application of the application with the application number 201911023052.X, the application date of October 25, 2019, and the title of "Method and Non - Transitory Computer - Readable Medium for Intelligent Metrology of Microscope Images". Technical Field

[0003] The technology disclosed herein generally relates to performing intelligent metrology using images, and more particularly, to performing intelligent metrology on images obtained using a charged - particle microscope. Background Art

[0004] The metrology of features in an image can be a difficult and time - consuming process, especially for images obtained by a charged - particle microscope. The difficulty and length of the process may be partly attributed to noisy images that are difficult for automatic imaging - processing algorithms to handle, which results in user manipulation and multiple steps. When only a small number of images are to be analyzed, this may not be a problem. However, in a manufacturing environment such as the semiconductor industry, where a large number of samples are imaged and need to be analyzed, this process really slows down the required analysis.

[0005] Although there have been some improvements in image processing over the years, which have improved metrology accuracy and efficiency, in today's semiconductor manufacturing environment, these improvements are not enough, such as node size and throughput. Therefore, there is a need for improved image processing and metrology automation across the industry. Summary of the Invention

[0006] The intelligent metrology methods and devices disclosed herein process images for the automatic metrology of desired features. Example methods at least include: extracting a region of interest from the image, the region including one or more boundaries between different segments; enhancing the extracted region of interest based on one or more filters; generating a multi - scale data set of the region of interest based on the enhanced region of interest; initializing a model of the region of interest; optimizing a plurality of active contours within the enhanced region of interest based on the model of the region of interest and further based on the multi - scale data set, the optimized plurality of active contours identifying the one or more boundaries within the region of interest; and performing metrology on the region of interest based on the identified boundaries.

[0007] Another embodiment includes a non - transitory computer - readable medium that includes code which, when executed by one or more processors, causes the one or more processors to: extract a region of interest from an image, the region including one or more boundaries between different segments of the region of interest; at least enhance the extracted region of interest based on one or more filters; generate a multi - scale data set of the region of interest based on the enhanced region of interest; initialize a model of the region of interest, the initialization of the model determining at least a first boundary and a second boundary of the region of interest; optimize a plurality of active contours within the enhanced region of interest based on the model of the region of interest and further based on the multi - scale data set, the optimized plurality of active contours identifying the one or more boundaries within the region of interest; and perform metrology on the region of interest based on the identified boundaries. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] This patent or application file contains at least one color drawing. Copies of this patent or patent application publication with color drawing(s) will be provided by the Office upon request and payment of the necessary fee.

[0009] Figure 1 is an example image sequence showing image processing and resulting metrology according to an embodiment of the present disclosure.

[0010] Figure 2 is an example image sequence showing image processing according to an embodiment of the present disclosure.

[0011] Figure 3 is an example method for processing an image and performing metrology on one or more features within the image according to an embodiment of the present disclosure.

[0012] Figure 4 is a method for optimizing a plurality of active contours within an ROI according to an embodiment of the present disclosure.

[0013] Figure 5 is an example image sequence according to an embodiment of the present disclosure.

[0014] Figure 6 is an example method according to an embodiment of the present disclosure.

[0015] Figure 7 is a block diagram of a computer system illustrating an embodiment in which the present invention can be implemented.

[0016] Figure 8 is an example charged - particle microscope environment for performing at least a portion of the methods disclosed herein and according to an embodiment of the present disclosure.

[0017] Throughout several views of the drawings, like reference numerals refer to corresponding parts. Detailed Implementation Modes

[0018] Embodiments of the present invention relate to intelligent metrology of microscope images. In some instances, an image is processed to enhance desired features, and then an active contour is optimized to locate boundaries formed at the feature interfaces. Subsequently, the active contour serves as an anchor point for performing accurate metrology on the features. However, it should be understood that the methods described herein are generally applicable to a variety of different AI-enhanced metrologies and should not be considered restrictive.

[0019] As used in this application and the claims, unless the context clearly indicates otherwise, the singular forms "a," "an," and "the" include plural forms. Additionally, the term "comprising" means "including." Further, the term "coupled" does not exclude the presence of intermediate elements between the coupling members. Additionally, in the following discussion and claims, the terms "comprising" and "including" are used in an open-ended manner and should therefore be interpreted to mean "including but not limited to...". An "integrated circuit" refers to a set of electronic components and their interconnections (collectively referred to as internal circuit elements) patterned on the surface of a microchip. The term "semiconductor device" generally refers to an integrated circuit (IC), which may be essential to a semiconductor wafer, separate from the wafer, or used in a package on a circuit board. The term "FIB" or "focused ion beam" is used herein to refer to any collimated ion beam, including an ion beam focused by ion optics and a shaped ion beam.

[0020] The systems, devices, and methods described herein should not be construed in any way as restrictive. On the contrary, the present disclosure relates to all novel and non-obvious features and aspects of the various disclosed embodiments (individually and in various combinations and sub-combinations with each other). The disclosed systems, methods, and devices are not limited to any particular aspect or feature or combination thereof, and the disclosed systems, methods, and devices do not require the presence of any one or more specific advantages or that problems be solved. Any theory of operation is for ease of explanation, but the disclosed systems, methods, and devices are not limited to such theory of operation.

[0021] Although the operations of some of the disclosed methods are described in a particular sequential order for convenience of presentation, it should be understood that this description encompasses rearrangements, unless the specific ordering is required by the specific language set forth below. For example, in some cases, the operations described in sequence may be rearranged or performed concurrently. In addition, for the sake of brevity, the drawings may not show the various ways in which the disclosed systems, methods, and devices may be combined with other systems, methods, and devices. Additionally, this specification sometimes uses terms such as "generate" and "provide" to describe the disclosed methods. These terms are high-level abstractions of the actual operations being performed. The actual operations corresponding to these terms will vary depending on the specific implementation and are readily discernible by those of ordinary skill in the art.

[0022] The intelligent metrology techniques disclosed herein enable the automatic metrology of features in images. Images can be obtained using charged particle microscopes (SEM, TEM, STEM, FIB, etc.) or optical microscopes (a few examples are provided). The images can typically contain one or more features, such as regions of interest, and the size of at least one or more features needs to be determined. For example, intelligent metrology can perform measurements by combining active contours based on multiple image processing algorithms. The active contours can be initialized based on the processed image, and the active contours can be iteratively optimized using one or more scale spaces. For example, the image can be initially processed to locate and isolate one or more regions of interest (ROIs). Subsequently, one or more of these ROIs can be subjected to data attribute normalization filtering to increase the signal-to-noise ratio (SNR) and / or improve the contrast between the various parts of the features in the image. After improving the SNR and / or contrast in the image, the image is subjected to one or more image processing algorithms to enhance the sharpness of the boundaries between the various parts of the features and / or distinguish and detect. The distinction and detection of the boundaries can be performed with or without transforming the data into a different representation space (such as Cartesian space or polar space), as may be required by the specific image processing algorithms implemented. Subsequently, the data obtained from the previous process is processed to generate a multi-scale dataset in Gaussian, geometric, non-linear, and / or adaptive shape scale spaces. Subsequently, the multi-scale dataset is used to initialize multiple active contours in the image. Subsequently, the active contours are iteratively initialized and optimized at each resolution level of one or more spatial scales, thereby allowing for the gradual optimization of the active contours. Once optimized, the active contours identify and locate all the boundaries within the ROI.

[0023] Subsequently, the optimized active contours can form the basis for performing metrology on the desired parts or aspects of the features in the original image or the enhanced image. The metrology can include geometric analysis of the segmented regions of the features based on the contours therein. In addition, the metrology can inform other analytical aspects of the feature analysis as well as statistical analysis, etc.

[0024] Additionally or alternatively, the active contours can be initialized based on a separate imaging modality (e.g., multimodal analysis). For example, energy-dispersive x-ray spectroscopy (EDX) can be used to analyze a sample containing the features to first determine the initial boundaries between the various parts of the features, and then the initial boundaries can be used as the positions for initializing the corresponding active contours. EDX can be a single-line scan across the sample, which will provide at least one point on the feature where the boundary may be located. Some samples may contain circular ROIs, which will then provide two points for each boundary from the EDX line scan. Of course, a full 2-D EDX scan can also be performed, but the time involved may not be ideal.

[0025] Figure 1FIG. 100 is an example image sequence showing image processing and resulting metrology in accordance with embodiments of the present disclosure. Image sequence 100 can be performed on any image type, but will be discussed in the context of images obtained by a charged particle microscope. More specifically but not restrictively, the images used to illustrate the disclosed techniques pertain to semiconductor structures. Image processing and metrology can be performed on any desired image content, and Figure 1 the elliptical features of the images in FIG. 100 are merely exemplary and not limited to the techniques disclosed herein. For image sequence 100, the elliptical regions are vertical memory devices, sometimes referred to as VNANDs. These devices contain various layers formed in high aspect ratio holes (e.g., holes formed in a material that extend through one or more epitaxial layers), where each of the various material layers is formed of a different material in order to form a working circuit device. For process control and defect detection, manufacturers of VNAND devices need to measure the thickness of the material layers that form the operating device, which requires a high-resolution microscope, such as SEM and / or TEM, to identify the material layers and their interfaces, such as the boundaries therebetween.

[0026] Image sequence 100 begins with an original image 102. Although image 102 appears as a dark elliptical shape surrounded by lighter regions, there are actually several rings within the dark elliptical shape, such as material layers, as can be seen in images 108 and 110 for example. Image 100 can be a dark field (DF) image or a HAADF image (high angle annular dark field), which typically results in a brighter background than the region of interest (ROI). For the image processing sequence, the goal is to determine the position and / or boundaries of each ring within the ROI, and to measure the width of one or more of the rings, e.g., perform metrology on a VNAND device. In order to perform the desired metrology, at least the boundaries between them need to be identified so that more accurate measurements can be made. However, in order to perform such measurements, the image may need to be enhanced, and the positions of the boundaries may need to be more accurately identified.

[0027] Typically, the sequence can include a preprocessing segment that enhances the image in terms of signal-to-noise ratio, contrast, region sharpness, etc., and also includes the extraction / identification of one or more ROIs. After preprocessing or general image enhancement, a multi-scale data set is generated so that an active contour can be initialized and optimized. The optimized active contour will locate the boundaries between the various materials within the image. Based on the identification of the boundaries, metrology of the desired layers or regions within the image can then be performed, which can be done automatically.

[0028] To begin image processing, the original image 102 can be analyzed to extract one or more ROIs. The extraction of the ROIs can provide a rough outer boundary for each ROI and specify the area where image processing can be concentrated, such as within the rough outer boundary. There are many techniques for ROI extraction, such as binarizing the image to define the outer boundary of an instance. Additional techniques will be discussed below. In some embodiments, the meta-information of the original image can also be considered in image processing. The meta-information includes information such as the data type (e.g., imaging mode), resolution, pixel size, etc.

[0029] After the ROIs are extracted, at least the portion of the image is subjected to further processing to enhance the image of the ROIs, as indicated by the dashed box 106. Generally, the enhancement of the ROIs includes reducing the signal-to-noise ratio, enhancing the contrast and image sharpness, so as to roughly identify the boundaries between different sections / materials. For example, image 108 shows the ROI of the original image (background removed) with improved contrast. Then the image 108 with improved contrast can be subjected to further processing to increase the sharpness, as shown in image 110. Image 110 shows the ROI after various filtering operations (such as reaction-diffusion filtering) have been performed.

[0030] In addition, image 110 can be used to generate a multi-scale data set, where different scales at least include a Gaussian scale space, a geometric scale space, a non-linear scale space, and an adaptive shape scale space. Then the images in each scale space are subjected to a series of blurring and subsampling to smooth the images. Such processing is performed to construct a potential surface for the smooth deformation of the active contour.

[0031] Furthermore, a model of the ROI is formed to provide the boundaries of the ROI for further processing. For example, various maps of the ROI can be generated to determine the inner and outer boundaries for establishing the area where further image processing will occur. An example map is the distance map of the ROI, which determines the center and outer edge of the ROI.

[0032] After enhancing sharpness and contrast in at least one or more ROIs of the original image, a large number of active contours can be initialized within the enhanced ROIs and placed in response to the generated ROI models, where the large number includes dozens to hundreds of active contours. The large number of active contours is used because the number and their positions of the boundaries (e.g., material layers) may not be known a priori. Generally, the number of initialized active contours will be greater than the number of boundaries. Image 112 shows the initialization of a plurality of active contours within the ROI, which can be placed on the original image or an enhanced version of the original image. The active contours will be allowed to optimize to locate the boundaries, which will coincide with the minimum energy positions of the image. Image 114 shows the optimized active contours, such as snakes. Some of the plurality of active contours will be optimized incorrectly and will thus be removed. For example, incorrectly optimized active contours may obscure different boundaries within the ROI.

[0033] Next, the optimized active contours can be used as a reference for measuring the thicknesses of different material layers within the ROI, as shown in Image 116.

[0034] Figure 2 Is an example image sequence 200 showing image processing according to an embodiment of the present disclosure. Sequence 200 is an example image processing sequence that illustrates parts of the entire sequence that are different from those shown in image sequence 100. Generally, each step of the processes discussed herein can be performed using any one or more of the image processing algorithms, and the selection of the algorithms implemented can be automatically selected based at least on the meta-information of the original image. Image sequence 200 begins with an original image 202, which, as can be seen, has an initial quality different from that of image 102. Additionally, image 202 can be obtained using bright field (BF) imaging or TEM imaging mode, and the TEM imaging mode produces an image with a darker background than the ROI. The VNAND structure of image 202 can be processed to extract the ROI, and the result is shown in image 204. As described above, the extraction of the ROI can generally establish the outer boundary of the ROI.

[0035] Next, the ROI of the extracted image 204 can be used as a template for forming a distance map on the image 202, e.g., initializing the model of the ROI. The distance map is centered on the ROI and can be used to restrict the initial placement of the active contour, as shown in image 212. Additionally, the model of the ROI shown in image 218 is used to establish the inner and outer boundaries, which are indicated by the innermost and outermost dashed lines in image 212. The active contour of image 212 can be placed on an image that has undergone ROI enhancement processing, e.g., as shown in box 106 of sequence 100, while generating a multi-scale dataset that provides energy values to the regions within the ROI. The active contour can be placed at positions that are increasingly farther from the background, e.g., outside the extracted ROI but within the regions of the ROI. Additionally, the model of the ROI shown in image 218 is used to establish the inner and outer boundaries, which are indicated by the innermost and outermost dashed lines in image 212.

[0036] After placing the active contours, they are allowed to optimize based on the multi-scale dataset. After optimizing the active contours, metrology can be performed on the layers of the VNAND shown in image 202, as shown in box 220.

[0037] Figure 3 is an example method 300 for processing an image and performing metrology on one or more features within the image according to an embodiment of the present disclosure. In Figure 1 and 2 The method 300, which is at least partially described, can be performed on images obtained using a charged particle microscope (e.g., SEM, TEM, STEM), to name just a few examples. However, the use of charged particle images does not limit the techniques disclosed herein. Additionally, the method 300 can be performed by imaging tool hardware, one or more servers networked to the imaging tool, a user's desktop workstation, or a combination thereof. Generally, the method 300 implements one or more image processing techniques automatically selected from a technology library to obtain an image that can be accurately and autonomously measured, e.g., performing metrology on one or more features of the image.

[0038] The method 300 can begin at processing block 301, which includes preprocessing the image. Generally, before using active contours to identify the boundaries within the ROI, the image can undergo some processing to enhance contrast and clarity, thereby enabling metrology of the layers that form these boundaries. The preprocessing can include multiple processes to extract the ROI, enhance contrast / region clarity, improve SNR, and distinguish and detect region boundaries. Image enhancement can be performed only within the ROI or on the entire image, and is a non-limiting aspect of the present disclosure. Of course, restricting the enhancement to within the ROI can improve the processing time and efficiency of the entire method 300.

[0039] Processing block 303 (an optional sub-step of processing block 301) includes extracting a region of interest from an image. The image can be an originally acquired image, such as image 102, 202, or can be a cropped portion of the original image. Along with the image data of the input image, processing 300 also receives meta-information about the input image, where the meta-information includes data type (e.g., imaging mode), pixel size, and other data about the image. The meta-information can be used to help automatically determine what image processing techniques to implement in the image processing steps of method 300, e.g., implementing at least steps 301 and 303. The meta-information attached to the image causes the preprocessing filter implemented in processing block 301 to automatically tune its parameters to the imaging mode of the image. The resolution helps determine which material layer is accurately segmented. For example, if the image has a low resolution, there may not be enough pixels in some material layers that are only one or two nanometers thick to be accurately resolved. Pixel size (which is also related to resolution) may be required to automate the process by normalizing the data to a standard pixel size and generating measurements in standard MKI units rather than pixel units.

[0040] There are many image processing techniques available for extracting the ROI. The following contains a non-exclusive list of techniques that can be implemented, but the techniques used are non-limiting aspects of the present disclosure: linear isotropic diffusion, histogram manipulation for contrast enhancement, automatic thresholding, component labeling, problem-specific size and shape criteria, part representation detection and elimination, etc., which are automatically selected and applied. The extraction of the ROI can provide a rough boundary within which additional image processing and metrology will be performed.

[0041] After processing block 303 can be processing block 305, which includes at least enhancing the ROI within the image. Generally, the enhancement is performed to obtain data attribute normalization and improve the detection of the region boundaries within the ROI. Additionally, the enhancement can improve image contrast, signal-to-noise ratio (SNR), region sharpness, discrimination and detection of region boundaries. The discrimination and detection of region boundaries can be performed with or without transforming the data to a different representation space such as Cartesian space or polar space, as may be required by the implemented image filtering techniques.

[0042] In some embodiments, processing block 305 can be divided into two processing steps, where in one step (step A), contrast and SNR are improved, and in another step (step B), enhancing regional sharpness and the differentiation and detection of regional boundaries are performed. Many image processing algorithms can be selected to implement step A, such as histogram manipulation, linear and nonlinear contrast enhancement, data normalization based on local low-frequency data distribution, gamma correction, logarithmic correction, brightness correction, etc., where the selected algorithm is at least automatically applied to the ROI of the image and is selected at least based on meta information. It should be noted that step A also performs a data attribute normalization step.

[0043] Similarly, step B includes selecting one or more algorithms from a set of automatically implemented similar algorithms at least based on meta information. The implemented algorithms can be selected from reaction-diffusion filtering, inherited isotropic and anisotropic diffusion, median filtering, nonlinear diffusion based on the Mumford-Shah model, background suppression and edge / boundary extraction, coherent enhancement filtering on object boundaries, and the application of amplitude features, texture (Gabor, Haralick, Laws, LCP, LBP, etc.) techniques. In some embodiments, the application of another imaging modality, such as energy-dispersive x-ray spectroscopy (EDX), electron energy loss spectroscopy (EELS), etc. (if available), can be used to differentiate regional boundaries. A more detailed discussion of the use of other imaging modalities is included below.

[0044] After processing block 301 can be processing block 307, which includes generating a multi-scale dataset of at least the ROI. The multi-scale dataset can be generated in one or more of multiple scale spaces, such as Gaussian scale space, geometric scale space, nonlinear scale space, and adaptive shape scale space. One or more of the scale spaces used to generate the multi-scale dataset will be used to optimize multiple active contours. The active contours can be initialized and optimized at each scale level to determine the boundaries within the ROI at the original scale level. Thus, the multi-scale dataset will be the basis for optimizing the active contours to identify the boundaries between parts of the ROI.

[0045] Following processing block 307 can be processing block 309, which initializes at least a model for the ROI. Processing block 309 initializes a general model for the ROI to form a first boundary and a second boundary of the ROI. Additional processing will mainly be performed within the boundaries set by the model. The model can be formed based on one or more techniques selected from: binary labeled maps, interactive maps, distance maps, CAD maps, statistical models from data, dye casting, random distribution of geometries, geometric models, etc. The distance map shown in image 218 provides an example of an initial model. It should be noted that processing block 309 can be executed in parallel with processing blocks 305 and 307 and does not need to be after processing block 307. Additionally, the model initialized in processing block 307 can be based on the original image or the enhanced image.

[0046] Following processing block 309 can be processing block 311, which includes optimizing a plurality of active contours within the enhanced ROI to locate the feature boundaries within the enhanced ROI. The process of optimizing the active contours or allowing the active contours to be optimized can start with initializing a first plurality of active contours, where the number of the first plurality of active contours is greater than the number remaining after optimization and is also greater than the number of boundaries within the ROI. Due to the lack of prior knowledge of the number of boundaries within the ROI and / or their positions within the ROI, it is possible to execute the initialization of more active contours than boundaries. Although all initialized active contours will be optimized, some may be combined due to being optimized to the same boundary, while others may be removed due to incorrect optimization, e.g., being optimized to different boundaries within the ROI. Thus, the optimized active contours will identify and locate the boundaries separating different segments / materials within the ROI.

[0047] The process of initializing and optimizing the active contours can be an iterative process that is performed and based on each scale level of the multi-scale dataset generated in processing block 307. For example, the first initialization and optimization of the active contours can be performed at the fourth scale level of the enhanced ROI image, where the fourth scale level has a resolution of 1 / 16 of the original enhanced ROI image. The active contours optimized on the fourth scale level image then become the initial active contours on the third scale level image (e.g., 1 / 8 resolution image), which are then allowed to be optimized. This process iterates until the original scale image has initialized and optimized active contours, thus identifying and locating the desired boundaries within the ROI.

[0048] After processing block 311 can be processing block 313, which performs metrology on the ROI within the original image based on a plurality of active contours optimized. The metrology can provide measurements of the widths of different sections based on the distances between different boundaries and further provide information about the overall shape of the boundaries. In some embodiments, the metrology can use segmentation, e.g., geometric analysis of the identified sections and contours. After the metrology, the obtained data can be used for statistical inference, hypothesis generation, defect detection, process control, time analysis, prediction, and other applications.

[0049] Figure 4 is method 400 for optimizing a plurality of active contours within an ROI according to an embodiment of the present disclosure. Method 400 can be implemented in combination with method 300, e.g., step 311, and can provide an example of the implementation of step 311. Method 400 can start at processing block 401, which includes generating a multi-scale dataset of the image. The multi-scale dataset will include datasets at multiple resolution levels and for each of a plurality of scale spaces (e.g., Gaussian, geometric, non-linear, and adaptive shape scale spaces, to name a few). Of course, other scale spaces can also be implemented. The multi-space dataset is generated from an enhanced image, e.g., generated by processing block 301 of method 300. Using the enhanced image provides better-defined boundaries that are highlighted and smoothed in the multi-scale dataset. After processing block 401 can be processing block 403, which indicates to complete the remainder of method 400 for each scale space.

[0050] For each scale space, processing blocks 405 and 407 can be performed for each resolution level. And once all the resolution levels of a scale space have been performed, the optimized active contours can be used as the initial active contours for the subsequent scale space. Of course, this is not restrictive, and each scale space can start with a new set of active contours.

[0051] After processing block 403 can be processing block 405, which includes initializing a plurality of active contours on the ROI. If this is the initial execution of processing block 403 for a scale space, the plurality of active contours will be initialized on the lowest resolution level data. The lowest resolution level can depend on the initial quality of the image but can be a resolution of 1 / 16 or less. However, the lowest level used is a non-limiting aspect of the present disclosure. If this is not the lowest level resolution image, the initialized active contours will be the optimized active contours from a lower resolution level image, e.g., a previous iteration of processing blocks 405 and 407.

[0052] The images of the ROIs at different resolution images in each scale space can be characterized as having been blurred by a certain amount based on the resolution level, where the lower the resolution, the more blur occurs. Blurring the boundaries of the enhanced image provides a larger energy band for the active contour to optimize accordingly. It should also be noted that higher-level datasets have less blur, which results in a narrower energy band for optimization. Thus, by continuously using the active contour optimized at a lower resolution, the active contour is optimized to full resolution in a step-by-step manner.

[0053] Following processing block 405 can be processing block 407, which includes optimizing multiple active contours. Optimizing the active contours allows the active contours to move / stabilize in the middle of the smooth boundaries provided at each resolution level. When processing blocks 405 and 407 are executed iteratively, the active contours are optimized in a stepwise function, and they are ultimately optimized to the boundaries in the original resolution image, thereby identifying and locating one or more boundaries in the image.

[0054] Following processing block 407 is processing block 409, which determines whether all resolution levels of the scale space are completed. If not, processing blocks 405 and 407 are repeated for the next resolution level image. If so, method 400 proceeds to processing block 411, which determines whether all scale spaces have been completed. If not, the process returns to processing block 403, otherwise it ends at processing block 413. Completion of method 400 locates and identifies all boundaries within the ROI of the enhanced image, which can then overlap or be associated with regions / boundaries in the original image.

[0055] Figure 5 is an example image sequence 500 according to an embodiment of the present disclosure. Image sequence 500 illustrates another process that can be used to initialize and optimize an active contour for anchoring metrology. Image 502 includes an original image of a VNAND and portions of two other VNANDs. Image 522 shows EDX data on the same VNAND as shown in image 502. As can be seen from image 522, the EDX data providing chemical analysis shows variations between the materials forming the different rings of the VNAND. This chemical information can be used to determine several boundaries within the VNAND and the approximate location of each boundary.

[0056] In some embodiments, the EDX data can be provided by a large area scan, as shown in image 522. However, this large area scan can be replaced by a simpler and more efficient EDX line scan. The EDX line scan can be performed on the VNAND or any imaging structure to identify the boundaries of each ring spanning the diameter of the structure.

[0057] Next, the EDX data of image 522 can be used to initialize a corresponding number of active contours at the indicated boundaries and place them on image 522 according to the positions of the boundaries. After placing the active contours, they are allowed to be optimized. The optimization of the active contours should more accurately identify the positions of the boundaries. In some embodiments, the boundaries identified by the EDX data will be used to place a corresponding number of active contours for optimization. For example, if the EDX data shows seven boundaries, seven active contours will be initialized, each initialized at the position of one of the identified boundaries. In other embodiments, the enhanced image data will be used to generate a multi-scale data set, and then the multi-scale data set serves as the basis for the initialization and optimization of the active contours. However, in such embodiments, the number and positions of the initialized active contours will be based on the boundaries identified by the EDX data.

[0058] As shown in image 514, the optimized active contours can be overlaid on the original image 502 or an enhanced version of image 502 to provide a basis for performing metrology, as indicated by block 516.

[0059] Figure 6 is an example method 600 according to an embodiment of the present disclosure. Method 600 is at least partially illustrated by image sequence 500 and includes an imaging modality different from the imaging modality used to form the image to assist in determining section boundaries and anchoring metrology. Although the present disclosure uses EDX as the different modality, other modalities such as EELS can also be used. Method 600 can start at processing block 601, which includes obtaining an image of at least the ROI of the sample. The image can be obtained using SEM, TEM, STEM, or other imaging techniques, which can be referred to as the first imaging technique. For a charged particle microscope, the image can be a grayscale image based on electrons passing through the sample (e.g., TEM or STEM), secondary electrons (e.g., SEM), and / or backscattered electrons (e.g., SEM). Of course, an image can also be obtained from an optical-based microscope.

[0060] After processing block 601 can be processing block 603, which includes performing at least a second imaging technique on the ROI. The second imaging technique can be any imaging / analysis modality different from the first imaging technique. In some embodiments, it may be desirable for the second imaging technique to be a chemical analysis tool such as EDX. Using EDX as an example second imaging technique, the EDX data will chemically show the changes in the material. These changes in the material provide an approximation of the boundaries within the ROI. Using EDX as the second imaging technique can be performed as a two-dimensional area scan over all ROIs or as a line scan over the ROI.

[0061] After processing block 603 can be processing block 605, which includes initializing a plurality of active contours within the ROI of the acquired image, where the acquired image is obtained using a first imaging technique. In some embodiments, the acquired image is the initially acquired image that has not been subjected to any additional image processing. However, in other embodiments, a plurality of active contours can be initialized on an enhanced image that has been processed to improve contrast, SNR, sharpness, etc., and is imaged using a second imaging technique to more precisely define the boundaries. In yet another embodiment, as described above, a plurality of active contours can be initialized and optimized iteratively and recursively on a series of resolution-adjusted images in one or more scale spaces. However, although previous methods (such as methods 300 and 400) initialized more active contours than boundaries, in method 600, the second imaging technique provides several boundaries within the ROI. Thus, the initialization of the active contours in method 600 includes initializing the active contours corresponding to the number of boundaries identified by the second imaging technique. Additionally, the active contours initialized in processing block 605 will be initialized at the positions determined by the second imaging technique data.

[0062] After processing block 605 can be processing block 607, which includes optimizing the plurality of active contours within the ROI to locate the feature boundaries within the ROI. The optimization of the active contours can be performed as previously discussed with respect to methods 300 and / or 400, but can also be performed based on the second imaging technique data. Regardless of the optimization process, processing block 605 generates identification and position information regarding the boundaries within the ROI, which are the interfaces between different materials within the ROI, such as Figure 5 VNAND.

[0063] After processing block 607 is processing block 609, which includes performing metrology on the features within the ROI based on the optimized plurality of active contours. The metrology provides measurements of various aspects of the features within the ROI, such as feature size, the overall shape of the features within the ROI, information regarding process control and / or defects, and other desired measurement-based information.

[0064] According to one embodiment, the techniques described herein are implemented by one or more special-purpose computing devices. The special-purpose computing devices can be hard-wired to perform the techniques, or can include digital electronic devices such as one or more application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or network processing units (NPUs) that are persistently programmed to perform the techniques, or can include one or more general-purpose hardware processors or graphics processing units (GPUs) that are programmed to perform the techniques in accordance with program instructions in firmware, memory, other storage devices, or a combination thereof. Such special-purpose computing devices can also combine custom hard-wired logic, ASICs, FPGAs, or NPUs with custom programming to implement the techniques. The special-purpose computing devices can be a desktop computer system, a portable computer system, a handheld device, a network device, or any other device that combines hard-wired and / or program logic to implement the techniques. In some embodiments, the special-purpose computing device can be part of a charged particle microscope or coupled to the microscope and other user computing devices.

[0065] For example, Figure 7 is a block diagram of a computer system 700 on which embodiments of the present invention can be implemented. The computing system 700 can be an example of computing hardware included in a Figure 8 charged particle environment as shown. The computer system 700 includes at least a bus or other communication mechanism for conveying information, and a hardware processor 730 coupled to the bus (not shown) for processing information. The hardware processor 730 can be, for example, a general-purpose microprocessor. The computing system 700 can be used to implement the methods and techniques disclosed herein, such as methods 300, 400, and / or 600, and can also be used to acquire images and process the images using one or more filters / algorithms.

[0066] The computer system 700 also includes a main memory 732 coupled to the bus for storing information and instructions to be executed by the processor 730, such as random access memory (RAM) or other dynamic storage devices. The main memory 732 can also be used to store temporary variables or other intermediate information during execution of instructions by the processor 730. When such instructions are stored in a non-transitory storage medium accessible to the processor 730, the computer system 700 becomes a special-purpose machine customized to perform the operations specified in the instructions.

[0067] The computer system 700 further includes a read-only memory (ROM) 734 or other static storage device coupled to the bus 740 for storing static information and instructions for the processor 730. A storage device 736, such as a magnetic disk or optical disk, is provided and coupled to the bus 740 for storing information and instructions.

[0068] The computer system 700 can be coupled via a bus to a display, such as a cathode ray tube (CRT), for displaying information to a computer user. An input device including alphanumeric keys and other keys is coupled to the bus for communicating information and command selections to the processor 730. Another type of user input device is a cursor control, such as a mouse, trackball, or cursor direction keys, which is used to convey direction information and command selections to the processor 730 and to control the movement of a cursor on the display. This input device typically has two degrees of freedom in two axes (a first axis (e.g., x) and a second axis (e.g., y)), which allows the device to specify a position in a plane.

[0069] The computer system 700 can implement the techniques described herein using custom hardwired logic, one or more ASICs or FPGAs, firmware, and / or program logic, which in combination with the computer system causes or programs the computer system 700 to be a special-purpose machine. According to one embodiment, the techniques herein are performed by the computer system 700 in response to execution by the processor 730 of one or more sequences of one or more instructions contained in the main memory 732. Such instructions can be read into the main memory 732 from another storage medium, such as the storage device 736. Execution of the instruction sequence contained in the main memory 732 causes the processor 730 to perform the processing steps described herein. In an alternative embodiment, hardwired circuitry may be used in place of or in combination with software instructions.

[0070] As used herein, the term "storage medium" refers to any non-transitory medium that stores data and / or instructions that cause a machine to operate in a particular fashion. Such storage media may include non-volatile media and / or volatile media. Non-volatile media includes, for example, optical or magnetic disks, such as the storage device 736. Volatile media includes dynamic memory, such as the main memory 732. Common forms of storage media include, for example, floppy disks, flexible disks, hard disks, solid state drives, magnetic tape, or any other magnetic data storage media, CD-ROM, any other optical data storage media, any physical media with hole patterns, RAM, PROM, and EPROM, flash-EPROM, NVRAM, any other memory chip or cartridge, content addressable memory, and ternary content addressable memory (TCAM).

[0071] Storage media is different from transmission media but can be used in combination with transmission media. Transmission media participates in the transfer of information between storage media. For example, transmission media includes coaxial cables, copper wire, and fiber optics, including the wires that make up a bus. Transmission media can also take the form of acoustic or light waves, such as acoustic or light waves generated during radio wave and infrared data communications.

[0072] Computer system 700 also includes a communication interface 738 coupled to the bus. The communication interface 738 provides for bi-directional data communication coupled to a network link, such as a network link (not shown) connected to a local network. As another example, the communication interface 738 can be a local area network (LAN) card to provide a data communication connection to a compatible LAN. A wireless link can also be implemented. In any such implementation, the communication interface 738 sends and receives electrical, electromagnetic, or optical signals that carry digital data streams representing various types of information.

[0073] Computer system 700 can send messages and receive data, including program code, through the network, network link, and communication interface 738. In an Internet example, a server can transmit request code for an application through the Internet, an ISP, a local network, and the communication interface 738.

[0074] The received code can be executed by the processor 730 when it is received, and / or stored in the storage device 736 or other non-volatile memory for later

[0075] Figure 8 is an example charged particle microscope environment 800 for performing at least a portion of the methods disclosed herein and in accordance with embodiments of the present disclosure. A charged particle microscope (CPM) environment 800 can include a charged particle microscope 850, a network 860, a user station 880, and a server 870. The various components of the CPM environment 800 can be co-located at a user's location or distributed. Additionally, some or all of the components will include a computing system, such as computing system 700, for performing the methods disclosed herein. Of course, the CPM environment is only an example operating environment for implementing the disclosed technology and should not be considered limiting of the implementation of the technology.

[0076] The CPM 850 can be any type of charged particle microscope, such as a TEM, SEM, STEM, dual beam system, or a focused ion beam (FIB) system. A dual beam system is a combination of an SEM and an FIB that allows for imaging and material removal / deposition. Of course, the type of microscope is a non-limiting aspect of the present disclosure, and the techniques disclosed herein can also be implemented on images obtained through other forms of microscopy and imaging. The CPM 850 can be used to obtain images of a sample and the ROI contained therein for processing with the methods disclosed herein. However, the disclosed methods can also be implemented by the CPM environment 800 on images obtained by other microscopes.

[0077] Network 860 can be any type of network, such as a local area network (LAN), a wide area network (WAN), the Internet, etc. Network 860 can be coupled between other components of the CPM environment 800 such that data and processing code can be transmitted to various components to implement the disclosed techniques. For example, a user at user station 880 can receive an image from CPM 850, which is intended to be processed according to the disclosed techniques. Subsequently, the user can retrieve code from server 870 directly or via network 860 to perform image processing and metering. Additionally, the user can initiate the process by providing the image from CPM 850 to server 870 directly or via network 860, such that processing and metering are performed by server 870.

[0078] In some instances, a numerical value, procedure, or device is referred to as "lowest", "optimal", "minimum", etc. It will be understood that such descriptions are intended to indicate that a selection can be made among a number of alternative functions in common use, and that such selection need not be better, smaller, or preferred to other selections. Additionally, the selected value can be obtained by numerical or other approximation means and can be only an approximation of the theoretically correct / value.

Claims

1. A method, which includes: extracting a region of interest from an image, the region of interest including one or more boundaries between different segments of the region of interest; generating a multi-scale dataset of the region of interest based on the region of interest; initializing a model of the region of interest, the initialized model at least determining a first boundary and a second boundary within the region of interest; optimizing a plurality of active contours within the region of interest based on the model of the region of interest and further based on the multi-scale dataset, the optimized plurality of active contours identifying one or more boundaries within the region of interest; and performing metrology on the region of interest based on the identified boundaries, wherein performing metrology on the region of interest includes automatically performing geometric analysis of the different segments separated by the one or more boundaries.

2. The method according to claim 1, further comprising at least enhancing the extracted region of interest based on one or more filters, and generating multi-scale data for the enhanced region of interest.

3. The method according to claim 2, wherein generating a multi-scale dataset of the region of interest based on the region of interest includes generating a plurality of image resolution levels of the region of interest using one or more scale spaces, wherein the one or more scale spaces are selected from a Gaussian scale space, a geometric scale space, a non-linear scale space, and an adaptive spatial scale space.

4. The method according to claim 2, wherein at least enhancing the extracted region of interest based on one or more filters includes: improving the contrast of at least the region of interest within the image; and improving the signal-to-noise ratio of at least the region of interest within the image.

5. The method according to claim 4, wherein improving the contrast and the signal-to-noise ratio includes automatically selecting and applying the one or more filters, wherein the one or more filters are selected from histogram manipulation, linear and non-linear contrast enhancement, data normalization based on local low-frequency data distribution, gamma correction, logarithmic correction, and brightness correction.

6. The method according to claim 1, wherein initializing a model of the region of interest, the initialized model at least determining a first boundary and a second boundary of the region of interest includes initializing the model of the region of interest based on one or more image maps selected from a binary-labeled map, an interactive map, a distance map, a CAD map, a statistical model from data, dye casting, a random distribution of geometric shapes, and a geometric model.

7. The method according to claim 1, further includes: improving the clarity of the region of interest; distinguishing and detecting one or more boundaries in the region of interest; and wherein the model of the region of interest is initialized at least in part based on the one or more boundaries distinguished and detected in the region of interest.

8. A method, which includes: extracting a region of interest from an image, the region including one or more boundaries between different segments of the region of interest; generating a multi-scale dataset of the region of interest based on the region of interest; Initialize the model of the region of interest, and the initialization model determines at least a first boundary and a second boundary of the region of interest; Based on the model of the region of interest and further based on the multi-scale data set, optimize multiple active contours within the region of interest, and the optimized multiple active contours identify one or more boundaries within the region of interest; Wherein optimizing multiple active contours within the region of interest based on the model of the region of interest and further based on the multi-scale data set includes: Initialize a first plurality of active contours within the first boundary and the second boundary of the initialization model, wherein there are more active contours in the first plurality of active contours than the number of boundaries within the region of interest; and Allow the first plurality of active contours to optimize to the multiple active contours to identify one or more boundaries within the region of interest; and Perform metrology on the region of interest based on the identified boundaries.

9. A non-transitory computer-readable medium storing instructions that, when executed by one or more processors, trigger the one or more processors to perform the following steps: Extract a region of interest from an image, the region including one or more boundaries between different segments of the region of interest; Generate a multi-scale data set of the region of interest based on the region of interest; Initialize the model of the region of interest, and the initialization model determines at least a first limit and a second limit within the region of interest; Based on the model of the region of interest and further based on the multi-scale data set, optimize multiple active contours within the region of interest, and the optimized multiple active contours identify one or more boundaries within the region of interest; And Perform metrology on the region of interest based on the identified boundaries, wherein performing metrology on the region of interest includes automatically performing geometric analysis of the different segments separated by the one or more boundaries.

10. The non-transitory computer-readable medium according to claim 9, optimizing multiple active contours within the region of interest based on the model of the region of interest and further based on the multi-scale data set includes: Initialize a first plurality of active contours within the first limit and the second limit of the initialization model; And Allow the first plurality of active contours to optimize to the multiple active contours to identify one or more boundaries within the region of interest.

11. The non-transitory computer-readable medium according to claim 10, wherein there are more active contours in the first plurality of active contours than the number of boundaries within the region of interest.

12. The non-transitory computer-readable medium according to claim 9, wherein the instructions further trigger the processor to enhance the performance of the extracted region of interest based on one or more filters, and wherein the multi-scale data is at least partially generated based on the enhanced region of interest.

13. The non-transitory computer-readable medium according to claim 12, wherein generating the multi-scale dataset of the region of interest based on the enhanced region of interest includes generating multiple image resolution levels of the enhanced region of interest using one or more scale spaces selected from a Gaussian scale space, a geometric scale space, a non-linear scale space, and an adaptive spatial scale space.

14. The non-transitory computer-readable medium according to claim 9, wherein enhancing the extracted region of interest based on one or more filters includes: improving the contrast of at least the region of interest within the image; and improving the signal-to-noise ratio of at least the region of interest within the image.

15. The non-transitory computer-readable medium according to claim 14, wherein improving the contrast and the signal-to-noise ratio includes automatically selecting and applying the one or more filters selected from histogram manipulation, linear and non-linear contrast enhancement, data normalization based on local low-frequency data distribution, gamma correction, logarithmic correction, and brightness correction.

16. The non-transitory computer-readable medium according to claim 9, wherein initializing the model of the region of interest, and initializing the model to determine at least a first boundary and a second boundary of the region of interest includes initializing the model of the region of interest based on one or more image maps selected from a binary-labeled map, an interactive map, a distance map, a CAD map, a statistical model from data, dye casting, a random distribution of geometric shapes, and a geometric model.

17. The non-transitory computer-readable medium according to claim 9, wherein the instructions further trigger the processor to cause the following performance: improving the sharpness of the region of interest; distinguishing and detecting one or more boundaries in the region of interest; and wherein the model of the region of interest is initialized at least in part based on the one or more boundaries distinguished and detected in the region of interest.

Images