Method and spectroscopic apparatus for acquiring and forming a spectroscopic image

Abstract

The invention relates to a method of acquiring and forming a spectrographic image (IS) of a sample (1), comprising the steps of: e1) acquiring an initial image (II) of a region of the sample, the initial image consisting of pixels and defining a maximum set of N spectrographic measurement positions (PM), 2 ≤ N, each measurement position comprising coordinates (CX; CY) and intensity (IR1; IR2; IG1; IG2; IB1; IB2) determined on a pixel basis; e2) assigning a classification value to each of the N measurement positions based on a bias calculated from intensity and coordinate differences between measurement positions; e3) determining a set of P measurement positions from the classification values, 1 ≤ P ≤ N; e4) for each measurement position of the set, successively positioning an excitation light beam (2) at said measurement position of the region of the sample, acquiring a spectrographic measurement and forming a spectrographic image.

Description

Technical Field

[0001] This invention generally relates to the field of methods and instruments for acquiring spectral or hyperspectral images of samples based on point-by-point measurement techniques using Raman spectroscopy, fluorescence, photoluminescence or anodization, and image reconstruction.

[0002] In particular, it relates to a method for acquiring and forming spectral or hyperspectral images of a sample.

[0003] In particular, this relates to a method for rapidly acquiring and reconstructing sample images using Raman microspectroscopy, wherein the high spatial resolution images are formed point by point.

[0004] It also relates to a spectroscopic apparatus designed to achieve this method. Background Technology

[0005] Specific spectroscopic measurements are usually sufficient to analyze spatially homogeneous samples. On the other hand, in the case of spatially heterogeneous samples, it is desirable to perform spectroscopic measurements in the form of point-by-point image construction with spatial resolution adjusted according to the sample's structure or microstructure.

[0006] Various spectral imaging techniques are known, based on point-by-point acquisition of spectra and spectral determination images based on reconstructed spatial resolution. These systems require either the displacement of the sample relative to the measuring instrument and an excitation beam focused at a point, or a scan of the excitation beam relative to the sample, or a combination of beam scanning and sample displacement.

[0007] In a sample displacement-based Raman microspectroscopic imaging system, the sample is placed on a motorized stage or equipped with a piezoelectric actuator to move the sample in two lateral directions (X and Y). A laser excitation beam is focused on a region approximately 1 micrometer in diameter on the sample. The sample displacement typically follows a periodic grid with a constant pitch in the X and Y directions. Spectra are acquired at each new position of the sample. Processing software reconstructs a hyperspectral image based on this set of measurements.

[0008] Regardless of the displacement type chosen, the acquisition duration for point-by-point imaging is determined by the acquisition duration for each point and the number of points in the image. The acquisition time for each point and the photometric intensity of each point are related to the spectral resolution. The number of points in the image depends on the displacement step size of the sample relative to the excitation beam or the beam relative to the sample, which determines the spatial resolution of the image.

[0009] Acquiring a single Raman spectrum typically takes 0.1 milliseconds to 1 minute, averaging 1 second. To obtain images with high spatial and spectral resolution, the acquisition time for Raman microspectroscopy can reach several hours or even tens of hours. These times are unsuitable for analyzing numerous samples.

[0010] The main options for reducing the time required to acquire spectroscopic images are to reduce the acquisition time per point and to reduce the number of points in the image.

[0011] However, these options typically result in a loss of quality in the reconstructed image, a decrease in the signal-to-noise ratio at each point, and / or a reduction in the spatial resolution of the reconstructed image.

[0012] In some applications, it is desirable to reduce the acquisition time of microspectroscopic or hyperspectral images without sacrificing the luminosity, signal-to-noise ratio, spectral resolution, or spatial resolution of the reconstructed images.

[0013] In other applications, it is desirable to improve the spatial resolution of hyperspectral or microspectral images without increasing the image acquisition time. Summary of the Invention

[0014] In this context, the present invention proposes a method for acquiring and forming a spectral image of a sample, comprising the following steps:

[0015] e1) Obtain an initial image of the sample region, the initial image consisting of pixels, and define a maximum set of N spectral measurement locations, 2≤N, each measurement location including at least one coordinate and at least one intensity, the coordinate representing the measurement location relative to the sample region, and the intensity being determined based on at least one pixel in the initial image located in the neighborhood of the measurement location;

[0016] e2) Based on the deviation between the measurement locations, a classification value is assigned to each of the N measurement locations. The deviation between the two measurement locations is calculated based on the intensity difference and the coordinate difference.

[0017] e3) Based on the classification values ​​assigned in step e2), determine a set of P measurement locations from the set of N measurement locations, where 1≤P≤N;

[0018] e4) For each measurement location in the group determined in step e3), the excitation beam is sequentially positioned at the measurement location in the region of the sample to acquire spectral measurements and construct a spectral image based on the acquired spectral measurements.

[0019] Here, the measurement position represents the specific position of the excitation beam relative to the sample for spectral measurement.

[0020] Therefore, due to the method of forming and acquiring spectroscopic images, classification values ​​are assigned to measurement locations. The classification value of a measurement location can quantify the contribution of that measurement location to the formation of the spectroscopic image. The contribution of a measurement location is an estimate of the weight (i.e., importance) of the spectroscopic measurements acquired at that location in the formation of the spectroscopic image.

[0021] The contribution of the measurement location is determined based on the initial image. The contribution of the measurement location is determined in relation to its position on the sample. For example, a measurement location located within a region of strong intensity variation in the sample is considered to contribute more than a measurement location located within a region of constant intensity. In other words, regions of uniform intensity in the sample can be accurately estimated through spectroscopic measurements, while regions exhibiting strong spatial variations in intensity must be estimated through several spectroscopic measurements.

[0022] Therefore, this method, by assigning them classification values, allows measurement locations to be organized or arranged, for example, from the measurement location that contributes the most to the measurement location that contributes the least.

[0023] Once the measurement locations are organized, spectral measurements can be acquired starting from the location that contributes the most. The spectral image can then be progressively built and updated with each newly acquired spectral measurement.

[0024] Therefore, during the acquisition and formation of spectroscopic images, regions in the sample exhibiting strong spatial variations in intensity are preferentially acquired with higher spatial resolution (on the spectroscopic image) than regions in the sample with spatially constant intensity. Consequently, even after several acquisitions, the quality of the spectroscopic image quickly becomes satisfactory.

[0025] Therefore, prioritizing the acquisition of spectroscopic measurements at the measurement locations that contribute the most makes it possible to obtain high-quality spectroscopic measurements within a given acceleration factor or a given acquisition time. For example, for a given acceleration factor, this method may be able to obtain spectroscopic images of better quality than acquiring measurements at locations distributed along a predetermined regular grid.

[0026] Users can stop acquiring new spectral measurements when they are satisfied with the image quality.

[0027] Once a certain acceleration factor is reached, acquiring new spectroscopic measurements can be stopped. The acceleration factor, sometimes referred to as the compression ratio, is defined as the ratio between the maximum number of measurement locations based on the spatial resolution of the spectrometer and the number of spectroscopic measurements acquired. Therefore, a given acceleration factor is inversely proportional to a given number of measurements acquired.

[0028] Other non-limiting and advantageous features of the method according to the invention, individually or based on all technically possible combinations, are as follows:

[0029] - The classification values ​​belonging to step e2) are all different from each other;

[0030] - The deviation between two measurement locations is calculated based on the weighted sum of the squares of the intensity differences between the two measurement locations and the squares of the coordinate differences between the two measurement locations;

[0031] -In step e3), the P measurement locations of the group are determined according to the ascending or descending order of the classification values;

[0032] - In step e2), initialize lists S and U, where list S includes measurement locations that still need to be assigned classification values ​​and list U includes measurement locations that have already been assigned classification values, and iterate through the following steps e21) and e22):

[0033] e21) For each measurement location in list S, determine the total error through iterative steps e211), e212), and e213):

[0034] e211) Move the measurement location under consideration from list S to list U;

[0035] e212) Calculate the total error associated with the measured locations under consideration based on the deviation between the measured locations from list S and the measured locations from list U;

[0036] e213) Move the measurement location under consideration from list U to list S;

[0037] e22) Based on each total error determined in step e21), select measurement locations from list S, assign numerical classifications to the selected measurement locations, and move the selected measurement locations from list S to list U;

[0038] - The classification value is initialized to a predetermined value, and in each iteration of step e22), the classification value that is strictly lower than the lowest classification value or strictly higher than the highest classification value is assigned to the measurement location with the lowest total error;

[0039] - In step e212), determining the total error associated with the considered measurement location includes the following steps:

[0040] e2121) For each measurement location in list U, determine the measurement location with the smallest deviation in list S. The deviation between these two measurement locations is defined as the minimum deviation.

[0041] e2122) Calculate the total error based on each minimum deviation determined in step e2121);

[0042] - The group includes at least three measurement locations and a predetermined trajectory for continuous positioning of the excitation beam;

[0043] - Iteration steps e3) and e4) In each new iteration, the group determined in step e3) only includes the measurement locations for which spectral measurements still need to be obtained;

[0044] - The method includes step e11), acquiring at least one spectral measurement at at least one predetermined measurement location, and assigning a predetermined classification value to each of the at least one predetermined measurement location;

[0045] - The construction of the spectroscopic image includes the following steps:

[0046] e41) For each measurement location for which a spectral measurement has been obtained, determine the spectral measurement value based on the spectral measurement;

[0047] e42) For each remaining measurement location, a spectral measurement value is determined based on at least one selected spectral measurement, and the selected spectral measurement is selected according to the coordinate difference between the remaining measurement location and the measurement location of the at least one selected spectral measurement.

[0048] - In step e42), the selected spectral measurement is a spectral measurement obtained at a measurement location where the coordinate difference from the remaining measurement location is less than a threshold.

[0049] - In step e42), the spectral measurement value is determined based on the weighted average of the deviations between the measurement location where the selected spectral measurement was acquired and the remaining measurement locations for each selected spectral measurement;

[0050] -The method includes step e43), interpolating the spectral measurement image, wherein multiple interpolation coefficients are determined based on the deviation between measurement locations;

[0051] - The method includes a stopping measure in step e4) to end the acquisition of additional spectroscopic measurements;

[0052] - The number of spectroscopic measurements to be obtained is predetermined;

[0053] - A region of interest, including only a portion of the measurement locations, is predefined in the initial image, and in step e3), this group is determined based on the measurement locations belonging to the region of interest.

[0054] - At least one spectroscopic measurement is spatially integrated over a sector of the sample;

[0055] -The initial image was acquired using one of the following imaging methods:

[0056] Spectral acquisition in one or more predetermined spectral bands; optical acquisition by reflection or transmission; hyperspectral acquisition; acquisition by atomic force microscopy; acquisition by scanning electron microscopy.

[0057] The present invention also proposes a spectroscopic measuring device, comprising:

[0058] - A light source designed to generate an excitation beam;

[0059] - A support suitable for receiving samples;

[0060] - Image capture device;

[0061] - A processor, programmed to perform steps e1) to e4) of the method to acquire and form a spectral measurement image.

[0062] Detailed description of the implementation method

[0063] The following description, with reference to the accompanying drawings, will, through non-limiting examples, make the content of the invention and how it is implemented clearly understandable.

[0064] Of course, different features, variations and implementations of the present invention can be associated with each other in various combinations, as long as they are not mutually incompatible or mutually exclusive.

[0065] In the attached diagram:

[0066] [ Figure 1 [ ] is a schematic diagram of a scanning beam imaging instrument;

[0067] [ Figure 2 [ ] is a schematic diagram of the spatial distribution grid of the measurement positions in the XY plane of the sample;

[0068] [ Figure 3 This is a flowchart of the steps involved in acquiring and forming a spectral measurement image;

[0069] [ Figure 4 The image shows the initial image of the sample;

[0070] [ Figure 5 ]yes Figure 4 An enlarged view of the initial image, showing the distribution of measurement locations;

[0071] [ Figure 6 ] represents an example of a classification of the measured location relative to the initial simulated image;

[0072] [ Figure 7A This explains the use of... Figure 3 The method constructed Figure 4 The Raman spectra of the sample were measured with an acceleration factor of 100.

[0073] [ Figure 7B This explains the use of... Figure 3 The method constructed Figure 4 The Raman spectra of the sample were measured with an acceleration factor of 50.

[0074] [ Figure 7C This explains the use of... Figure 3 The method constructed Figure 4 The Raman spectra of the samples were measured with an acceleration factor of 20.

[0075] [ Figure 7D This explains the use of... Figure 3 The method constructed Figure 4 The Raman spectra of the sample were measured with an acceleration factor of 10.

[0076] [ Figure 7E This explains the use of... Figure 3 The method constructed Figure 4 The Raman spectra of the sample were measured with an acceleration factor of 5.

[0077] [ Figure 7F This explains the use of... Figure 3 The method constructed Figure 4 The Raman spectra of the sample were measured with an acceleration factor of 1.

[0078] [ Figure 8 [Explanation] Figure 4 The initial image and the distribution of source pixels and their associated networks are obtained, with an acceleration factor of 20;

[0079] [ Figure 9 This explains that in Figure 4 The initial image and Figure 8 The estimated structural image is constructed based on the source pixel distribution, with an acceleration factor of 20;

[0080] [ Figure 10 ]yes Figure 7C A magnified view of the spectral measurement image, corresponding to Figure 9 The estimated structural image at the same acceleration factor of 20;

[0081] [ Figure 11A Explanation of Figure 7A The interpolation step performed on the Raman spectroscopy images;

[0082] [ Figure 11B Explanation of Figure 7B The interpolation step is performed on the Raman spectroscopy image.

[0083] Device

[0084] exist Figure 1The image schematically illustrates an example of an analytical instrument based on a scanning beam microscope, designed to sequentially position an incident excitation beam 2 at different points on a sample 1. This scanning microscope includes a stage system that displaces the sample 1 relative to the stationary excitation beam 2. This mode of movement of the excitation beam 2 relative to the sample allows for precise positioning of the excitation beam 2. Other microscopes include scanning devices (typically scanners) for changing the angle of incidence of the excitation beam relative to the sample, while the sample remains stationary as the excitation beam scans. Finally, some microscopes combine the scanning of the excitation beam with sample movement to cover an analytical surface with a larger spatial extent.

[0085] Generally, point-by-point spectral imaging microscopy is configured to perform relative displacement of the excitation beam 2 relative to the sample 1, so as to sequentially guide the excitation beam 2 to different points of the sample 1 to be analyzed. These different points of the sample 1 to be analyzed, where the excitation beam 2 can be placed, correspond here to measurement positions, denoted by PM. More precisely, the measurement positions form a maximum set of measurement positions. As detailed below, the method of this disclosure can select the measurement positions that contribute the most from these N measurement positions, preferentially acquire spectroscopic measurements at these positions, and determine the order in which these measurements are acquired. This method is practically designed to acquire an effective number of spectroscopic measurements below the maximum number N to reduce the total acquisition time while producing high-quality spectroscopic images, that is, high resolution in the maximum contributing region of the sample.

[0086] The analytical instrument also includes a spectrometer, which makes it possible to acquire spectral signals at each measurement location PM. The spectral signal at a point on sample 1 is referred to as a spectral measurement.

[0087] The analytical instrument ultimately includes a calculator and processing software used to generate hyperspectral images or even one or more images of different spectral bands or different wavelengths of sample 1 based on spectroscopic measurements. Therefore, each point in the hyperspectral image represents a spectroscopic measurement, such as Raman scattering, fluorescence, photoluminescence, or anodizing.

[0088] Figure 2 This illustrates a typical example of a grid for the measurement position PM where the excitation beam 2 should be positioned. The measurement position PM is represented by a measurement disk or point 6 on the orthogonal reference frame plane (XY) of the sample surface. The grid is generally regular, rectangular or square, with a pitch DX along the X-axis and a pitch DY along the Y-axis. The pitch DX is usually chosen to be equal to the pitch DY.

[0089] The classic method for forming a spectral image is to place the incident beam 2 at the measurement position PM(1,1) and acquire the first spectral measurement, and then move the excitation beam to the measurement position PM(1,2) to acquire the second spectral measurement. The operation of repositioning the excitation beam 2 relative to the sample 1 and acquiring spectral measurements can then be repeated until spectral measurements are acquired at each measurement position PM on the scanning surface 4 of the sample 1.

[0090] The spatial resolution of the resulting image depends on the size of the excitation beam 2 on sample 1 and the selected displacement pitches DX and DY.

[0091] The microscope may be an optical microscope, including a microscope objective 3, to focus the laser beam 2 onto the sample 1 to be analyzed. The optical microscope is combined with a Raman spectrometer, or with optical fluorescence or photoluminescence respectively, and is configured to measure Raman scattering spectra or optical fluorescence or photoluminescence spectra in the ultraviolet-visible range respectively.

[0092] Alternatively, the microscope can be an X-ray fluorescence microscope, including an X-ray beam focusing and scanning system. The X-ray fluorescence microscope is then combined with a spectrometer configured to measure X-ray spectra (XRF).

[0093] Alternatively, the microscope can be a scanning electron microscope (SEM), which includes an electron beam focusing and scanning system. The scanning electron microscope is then combined with a spectrometer, configured to measure X-ray spectra (EDX), or separately to measure cathode ray spectra.

[0094] These analytical instruments can acquire point-by-point spectral measurements of sample 1 on scanning surface 4, with spatial resolution determined in particular by the size of the incident excitation beam on the sample.

[0095] Point-by-point spectral imaging provides spatial and spectral resolution images of a sample. For example, the application of Raman microscopy makes it possible to analyze the spatial distribution of chemical components within a region of a pharmaceutical tablet, with a spatial resolution of approximately λ, where λ represents the wavelength of the excitation beam.

[0096] Traditionally, the spatial resolution of spectral or hyperspectral images acquired by scanning microscopy is spatially uniform across the entire surface being scanned and analyzed. The acquisition time for scanning microspectrometry images increases with the number of points in the resulting spectral image. In practice, the acquisition time is approximately equal to the product of the number of spectral measurements and the acquisition time for those measurements.

[0097] The analytical instrument of the present invention may also advantageously include another imaging device, making it possible to acquire an initial image of a region of the sample under study with high spatial resolution. This initial structural image may be a video image acquired by a CCD camera combined with a scanning microscope. In SEM, a secondary electron detector can quickly provide an initial high spatial resolution image. However, certain structural details detected on a high-resolution video image, such as local variations in luminosity, can indicate variations in measurements acquired by spectroscopy, although video images generally do not provide information about the type of spectral determination.

[0098] method

[0099] Figure 3 A method for acquiring and forming a spectroscopic image IS of sample 1 is shown.

[0100] Step e1)

[0101] like Figure 3 As shown, the method includes step e1, namely, acquiring an initial image II of the region of sample 1, the initial image II consisting of pixels, and defining a maximum set of N (2≤N) spectral measurement positions PM, each measurement position PM including at least one coordinate (representing the measurement position PM relative to the region of the sample) and at least one intensity (determined based on at least one pixel in the initial image II located in the neighborhood of the measurement position PM).

[0102] The acquisition of the initial image II, as referred to here, means detecting an image of sample 1 on an image sensor, such as a CCD-type image, or even retrieving an image file of sample 1 acquired on another metrology instrument, or obtaining it from a database containing images of sample 1. Preferably, an initial digital image II is used. If the detected or collected initial image II is analog, it is digitized to obtain a digital initial image II of sample 1. If necessary, the size of the initial image II is adjusted according to the scanning excitation beam 2 and / or the area of ​​sample 1 to be analyzed that is moved relative to the excitation beam 2.

[0103] In practice, the initial image II can be obtained using one of the following imaging methods:

[0104] - Spectral acquisition in one or more predetermined spectral bands;

[0105] - Optical acquisition through reflection or transmission;

[0106] - Hyperspectral acquisition;

[0107] - Obtained using an atomic force microscope;

[0108] - Obtained using a scanning electron microscope.

[0109] Initial Image II can be acquired using the spectrometer's own spectrometer, for example, if Initial Image II is a spectrometer image with a single spectral band. Other modes require additional imaging equipment.

[0110] For example, here, the initial image II is a video image acquired via a CCD sensor and by illuminating the sample with white light. The settings of the illumination and detection systems are adjusted to increase image contrast and bring out structural details. Advantageously, the intensity, polarization, and / or illumination and / or detection wavelengths are adjusted to obtain a highly contrasting initial structural image.

[0111] An example of this initial image II is... Figure 4 The text appears to be a mix of Chinese characters and symbols, possibly representing a corrupted or incomplete sentence. A direct translation wouldn't be Figure 4 In the sample, sample 1 represents the dark letters J, O, B, I, and N on a light background.

[0112] Here, the initial image II is an RGB image, and the intensity of each pixel consists of three components: red, green and blue, all of which are 256-bit encoded.

[0113] Regardless of how the initial image II was acquired, it consists of pixels with one or more intensities; in this case, there are three intensities. Here, the pixels are conventionally distributed in rows and columns.

[0114] Preferably, the acquisition time of the initial image II is shorter than that of the acquisition time of the spectroscopic image IS. For example, the initial image II is acquired in a fraction of a second, while the spectroscopic image IS is acquired in a few minutes.

[0115] In step e1, a maximum set of N (2≤N) spectral measurement positions PM is defined. Here, the maximum number of measurement positions PM depends on the pitch DX along the X-axis and the pitch DY along the Y-axis. The pitches DX and DY are selected based on the resolution of the analytical instrument. For example, the pitches DX and DY can be determined by the diameter of the measurement point 6 and the accuracy of the relative displacement of the excitation beam 2 of the analytical instrument relative to the sample 1.

[0116] Here, the measurement location PM is therefore defined by a grid with pitches DX and DY, the size of which corresponds to the area of ​​sample 1 to be analyzed.

[0117] Each measurement position PM is identified relative to sample 1 by at least one coordinate. Here, the measurement position is identified by two coordinates: a CX coordinate along the X-axis and a CY coordinate along the Y-axis.

[0118] The initial image II is preferably a high-resolution image. In practice, the resolution of the initial image II is at least as high as the grid pitch of the measurement position PM. Then the coordinates of the measurement position PM can correspond to pixels in the initial image II. If the resolution of the initial image II is higher than the grid pitch of the measurement position PM, the coordinates of the measurement position PM can correspond to several pixels.

[0119] exist Figure 5 In the image, measurement locations PM, indicated by "+", are superimposed on the initial image II. To avoid excessive content in the image, only a portion of the measurement locations PM are shown. In practice, the grid of the measurement locations PM can be larger than... Figure 5 The examples in the text are at least 4 or 5 times denser.

[0120] For example, in a given direction, two measurement positions PM can be approximately 4 pixels apart. Thus, the resolution of the initial image II is about 16 times greater than the maximum spatial resolution of the spectroscopic image IS.

[0121] Each measurement location PM also includes at least one intensity I. The intensity of the measurement location PM is calculated based on the intensity of pixels in the neighborhood of the measurement point in the initial image II.

[0122] For example, a pixel is in the nearest neighborhood to the measurement location PM. The number of pixels in the neighborhood of measurement location PM can be estimated by the ratio between the number of pixels in the initial image II and the maximum number N of measurement locations PM. Again, as an example, the neighborhood can depend on the size of measurement point 6. Therefore, the neighborhood of measurement location PM can be defined by all pixels included within a circle centered on measurement location PM, with a size substantially similar to that of measurement point 6.

[0123] Here, the intensity of the measurement location PM is calculated by interpolating (e.g., averaging or bicubic interpolation) the intensity of the pixels PI located in its neighborhood V. In other words, the initial image II is subsampled such that the number of subsampled pixels equals the maximum number N of measurement locations PM. Once subsampling is performed, the intensity of the measurement location PM corresponds one-to-one with the intensity of a pixel in the initial image.

[0124] Therefore, the intensity of the measurement location PM has the same properties as the intensity of the pixels in the initial image II. Then, the intensity of the measurement location PM can have the same number of components as the pixel intensity. Here, for example, the intensity of the measurement location PM includes three components: a red component IR, a green component IG, and a blue component IB.

[0125] In one implementation, the intensity of the measured location PM has additional components, such as a component representing the average intensity of neighboring pixels, a component representing the maximum intensity of neighboring pixels, and even a component representing the minimum intensity of neighboring pixels. Such additional components can be defined for each color: red, green, and blue.

[0126] Step e2)

[0127] Once the coordinates and intensity of PM at each measurement location are determined, the method can assign a classification value to PM at each measurement location.

[0128] Therefore, such as Figure 3 As shown, the method includes step e2, which assigns a classification value to each of the N measurement locations PM based on the deviation between the measurement locations PM, wherein the deviation between two measurement locations PM is calculated based on the intensity difference and the coordinate difference.

[0129] Step e2 allows all measurement locations PM to be sorted or classified according to their classification values, from the largest contributor to the smallest contributor (or vice versa). As explained in the introduction, the classification values ​​of the measurement locations PM enable the quantification of their contribution to the formation of the spectroscopic image IS.

[0130] The contribution of the measurement location PM represents the importance or interest in acquiring spectral measurements at that location to improve the quality of the spectral image IS. Spectral measurements are then acquired based on classification values, typically starting with the location contributing the most. The spectral image is then reconstructed and updated in real time based on the actual acquired spectral measurements.

[0131] For example, at a given time, when j (1≤j≤N) spectroscopic measurements have been acquired, the (j+1)th measurement position PM that contributes the most is the measurement position where acquiring the (j+1)th spectroscopic measurement would increase the quality of the spectroscopic image IS the most. As described below, the quality of the spectroscopic image IS at j+1 measurements can be estimated based on the initial image II. The quality of the spectroscopic image IS with j measurements can correspond to the difference between it and the spectroscopic image IS with N measurements, that is, the difference between it and the spectroscopic image IS acquired with an acceleration factor of 1.

[0132] This classification is based on the coordinates and intensity of the PM at the measured location. Therefore, this classification is based on the initial image II.

[0133] The deviation between two measurement locations is calculated based on the difference in intensity and coordinates. Therefore, the deviation represents the spatial and intensity similarity between the two measurement locations (PM). The smaller the deviation, the more similar the measurement locations (PM) are.

[0134] In practice, if the deviation between the two measurement locations (PM) is small, obtaining spectral measurements from only one of the two PM locations is sufficient to construct a spectral image. The missing spectral measurements can be estimated based on the acquired ones. Thus, one of the two PM locations can be considered to contribute little, as obtaining spectral measurements there is not a priority.

[0135] To calculate the deviation between two measurement locations PM, the calculator performs a weighted sum of the squares of the intensity difference between the two measurement locations and the squares of the coordinate difference between the two measurement locations.

[0136] For example, here, the first measurement position PM is located at coordinates CX1, CY1, and its intensity components are IR1, IG1, IB1; the second measurement position PM is located at coordinates CX2, CY2, and its intensity components are IR2, IG2, IB2; the deviation E can be given by the formula:

[0137] E = (IR1 - IR2) 2 +(IG1-IG2) 2 +(IB1-IB2) 2 +R×((CX1-CX2) 2 +(CY1-CY2) 2 ).

[0138] Here, the intensity components IR1, IG1, IB1, IR2, IG2, and IB2 are 8-bit encoded values ​​between 0 and 255. The coordinates CX1, CY1, CX2, and CY2 define the measurement location as follows: Figure 2 The exponent values ​​on the grid shown. Here, the coordinates CX1, CY1, CX2, and CY2 are integers, for example, between the square roots of 1 and N for a square grid.

[0139] Alternatively, the deviation can be calculated as a weighted sum of the intensity difference between the two measurement locations and the squares of the coordinate differences between the two measurement locations. As a variation, the deviation can also be calculated as a weighted sum of the average intensity at the two measurement locations and the average coordinates between the two measurement locations.

[0140] Here, the weighting factor R allows the coordinate difference to be weighted relative to the intensity difference. The weighting factor R can be manually adjusted by the user or automatically adjusted based on the contrast variations present in the initial image II. In practice, for an initial image including an intensity difference greater than 100 between bright and dark pixels, the weighting factor R can take a value between 30 and 100.

[0141] The weighting factor R used to calculate the deviation remains unchanged in step e2, that is, for the classification of all measurement locations PM.

[0142] Advantageously, the weighting coefficient R allows the bias calculation to be adapted to sample 1. For example, the higher the spatial frequency of the initial image II, the lower the weighting coefficient R. Then the similarity between the two measurement locations PM is primarily based on intensity. Again, as an example, if the contrast variation in a relatively large portion of the initial image II is small, the weighting coefficient can be high. Then the similarity between the two measurement locations PM is primarily based on coordinates.

[0143] The weighting coefficient R can also be proportional to the overall contrast of the initial image II to maintain a balance between the coordinate difference and the intensity difference. This makes it possible, for example, to assign a significant weight to the coordinate difference even when the intensity difference is very large.

[0144] In embodiments where the intensity at the measurement location has an additional component, the deviation calculation takes these additional components into account.

[0145] Advantageously, during step e2, the calculator constructs a list S, which includes measurement locations PM whose classification values ​​are yet to be assigned, that is, measurement locations that are yet to be sorted. The calculator also constructs a list U, which consists of measurement locations PM whose classification values ​​have already been assigned (that is, measurement locations PM that have been sorted).

[0146] For example, after defining N measurement locations PM, all measurement locations PM still need to be sorted. Then, list S contains N measurement locations PM, and list U contains no measurement locations PM. Alternatively, if the calculator assigns category values ​​to j measurement locations, then list S contains Nj measurement locations PM, and list U contains j measurement locations PM.

[0147] In the first implementation, such as Figure 3 As shown, in order to assign a classification value to each measurement location PM in list S, the calculator iterates through steps e21 and e22:

[0148] e21) For each measurement location PM in list S, determine the total error;

[0149] e22) Based on each total error determined in step e21), select measurement location PM from list S, assign a classification value to the selected measurement location PM, and move the selected measurement location PM from list S to list U.

[0150] At the end of step e21, each measurement location PM in list S is associated with the total error.

[0151] For example, if steps e21 and e22 have been performed j-1 times, this means the calculator has assigned classification values ​​to j-1 measurement locations PM. In iteration j of step e21, the calculator determines j total errors. The determination of the total errors will be described further below. Then, in iteration j of step e22, based on the total errors, the calculator selects a measurement location PM from the j measurement locations PM in list S and assigns a classification value to the selected measurement location PM.

[0152] More specifically, during the iteration of step e22, the calculator selects the measurement position PM that is associated with the lowest total error.

[0153] As described below, the total error here represents the contribution of all the measurement locations PM that have been classified (i.e., those measurement locations that have been assigned classification values). During the iteration of step e22, the contribution of the selected measurement location PM can be estimated, for example, by the difference between the total error before classification (i.e., before moving the measurement location into list U) and the total error after classification (i.e., after moving the measurement location into list U).

[0154] Here, prior to the iterations in steps e21 and e22, the classification values ​​are initialized to predetermined values, and in each iteration of step e22, classification values ​​that are strictly below the lowest classification value or strictly above the highest classification value are assigned to the measurement location PM associated with the lowest total error. Two examples of classification value assignment are described in detail below.

[0155] The calculator then assigns classification values ​​to the selected measurement locations. For example, the calculator assigns only classification values ​​that are strictly higher than the previous classification value or only classification values ​​that are strictly lower than the previous classification value. The previous classification value is either a classification value that has already been assigned to a previous iteration or is equal to a predetermined initial value.

[0156] If the classification value assigned by the calculator is significantly greater than the previous classification value, then at the end of step e2, the measurement locations PM are sorted from the smallest to the largest contribution according to the increased classification value.

[0157] In practice, to achieve this, the classification values ​​can be initialized to zero. Then, each selected measurement location (PM) is assigned a classification value incremented by 1 from the previous value. Therefore, the measurement location PM with the smallest contribution has a classification value of 1, and the PM with the largest contribution has a classification value of N.

[0158] If the classification value assigned by the calculator is significantly lower than the previous classification value, then at the end of step e2, the measurement location PM is sorted from the largest contributor to the smallest contributor according to the increase of the classification value.

[0159] In practice, to achieve this, the classification values ​​can be initialized to N+1. Then, each selected measurement location PM is assigned a classification value that increments by -1 from the previous classification value. Thus, the measurement location PM with the smallest contribution has a classification value of N, and the measurement location PM with the largest contribution has a classification value of 1.

[0160] Once the calculator assigns a classification value to the selected measurement location PM, the selected measurement location PM is moved from list S to list U. After j iterations of steps e21 and e22, list S contains Nj measurement locations PM, and list U contains j measurement locations PM.

[0161] To assign a classification value to each measurement location PM in list S, the calculator iterates through steps e21 and e22 until there are no more measurement locations PM in list S. This corresponds to N iterations of steps e21 and e22.

[0162] In this first embodiment, in step e21, the total error associated with the considered measurement location PM in list S is determined by iterating through steps e211, e212, and e213:

[0163] e211) Move the considered measurement location PM from list S to list U;

[0164] e212) Based on the deviation between the measurement location PM in list S and the measurement location PM in list U, calculate the total error associated with the considered measurement location PM;

[0165] e213) Move the considered measurement location PM from list U to list S.

[0166] Here, in step e221, the total error associated with the considered measurement position PM is calculated based on the deviation between each measurement position PM in list U and each measurement position PM in list U, or between each measurement position PM in list U and the measurement position PM in list S.

[0167] like Figure 3 As shown, step e212 for determining the total error associated with the considered measurement location includes sub-steps e2121 and e2122:

[0168] e2121) For each measurement location PM in list U, determine the measurement location in list S with the smallest deviation from it. The deviation between these two measurement locations is defined as the minimum deviation.

[0169] e2122) Calculate the total error based on each minimum deviation determined in step e2121);

[0170] In step e2121, for each measurement location PM in list U, the calculator determines the measurement location in list S with the smallest deviation. The calculation of the deviation will be detailed below. This minimum deviation value is saved.

[0171] For example, during iteration j in step e21, after the considered measurement position PM is moved from list S to list U in step e212, list S contains Nj measurement positions PM and list U contains j measurement positions PM. The calculator can then calculate the (Nj)×j deviation between the measurement positions PM in list S and the measurement positions PM in list U.

[0172] It is worth noting that, in order to reduce the number of calculations and thus the computation time, only certain deviations can be calculated to determine the minimum deviation of the measurement position PM in list U. For example, here, the deviations calculated first are those between the spatially closest measurement positions PM in list U and PM in list S; that is, those coordinate differences (hereinafter referred to as distances) are the lowest. Then, the deviations are calculated in ascending order of coordinate differences. Based on the already calculated deviations, if the contribution of the coordinate difference is greater than the already calculated minimum deviation, the calculation may be stopped. Therefore, the minimum deviation must be a portion of the already calculated deviations.

[0173] Here, the total error is equal to the sum of each minimum deviation determined in step e2121. For example, during iteration j of step e21, after calculating the (Nj)×j deviation and determining the j minimum deviations, the calculator calculates the total error associated with the measurement location PM by summing the j minimum deviations of the measurement location PM.

[0174] Once the total error associated with the considered measurement location PM is determined, in step e213, the calculator moves the considered measurement location PM from list U to list S.

[0175] Through iterative steps e211, e212, and e213, the calculator thus associates the total error with each measurement location PM in list S. As previously described, this makes it possible to assign a classification value to the relevant measurement location PM with the smallest total error. For example, during iteration j of step e21, this makes it possible to assign a classification value to the measurement location PM among the j measurement locations in list S that is associated with the lowest total error.

[0176] In the second embodiment (not shown), in step e21, the calculator determines the change in total error associated with each measurement position PM. The change in total error here is the difference between two total errors calculated for two measurement positions PM, for example, according to steps e2121 and e2122. Alternatively, the change in total error could be the difference between a first total error (referred to as the initial total error) calculated before the considered measurement position PM is moved from list S to list U and a second total error calculated after the considered measurement position PM is moved from list S to list U.

[0177] During the given iterations in step e21, the measurement location in list S considered to contribute the least is associated with the minimum change in the total error. This means that in step e22, the selected measurement location is associated with the minimum change in the total error.

[0178] Calculating the total error change, rather than the total error, allows for the study of the total error change from one considered measurement location PM to another considered measurement location PM, rather than calculating the total error for each considered measurement location.

[0179] In fact, for example, it is not necessary to fully calculate the total error associated with the considered measurement location PM. Therefore, the local error is calculated only based on the deviation within the neighborhood of the considered measurement location PM. For example, this neighborhood is defined for each considered measurement location PM based on the minimum deviation. Measurement locations PM that are spatially too far apart due to coordinate differences to correspond to the minimum deviation are excluded from this neighborhood. This neighborhood can also be defined by a predetermined threshold for the coordinate difference.

[0180] Local errors make it possible to estimate the variation in total error, since the measurement location under consideration only has an effect in its neighborhood.

[0181] Furthermore, in each new iteration of step e21, it is possible to estimate a new initial total error without having to calculate the deviation. This can be achieved by subtracting the change in the total error related to the measurement location chosen in the previous iteration from the initial error of the previous iteration. This makes it possible to avoid recalculating the initial total error, thereby reducing the number of calculations.

[0182] Figure 6 This illustrates an example of classifying measurement locations PM based on a simulated initial image. The simulated initial image has a dark circle on a light background, meaning there are two regions of different intensities. For ease of understanding, the grid containing N measurement locations PM is intentionally designed with a relatively high pitch.

[0183] exist Figure 6In the diagram, the six measurement locations PM that contribute the most are displayed in ascending order of their classification values. Two measurement locations PM(1) and PM(2) can respectively estimate the bright and dark areas represented by the shading. Measurement locations PM(3), PM(4), PM(5), and PM(6) are the next most contributing because these measurement locations PM(3), PM(4), PM(5), and PM(6) improve the resolution of the dark area edges.

[0184] Once all measurement locations are sorted, the acquired spectroscopic image IS with a defined number of spectroscopic measurements less than the maximum number N can be estimated. This will be further described after the steps for constructing the spectroscopic image IS are described.

[0185] Step e3

[0186] Once all measurement locations PM are sorted, for example from the largest contributor to the smallest contributor, the calculator can define a set of measurement locations PM from the largest contributor measurement locations.

[0187] Therefore, in step e3, based on the classification values ​​assigned in step e2, the calculator determines a set of P (1≤P≤N) measurement locations PM from the set of N measurement locations PM.

[0188] To determine a set of P measurement locations PM, the calculator selects the P measurement locations that contribute the most, based on the classification values ​​assigned to the measurement locations PM in step e2.

[0189] In other words, in step e3, the P measurement locations PM of the group are determined according to the ascending and descending order of the classification values, respectively.

[0190] For example, in step e2, the calculator increments the classification value by +1 for each measurement location PM, from the smallest to the largest contribution. The P measurement locations in this group are those assigned the P highest classification values. In this case, the P measurement locations PM in the group are determined in descending order of classification values ​​from highest to lowest.

[0191] Continuing as an example, in step e2, if the calculator increments the classification value of the measurement positions PM from the smallest to the largest contribution by -1, then the P measurement positions in this group are those that are assigned the P lowest classification values. In this case, the P measurement positions PM in the group are determined according to the classification values ​​in ascending order from lowest to highest.

[0192] For example, the number P of measurement locations PM can be determined to achieve a given acceleration factor. For instance, to achieve an acceleration factor of 10, the number P of measurement locations PM in this group is equal to N / 10. This means that in step e4, spectroscopic measurements are obtained only at 10% of the measurement locations PM.

[0193] For example, the number P of measurement locations PM can be predetermined by the user.

[0194] For example, the number of measurement positions PM can be determined to achieve a given acquisition duration or the distance the excitation beam 2 moves relative to sample 1, as given in step e4. In these cases, the number P of measurement positions PM in the group depends on the position of the measurement positions PM relative to sample 1.

[0195] The number P of measurement locations PM can also be determined by defining a threshold for the total error. After the classification in step e2), the total error associated with each measurement location PM can be stored in memory. A number P can then be determined such that the sum of the total errors associated with measurement locations not selected for that group is less than the threshold.

[0196] When determining the number P of measurement locations PM, additional criteria can be used. The number P of measurement locations PM can be determined such that the maximum distance between two measurement locations PM is less than a threshold. For example, this threshold could be equal to 15 times the pitch DY or DY.

[0197] Step e4

[0198] Once the group containing the P measurement locations PM that contribute the most can be identified, spectral measurements can be obtained at the P measurement locations PM in that group.

[0199] Therefore, in step e4, for each measurement position of the group determined in step e3, the excitation beam 2 is continuously positioned at the measurement position PM of the region of sample 1 to obtain spectral measurements, and a spectral image IS is constructed based on the obtained spectral measurements.

[0200] Here, the spectral image IS is updated after each new spectral measurement is acquired. Therefore, the spectral image IS is constructed progressively and dynamically in step e4, that is, simultaneously with the acquisition of the spectral measurements. This makes it possible to continuously improve the quality of the spectral image IS as each new spectral measurement is acquired.

[0201] For example, the order in which spectral measurements are acquired can be based on the classification values ​​assigned to the P measurement locations PM in the group, so that spectral measurements are acquired from the measurement location PM that contributes the most to the location that contributes the least.

[0202] According to a specific and advantageous variation, when the group includes at least three measurement positions PM, a trajectory for continuous positioning of the excitation beam can be predetermined. This means that at the P measurement positions PM in the group, the acquisition sequence of spectroscopic measurements can be determined based on the predetermined trajectory of the excitation beam 2 relative to the sample 1.

[0203] This trajectory can be specifically determined to optimize the displacement of the excitation beam 2 relative to the sample 1. For example, this optimization could include minimizing the distance the excitation beam 2 moves relative to the sample 1 during the continuous positioning of the excitation beam 2 at all measurement positions PM in the group. For example, this optimization could also include minimizing the number of directional changes of the sample holder stage. This optimization could also include minimizing the time required for the excitation beam 2 to be continuously positioned at all P measurement positions PM in the group.

[0204] Step e4 involves constructing the spectroscopic image IS. The maximum spatial resolution of the spectroscopic image IS depends in particular on the maximum set N of measurement locations PM. Here, the spatial resolution depends, for example, on the selected displacement pitches DX and DY. The spatial resolution of the spectroscopic image IS also depends on the acquired spectroscopic measurements.

[0205] Generally, for methods of acquiring and forming spectral measurement images (IS), the spatial resolution is not homogeneous across the entire spectral measurement image IS. In fact, since regions in sample 1 with strong spatial variations in intensity correspond to the measurement locations PM that contribute the most, more spectral measurements are acquired in these regions, and therefore the spectral measurement image IS has higher spatial resolution in these regions.

[0206] Therefore, the spatial resolution of the spectral image increases locally with each spectral measurement acquired at the measurement location PM.

[0207] Here, the spectral image IS is composed of spectral measurement pixels. Specifically, the spectral image IS consists of the same number of spectral measurement pixels as the measurement locations in the largest set N.

[0208] Here, because the measurement location PM is arranged in a grid, that is, in rows and columns, such as... Figure 5 or Figure 6 As shown, a spectroscopic image IS is composed of spectroscopic pixels arranged in rows and columns. Each spectroscopic pixel contains a spectroscopic value representing a spectroscopic measurement. For example, a spectroscopic value may represent a spectrum, a portion of a spectrum, the intensity of a narrow spectral band, the intensity of a single wavelength, or information from a multivariate analysis or a specific analysis.

[0209] like Figure 3 As shown, in a preferred embodiment, the construction of the spectral determination image IS includes sub-steps e41 and e42:

[0210] e41) For each measurement location PM for which spectral measurements have been obtained, determine the spectral measurement value based on the spectral measurement;

[0211] e42) For each remaining measurement location, a spectral measurement value is determined based on at least one selected spectral measurement, and the selected spectral measurement is selected according to the coordinate difference between the remaining measurement location and the range location of the at least one selected spectral measurement.

[0212] The remaining measurement positions PM are those for which no spectral measurements were taken. Therefore, the remaining measurement positions here refer to the measurement positions PM that were not considered in step e41.

[0213] In steps e41 and e42, a spectral measurement image IS is constructed by determining a spectral measurement value for each measurement location PM (i.e., for each spectral measurement pixel).

[0214] In step e41, the calculator determines the spectral measurement value for each measurement location PM (referred to as the acquired measurement location) based on the spectral measurement. The spectral measurement value for the acquired measurement location may, for example, be equal to the spectral measurement or equal to a coefficient, that is, equal to the spectral measurement at a certain wavelength or in a certain narrow band.

[0215] If k (1≤k≤N) spectral measurements have been obtained, then determine k spectral values ​​for the k obtained measurement locations based on their respective spectral measurements.

[0216] In step e42, the calculator determines the spectral measurements for each remaining measurement location (that is, measurement location PM for which no measurement has yet been obtained). Subsequently, the term "considered" remaining measurement location refers to the location for which a spectral measurement is determined at a given time.

[0217] To determine the spectral measurements for the remaining measurement locations under consideration, the calculator considers at least one spectral measurement acquired at an already acquired measurement location. One or more acquired measurement locations are selected based on their coordinate differences from the remaining measurement locations under consideration. The coordinate difference between two measurement locations PM is referred to as the distance.

[0218] Here, in step e42, the selected spectral measurements are those acquired at locations less than a minimum distance threshold from the remaining considered measurement locations. This means selecting all acquired measurement locations less than the minimum distance threshold from the remaining considered measurement locations. The spectral measurement value can then be determined based on the average of all selected spectral measurements.

[0219] Furthermore, regardless of the selected spectroscopic measurement, the spectroscopic values ​​here are determined by weighting the deviation between the measurement location at the selected spectroscopic measurement and the remaining measurement locations under consideration for each selected spectroscopic measurement.

[0220] Here, the calculation of the deviation between the two measurement positions PM is the same as described above, that is, a weighted sum of the squares of the intensity difference between the two measurement positions and the squares of the coordinate difference between the two measurement positions. Here, the weighting coefficient R is the same in steps e2 and e4. As a variation, the weighting coefficient R can be specified to be different in steps e2 and e4.

[0221] Here, the weight is inversely proportional to the deviation. The farther the acquired measurement location is from the remaining measurement locations under consideration, that is, the greater the deviation, the lower the weight of its spectral measurement.

[0222] For example, if a spectral measurement is determined based on two spectral measurements, and the deviation between the first measurement location and the remaining considered measurement locations is twice that of the second measurement location, then when determining the spectral measurement, the weight of the spectral measurement obtained at the second measurement location can be twice that of the spectral measurement obtained at the first measurement location.

[0223] In other words, the spectral measurements of the remaining measurement locations are determined based on the spectral measurements of one or more nearby measurement locations, and the proximity between the measurement locations (PMs) is quantified by the deviation between these measurement locations (PMs).

[0224] In some cases, it is planned not to assign spectral measurements to some remaining measurement locations. This would result in incomplete spectral images. Not assigning spectral measurements to remaining measurement locations can be based, for example, on a minimum distance threshold. If no acquired measurement location is within the minimum distance threshold of the remaining measurement locations, then no spectral measurements will be assigned to the remaining measurement locations. This means that if no acquired measurement location is spatially close enough to the remaining measurement locations, no spectral measurements will be assigned to them.

[0225] Figure 7A , Figure 7B , Figure 7C , Figure 7B , Figure 7E and Figure 7F It shows Figure 4 The image shows the spectral measurements of sample 1 at several acceleration coefficients. In these images, the spectral measurements at the remaining measurement locations are equal to the spectral measurements at the closest known measurement location.

[0226] exist Figure 7A In the middle, the acceleration coefficient is 100, in Figure 7B In the middle, the acceleration coefficient is 50, in Figure 7C In the middle, the acceleration coefficient is 20, in Figure 7D In the middle, the acceleration coefficient is 10, in Figure 7E In the middle, the acceleration coefficient is 5, in Figure 7F In this case, the acceleration factor is 1. Note that the quality of the spectral measurement image will not be improved in the conventional way.

[0227] In fact, Figure 7A In the measurement, the spectral image IS constructed at the PM position, where the contribution is greatest (1%), is already able to distinguish the contrast of sample 1. If like... Figure 7B and 7C In this way, constructing spectral images (IS) at 2% or 5% of the measurement locations (PM) will significantly increase their quality. Then, as follows... Figure 7D At 10%, such as Figure 7E At 20%, or as Figure 7F Acquiring spectral measurement data at all measurement locations does not significantly improve the quality of the spectral image IS, because the spectral measurements are already acquired at the measurement location PM, which contributes the most to the image. Even with a high acceleration factor, the word JOBIN can still be distinguished here.

[0228] As described above, once all measurement locations PM are sorted, it becomes possible to estimate the spectroscopic image IS that can be obtained with a given number of spectroscopic measurements. In practice, it is possible to calculate the pixel-to-pixel deviations of the initial image II in the same manner as between the measurement locations PM. Therefore, it is possible to construct an estimated structural image IE based solely on certain pixels (called source pixels PS) corresponding to the measurement locations that contribute the most.

[0229] In the estimated structural image IE, the intensity of pixels that are not source pixels PS can be determined based on the nearest (i.e., the one with the lowest bias) source pixel PS. All pixels whose intensity is estimated based on the source pixel PS constitute a mesh M associated with that source pixel PS. Figure 8 The image shows the location of the source pixel PS, centered on the letter B, with an acceleration factor of 20, and its associated mesh M, superimposed on a portion of the initial image II.

[0230] Figure 9 It shows the corresponding Figure 8 The estimated structural image IE of sample 1. The acceleration factor, here 20, makes it possible to determine the N / 20 most contributing measurement locations, corresponding to the N / 20 source pixels PS of the initial image II.

[0231] Figure 10 yes Figure 7C The magnified image of the spectral determination IS, centered on the letter B, was acquired with the same acceleration factor of 20.

[0232] It is worth noting that, Figure 9 The estimated spatial resolution of the structural image IE and Figure 10They are very close because the two images were constructed using a similar process: one based on N / 20 source pixels (PS) and the other based on N / 20 spectral measurements obtained at the measurement location (PM) that contributes the most to the measurement.

[0233] This estimation of the spectral image IS, due to the estimation of the structural image IE, can, for example, help the user select a given acceleration factor.

[0234] It is worth noting that steps e3 and e4 are iterated to improve the spatial resolution IS of the spectral image. In each new iteration, the group determined in step e3 only includes the measurement locations PM that still require spectral measurements; that is, the remaining measurement locations. In other words, for each iteration of step e3, the measurement locations PM that have already been acquired for spectral measurements are not considered when determining the group. In each new iteration of step e3, due to the classification performed in step e2, the group is determined by selecting the remaining measurement locations that contribute the most.

[0235] For example, in the first iteration of steps e3 and e4, N / 100 spectral measurements are acquired at the N / 100 measurement locations PM that contribute the most. This provides a spectral image IS with an acceleration factor of 100. In the second iteration, N / 100 additional spectral measurements are acquired at the N / 100 subsequent measurement locations that contribute the most. The second iteration provides a spectral image IS with an acceleration factor of 50.

[0236] In this example, the trajectory of the excitation beam 2 relative to the sample 1 can be optimized in each iteration, that is, for each group of N / 100 measurement positions PM.

[0237] In different iterations of steps e3 and e4, different groups may also have different numbers of measurement positions PM. This would happen, for example, if the number of measurement positions PM for each group is determined to obtain a given distance of excitation beam 2 relative to sample 1.

[0238] Several conditions can be used to implement methods for stopping the acquisition and formation of spectral measurement images (IS), and these conditions can sometimes be combined.

[0239] The first stopping condition could be that the number of spectroscopic measurements to be obtained is predetermined.

[0240] The first stopping condition corresponds, for example, to the desired acceleration factor for a given spectral measurement image. For instance, if the desired acceleration factor is 10, then the number of spectral measurements to be acquired is N / 10. Because of this acquisition method, N / 10 measurements are acquired at N / 10 of the most contributing measurement locations, making it possible to acquire very high-quality images at a given acceleration factor.

[0241] For this first stopping condition, it can be specified that the group determined in step e3 contains a number P of measurement positions PM equal to the number of spectral measurements to be acquired, and that the trajectory of the excitation beam is optimized at all P measurement positions PM in this group. This makes it possible to acquire the spectral image IS at a determined acceleration factor in the shortest possible time.

[0242] The second stopping condition could be that the acquisition duration is predetermined. In this way, the method for acquiring and forming spectroscopic images (IS) could potentially acquire spectroscopic images of better quality than existing methods within a given acquisition duration.

[0243] The number of PMs at each measurement location can be predetermined, for example, fixed at N / 100, to correspond to the acquisition steps of the available 1% spectral determination measurements.

[0244] The third stopping condition could be that the method includes stopping measures to end (e.g., during step e4) the acquisition of additional spectroscopic measurements. Acquiring new spectroscopic measurements can then be stopped at any time, including before reaching a given acceleration factor or a given acquisition duration.

[0245] This stopping mechanism could, for example, include user input. User input could be made via a keyboard-type peripheral device connected to a calculator. Therefore, if the user deems the quality of the spectral image IS constructed from a certain number of spectral measurements satisfactory at a given time, they can decide to stop acquiring new spectral measurements. This means that the user can terminate the acquisition and formation process of the spectral image IS whenever they wish.

[0246] According to a specific and advantageous implementation method, such as Figure 3 As shown, step e4 here includes sub-step e43.

[0247] In step e43, the calculator interpolates the spectroscopic image IS by determining multiple interpolation coefficients based on the deviation between the measurement locations PM. Once the spectroscopic image IS has been constructed in steps e41 and e42, interpolation can potentially improve the visual rendering of the spectroscopic image IS.

[0248] This interpolation makes it possible to adjust spectral measurements based on neighboring spectral measurements. Figure 11A and 11B The interpolation steps are explained, making it possible to form an interpolated image IP, which is then applied to... Figure 7A and 7B Spectral image IS.

[0249] For example, interpolation can be polynomial interpolation.

[0250] Here, the interpolation coefficients are determined based on the deviation between the measurement positions PM. The interpolation coefficients are inversely proportional to the deviation.

[0251] For example, the smaller the deviation between two measurement locations PM, the larger the interpolation coefficient between the two spectral measurements associated with those two PM locations. Therefore, for a given measurement location PM, its spectral measurements are primarily adjusted spatially and in intensity based on nearby measurement locations.

[0252] like Figure 11A and 11B As shown, step e43, which interpolates the spectral measurement image IS, allows for faster acquisition of the interpolated image IP, which the user deems to be of satisfactory quality.

[0253] In the context of the third stopping condition, stopping measures can then be triggered more quickly based on the interpolated image IP, which means time is saved.

[0254] Optional steps

[0255] like Figure 3 As shown, the acquisition method can be specified to include step e11.

[0256] In step e11, at least one initial spectral measurement is obtained at at least one predetermined initial measurement location, and a predetermined classification value is assigned to each initial measurement location.

[0257] During the construction of lists S and U, the calculator places the initial measurement locations in list U because these measurement locations have categorical values.

[0258] In practice, several initial spectroscopic measurements are acquired in step e11. For example, these spectroscopic measurements are distributed on an initial grid with a pitch greater than the pitches DX and DY of a grid with a set of N measurement locations PM.

[0259] For example, the pitch in the X and Y directions is 10 times larger than DX and DY, respectively. The initial grid comprises N / 100 of the N measurement positions PM. By acquiring measurements at these N / 100 measurement positions PM, it is possible to acquire the spectral image IS with an acceleration factor of 100, that is, within 1% of the standard acquisition time.

[0260] This makes it possible to quickly generate a low-quality but comprehensive spectral image IS covering the entire sample 1. Then, spectral measurements are acquired at other locations that contribute the most to the measurement, thereby rapidly improving the quality of the spectral image IS.

[0261] The initial classification values ​​for measurement locations are assigned to avoid interfering with the classification of PMs at other measurement locations. For example, all classification values ​​can be set to the same value, such as the initial values ​​described above.

[0262] Then, criteria are simply defined for these initial measurement locations so that the initial measurement locations are not considered when determining the groups in step e3. This means that the initial measurement locations are considered as the acquired measurement locations.

[0263] The initial measurement location is placed in list U. In fact, in step 2, the initial measurement location is taken into account when assigning classification values ​​to other measurement locations PM.

[0264] It is worth noting that a region of interest can be predefined in the initial image II, which includes only a portion of the measurement locations PM, and then in step e3, this group can be determined based on the measurement locations PM that belong to the region of interest.

[0265] For example, the region of interest can be defined by the user on the initial image II. For instance, the user can draw a rectangle or circle on the initial image. By measuring the coordinates of location PM, it can be determined which areas are located within the region of interest.

[0266] The area of ​​the region of interest is smaller than the area of ​​sample 1. Therefore, only a portion of all N measurement locations PM fall within the region of interest. In step e3, the measurement locations PM in this group are selected only from those locations belonging to the region of interest.

[0267] Therefore, in step e4, spectral measurements are only acquired at the measurement location PM, which belongs to the region of interest.

[0268] Therefore, identifying the region of interest makes it possible to acquire spectral measurements only at a specific location in sample 1. Consequently, the spatial resolution and the quality of the spectral image IS are increased only at this specific location. To improve the quality of the spectral image IS within the region of interest, steps e3 and e4 can be performed iteratively by considering only the measurement location PM within the region of interest.

[0269] Defining the region of interest makes it possible to refine the spectral measurement image IS only in a specific location of Sample 1, that is, to improve its spatial resolution. For example, a user can define the region of interest where Sample 1 exhibits strong contrast and high spatial frequencies.

[0270] Regions of interest (ROIs) can also be defined based on the spectral image IS to refine areas in IS that do not show strong contrast in the initial image II. In other words, the measurement location PM in such areas may not necessarily be considered highly contributing. For example, a user can define a ROI where they perceive strong contrast in the spectral image IS. ROIs can also be defined automatically. For instance, a region of sample 1 that shows changes in spectral measurements in the spectral image IS but does not show any particular intensity change in the initial image II can be defined as a ROI.

[0271] New spectroscopic measurements can be acquired outside the region of interest, for example, once all measurement locations of the region of interest have been acquired, or after a defined time period, or once a portion of the spectroscopic image IS corresponding to the region of interest reaches a defined acceleration factor.

[0272] It is worth noting that spatial integrated spectral measurements can be specified for the sector of sample 1.

[0273] Spatial integration of measurements can be achieved by moving the excitation beam 2 relative to sample 1 during the measurement acquisition time. Spatial integration of measurements can also be achieved by widening the lateral dimension of the excitation beam 2 during the measurement acquisition process.

[0274] This is particularly useful when spectroscopic measurements are used to estimate spectroscopic measurements at several adjacent remaining measurement locations. Spatially integrating the measurements makes it possible to obtain an average spectroscopic measurement for this neighborhood. Such an average may be more representative of sample 1 than point spectroscopic measurements applied to adjacent measurements. The estimated image can then be used to estimate whether the spectroscopic measurements could be used to determine the PMR spectroscopic measurements at the remaining measurement locations.

Claims

1. A method of acquiring and forming a spectroscopic image of a sample, comprising the steps of: el) acquiring an initial image of a region of the sample, and defining a maximum set of N spectroscopic measurement locations, the initial image consisting of pixels, 2 < N, each measurement location comprising at least one coordinate and at least one intensity, the coordinate representing a measurement location with respect to the region of the sample, the intensity being determined based on at least one pixel of the initial image located in a neighborhood of the measurement location; e2) assigning a classification value to each of the N measurement locations based on a deviation between the measurement locations, the deviation between two measurement locations being computed based on an intensity difference and a coordinate difference; e3) determining, among the set of N measurement locations, a group of P measurement locations according to the classification values assigned in step e2), 1 < P < N; e4) for each measurement location of the group determined in step e3), successively positioning an excitation light beam at the measurement location on the region of the sample, acquiring a spectroscopic measurement, and constructing the spectroscopic image based on the acquired spectroscopic measurements; wherein, in the step e2), the classification values assigned are all different from each other.

2. The method of claim 1, wherein, The deviation between two measurement locations is computed based on a weighted sum of the square of the intensity difference between the two measurement locations and the square of the coordinate difference between the two measurement locations.

3. The method of claim 1, wherein, In the step e3), the P measurement locations of the group are determined according to an ascending order, respectively a descending order, of classification values.

4. The method of claim 1, wherein, In the step e2), a list S and a list U are initialized, the list S comprising measurement locations for which a classification value is to be assigned, the list U comprising measurement locations for which a classification value has been assigned, and the following steps e21) and e22) are iterated: e21) determining, for each measurement location of the list S, a total error by iterating the following steps e211), e212) and e213): e211) moving the considered measurement location from the list S to the list U; e212) computing, based on the deviations between the measurement locations of the list S and the measurement locations of the list U, a total error related to the considered measurement location; e213) moving the considered measurement location from the list U to the list S; e22) based on each total error determined in the step e21), selecting a measurement location from the list S, assigning a classification value to the selected measurement location, and moving the selected measurement location from the list S to the list U.

5. The method of claim 4, wherein, The classification values are initialized to predetermined values, and, at each iteration of the step e22), a classification value strictly lower than the lowest classification value or strictly higher than the highest classification value is assigned to the measurement location having the lowest total error.

6. The method of claim 4, wherein, In the step e212), determining a total error related to the considered measurement location comprises the steps of: e2121) for each measurement position in the list U, determining the measurement position in the list S having the smallest deviation from the measurement position in the list U, the value of the deviation between the measurement position in the list U and the measurement position in the list S having the smallest deviation from the measurement position in the list U defining the smallest deviation; e2122) based on each smallest deviation determined in step e2121), calculating the total error.

7. The method of claim 1, wherein, The set comprises at least three measurement positions, and wherein a trajectory for the successive positioning of the excitation beam is predetermined.

8. The method of claim 1, wherein, The iteration steps e3) and e4) are performed, and wherein in each new iteration, the set determined in the step e3) comprises only measurement positions for which a spectrometric measurement is to be acquired.

9. The method according to claim 1, comprising a step e11) of acquiring at least one spectrometric measurement at at least one predetermined measurement position, each of said at least one predetermined measurement position being assigned a predetermined classification value.

10. The method of claim 1, wherein, The construction of the spectrometric image comprises the following steps: e41) for each measurement position for which a spectrometric measurement has been acquired, determining a spectrometric value based on the spectrometric measurement; e42) for each remaining measurement position, determining a spectrometric value based on at least one selected spectrometric measurement, the selected spectrometric value being selected as a function of the coordinate difference between the remaining measurement position and the measurement position at which the at least one selected spectrometric measurement has been acquired.

11. The method of claim 10, wherein, In step e42), the selected spectrometric measurements are the spectrometric measurements acquired at the measurement positions having a coordinate difference with the remaining measurement position less than a threshold value.

12. The method of claim 10, wherein, In step e42), the spectrometric values are determined based on weighting each selected spectrometric measurement by the deviation between the measurement position at which the selected spectrometric measurement has been acquired and the remaining measurement position.

13. The method according to claim 1, comprising the step e43) interpolating the spectroscopic measurement image, wherein, A plurality of interpolation coefficients is determined based on the deviations between the measurement positions.

14. The method according to claim 1, comprising in step e4) a stopping measure to end the acquisition of additional spectrometric measurements.

15. The method of claim 1, wherein, The number of spectrometric measurements to be acquired is predetermined.

16. The method of claim 1, wherein, In the initial image, a region of interest is predetermined comprising only a portion of the measurement positions, and in step e3), the set is determined based on the measurement positions belonging to the region of interest.

17. The method of claim 1, wherein, At least one spectrometric measurement is spatially integrated over a sector of the sample.

18. The method of claim 1, wherein, The initial image is acquired according to one of the following imaging methods: - spectral acquisition in one or more predetermined spectral bands; - optical acquisition in reflection or transmission; - hyperspectral acquisition; - acquisition with an atomic force microscope; - acquisition with a scanning electron microscope.

19. A spectrometric device comprising: - a light source designed to produce an excitation beam; - a support adapted to receive a sample; - an image capture device; - a processor programmed to implement steps e1) to e4) of the method according to claim 1.

Images