Imaging system, learning model, server, imaging method, and learning model generation method

Abstract

To provide an imaging system capable of changing speckle noise between two images.SOLUTION: An imaging system (1) includes an imaging device (60) that captures an image in which signal light and reference light interfere with each other, and a displacement mirror (32) that displaces an irradiation position at which an object is irradiated with the signal light, wherein the imaging device captures a first image in which the irradiation position of the signal light is a first position, and a second image in which the irradiation position is a second position displaced from the first position by less than a spot size.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to an imaging system capable of capturing an image with reduced noise, a learning model using the image, a server storing the learning model, an imaging method, and a learning model generation method. [Background technology]

[0002] Optical Coherence Tomography (OCT) is widely used in biological structural analysis, art preservation, medical care, and industrial product quality inspection. Images captured using OCT contain various types of noise, making it difficult to obtain a clear image in a single capture. The Noise2Noise method, as described in Non-Patent Document 1, is one method for reducing noise. The Noise2Noise method considers areas where there are differences between paired images to be noise, and generates a learning model that learns the image features of the noise areas. [Prior art documents] [Non-patent literature]

[0003] [Non-Patent Document 1] Qiu, B.et al., J. biophoto, 14(11), e202100151.(2021) https: / / doi.org / 10.1002 / jbio.202100151 Summary of the Invention [Problem to be solved by the invention]

[0004] However, the noise contained in images captured using the OCT method includes speckle noise, which always appears in the same position if the object position and optical system are the same. The training model generated by the Noise2Noise method determines that noise that does not change in the paired images used for training is not noise. Therefore, a training model generated by the Noise2Noise method using images captured by a conventional OCT imaging device cannot remove speckle noise, resulting in a problem of image quality degradation.

[0005] An object of one aspect of the present invention is to realize an imaging system or the like that is capable of changing speckle noise between two images. [Means for solving the problem]

[0006] In order to solve the above problem, an imaging system according to one embodiment of the present invention includes a light source, a beam splitter that splits light emitted from the light source into signal light and reference light, an imaging device that captures an image in which the signal light irradiated onto an object and the reference light that is not irradiated onto the object interfere with each other, and a displacement mirror that displaces an irradiation position at which the signal light is irradiated onto the object by an amount less than the spot size of the signal light on the surface of the object, and the imaging device captures, as the images, a first image in which the irradiation position is at a first position, and a second image in which the irradiation position is at a second position displaced from the first position by an amount less than the spot size.

[0007] Furthermore, an imaging method according to one aspect of the present invention includes a first image capturing step of capturing a first image, which is an image in which signal light irradiated onto an object and reference light not irradiated onto the object interfere with each other, and in which an irradiation position where the signal light is irradiated onto the object is a first position; a displacement step of displacing the irradiation position by an amount less than the spot size of the signal light on the surface of the object; and a second image capturing step of capturing a second image in which the irradiation position is displaced from the first position by an amount less than the spot size. [Effects of the Invention]

[0008] According to the imaging system and the like according to one aspect of the present invention, speckle noise can be changed between two images. [Brief explanation of the drawings]

[0009] [Figure 1] 1 is a diagram illustrating an example of the configuration of an imaging system according to an embodiment of the present invention. [Figure 2] FIG. 10 is a diagram for explaining a change in the position of speckle noise. [Figure 3] FIG. 10 is a diagram for explaining the Noise2Noise method. [Figure 4] FIG. 2 is a block diagram illustrating an example of the configuration of a main part of a server. [Figure 5] 1 is a flowchart illustrating an imaging method in an imaging system according to an embodiment of the present invention. [Figure 6] 10 shows an example of an image in which the signal light and the reference light interfere with each other. DETAILED DESCRIPTION OF THE INVENTION

[0010] Hereinafter, one embodiment of the present invention will be described in detail.

[0011] (Configuration of imaging system 1) 1 is a diagram illustrating the configuration of an imaging system 1 according to an embodiment of the present invention. The imaging system 1 is a system that captures an image of an object 100 using the OCT method. As shown in FIG. 1, the imaging system 1 includes a light source 10, a beam splitter 20, an imaging device 60, and a displacement mirror 32. The imaging system 1 also includes a scanning mirror 31, a light-guiding mirror 33, an objective lens 34, a light-guiding mirror 41, a crown glass 42, a retroreflector 43, a collimator lens 51, and a spectroscope 52.

[0012] The light source 10 emits coherent light for capturing an image of the interference light of the object 100. The light emitted by the light source 10 is parallel light. The light source 10 is, for example, a so-called supercontinuum light source that includes light of multiple wavelengths. However, the light source 10 is not limited to this and may be a monochromatic laser light source or the like.

[0013] The beam splitter 20 is an optical element that transmits a portion of incident light and reflects the remaining portion. The intensity ratio between the light transmitted and reflected by the beam splitter 20 is, for example, 50:50. Here, 50:50 means that the intensity ratio can be considered to be substantially equal to each other, and does not necessarily have to be completely equal. For example, if the intensity ratio between the transmitted light and the reflected light is about 55:45 to 45:55, the light intensities can be considered to be substantially equal to each other. Of course, the intensity ratio between the transmitted light and the reflected light can be any ratio as long as it is possible to capture the interference light.

[0014] The beam splitter 20 splits the light from the light source 10 into signal light and reference light. One of the light transmitted and reflected by the beam splitter 20 becomes the signal light, and the other becomes the reference light. The signal light is light that is irradiated onto the object 100. The reference light is light that is not irradiated onto the object 100.

[0015] 1, the light reflected by the beam splitter 20 is the signal light, and the light transmitted by the beam splitter 20 is the reference light. However, the imaging system 1 may be configured so that the light transmitted by the beam splitter 20 is the signal light, and the light reflected by the beam splitter 20 is the reference light.

[0016] After the signal light is split by the beam splitter 20, a scanning mirror 31, a displacement mirror 32, a light-guiding mirror 33, and an objective lens 34 are arranged on the path from when the signal light is irradiated onto the object 100 to when the light returns to the beam splitter 20. Furthermore, after the reference light is split by the beam splitter 20, a light-guiding mirror 41, a crown glass 42, and a retroreflector 43 are arranged on the path from when the reference light returns to the beam splitter 20.

[0017] The scanning mirror 31 scans the signal light on the object 100 when capturing a first image and a second image, which will be described later. The interval between scans by the scanning mirror 31 corresponds to the size of a pixel in the image captured by the imaging system 1. The spot size of the signal light is equal to or smaller than the scan interval. Here, the spot size of the signal light is equal to the scan interval and corresponds to the size of a pixel in the image.

[0018] The displacement mirror 32 displaces the irradiation position where the signal light is irradiated on the object 100 by an amount less than the spot size of the signal light on the surface of the object 100. The scanning mirror 31 scans the signal light within a certain range. The displacement mirror 32 displaces the entire range where the scanning mirror 31 scans the signal light.

[0019] As described above, the spot size of the signal light corresponds to the size of a pixel in the image. Therefore, it may be expressed that the displacement mirror 32 displaces the irradiation position where the signal light is irradiated on the object 100 so that the position of the object 100 in the image is displaced by less than the size of one pixel in the image.

[0020] The displacement mirror 32 may include multiple MEMS (Micro Electro Mechanical System) mirrors. By configuring the displacement mirror 32 with multiple MEMS mirrors, the displacement mirror 32 can be displaced at a higher frequency than when the displacement mirror 32 is displaced by, for example, a piezoelectric element or the like.

[0021] For example, the displacement mirror 32 has a structure in which 12 MEMS mirrors are arranged vertically and horizontally, resulting in 144 mirrors, with four mirrors missing from the four corners, for a total of 140 MEMS mirrors. However, the number and arrangement of MEMS mirrors included in the displacement mirror 32 are not limited to this.

[0022] In the displacement mirror 32, the displacement amounts of the multiple MEMS mirrors may be the same. In this case, the displacement mirror 32 can achieve an operation similar to that of a single planar mirror displacing at high frequency. However, in the displacement mirror 32, the displacement amounts of the multiple MEMS mirrors do not necessarily have to be the same, and may be different for each MEMS mirror.

[0023] The displacement frequency of the displacement mirror 32 may be 100 Hz or higher. In other words, the time required for one displacement of the displacement mirror 32 may be 0.01 seconds or less. If the object 100 is a living organism, the position of the object 100 in the image may be displaced due to vital activities such as breathing, regardless of the position of the signal light irradiation. By setting the displacement frequency of the displacement mirror 32 to 100 Hz or higher, even if the object 100 is a living organism, it is possible to capture multiple images in which the position of the signal light irradiation is displaced by the displacement mirror 32 before the position of the object 100 in the image is displaced due to vital activities.

[0024] The light-guiding mirror 33 reflects the signal light reflected by the displacement mirror 32 toward the object 100. The light-guiding mirror 33 may be a general mirror. The number and arrangement of the light-guiding mirrors 33 may be determined appropriately depending on the other components included in the imaging system 1 and the position of the object 100.

[0025] The objective lens 34 reduces the signal light to an appropriate spot size on the surface of the object 100. The magnification of the objective lens 34 may be set appropriately taking into consideration the spot size of the light emitted from the light source 10, the size of the object 100, the desired resolution, etc., and is, for example, 10x.

[0026] The signal light irradiated onto the object 100 is reflected by the object 100. The reflected signal light then passes through the objective lens 34, the light guiding mirror 33, the displacement mirror 32, and the scanning mirror 31 and returns to the beam splitter 20.

[0027] The light-guiding mirror 41 reflects the reference light that has passed through the beam splitter 20. The light-guiding mirror 41 may be a general mirror. The crown glass 42 is crown glass formed into a plate shape. The crown glass 42 may be formed of borosilicate crown glass, but is not limited to this. The crown glass 42 is not essential and may be omitted. The retroreflector 43 reflects (retroreflects) the incident reference light in a direction parallel to and opposite to the incident direction. Therefore, after being reflected by the retroreflector 43, the reference light returns to the beam splitter 20 via the crown glass 42 and the light-guiding mirror 41, regardless of the angle of incidence with respect to the retroreflector 43.

[0028] The crown glass 42 is described below in more detail. In the optical path of the signal light, multiple lens elements are implemented inside the objective lens 34. The signal light is dispersed according to the total thickness of the multiple lens elements. If the difference between the dispersion of the reference light and the dispersion of the signal light is large, the OCT image will become blurred. The crown glass 42 causes dispersion in the reference light to the same extent as the dispersion of the signal light caused by the lens elements. This reduces blurring in the OCT image.

[0029] The beam splitter 20 combines the signal light that is irradiated onto the object 100 and returned, and the reference light that is reflected by the retroreflector 43 and returned. In the example shown in Fig. 1, the signal light that is transmitted by the beam splitter 20 and the reference light that is reflected are combined.

[0030] Collimator lens 51 focuses the signal light and reference light that have been combined by beam splitter 20. Spectrometer 52 separates the signal light and reference light into individual wavelengths. Specifically, spectrometer 52 includes an optical fiber into which light is incident, a collimator lens separate from collimator lens 51, and a transmission grating. Collimator lens 51 focuses the signal light and reference light into the optical fiber of spectrometer 52. The signal light and reference light that have entered the optical fiber of spectrometer 52 are collimated again by the collimator lens of spectrometer 52, and then separated into individual wavelengths by the transmission grating.

[0031] The imaging device 60 captures an image in which the signal light and the reference light, which have been separated by wavelength, interfere with each other. The imaging device 60 is, for example, a device that captures an image in which the signal light and the reference light, which have been scanned across the object 100 by the scanning mirror 31, interfere with each other using a single pixel for each of the separated wavelength ranges. Alternatively, the imaging device 60 may be a device that captures an image in which the signal light and the reference light, which have been reflected by the object 100, interfere with each other using a plurality of pixels that are two-dimensionally arranged. When the imaging device 60 is a device that captures an image using a plurality of pixels that are two-dimensionally arranged, the scanning mirror 31 may be omitted. In this case, the displacement mirror 32 displaces the entire imaging range by an amount less than the spot size of the signal light.

[0032] Generally, in an image, a single pixel contains multiple particles such as cells, molecules, or atoms. In images captured using the OCT method, speckle noise occurs due to the overlap of reflected light from these particles. Speckle noise depends on the optical system and the position of the object 100. Therefore, unlike random noise caused by, for example, a photodetector in the imaging device 60 or air turbulence, speckle noise does not disappear even when multiple images of the object 100 at the same position captured using the same optical system are averaged.

[0033] As described above, in the imaging system 1, the displacement mirror 32 displaces the irradiation position where the signal light is irradiated on the object 100 by an amount less than the spot size of the signal light on the surface of the object 100. The imaging device 60 captures a first image in which the irradiation position is at a first position, and a second image in which the irradiation position is at a second position displaced from the first position by an amount less than the spot size.

[0034] FIG. 2 is a diagram for explaining changes in the position of speckle noise. In FIG. 2, spot SP is a spot of signal light on object 100. In FIG. 2, three different examples of the position of spot SP are indicated by reference numerals 201 to 203. Particles PA1 to PA4 are particles present on the surface of object 100. Spot SP displaces from the position indicated by reference numeral 201 to the position indicated by reference numeral 202 or 203 along the displacement direction. The dashed lines in FIG. 2 indicate both ends of spot SP at reference numeral 201 in the displacement direction.

[0035] In the example shown by reference numeral 201, particles PA1 and PA2 are included in spot SP. In the example shown by reference numeral 202, spot SP has been displaced to one side in the displacement direction, so that particle PA3 is included in spot SP in addition to particles PA1 and PA2. In the example shown by reference numeral 203, spot SP has been displaced from the position shown by reference numeral 201 to the other side in the displacement direction, so that particle PA3 is not included in spot SP, but particle PA4 is included in spot SP.

[0036] Speckle noise is affected by particles in each pixel in an image. In the imaging system 1, the size of the spot SP corresponds to the size of the pixel. Therefore, as the combination of particles contained in the spot SP changes, the overlap of reflected light by the particles at the pixel corresponding to the spot SP also changes. As a result, the presence or absence of speckle noise at that pixel also changes. For example, when speckle noise no longer occurs at the pixel corresponding to the spot SP, speckle noise may occur at a pixel in another position.

[0037] In the first and second images captured by the imaging system 1, the position of the object 100 in the first image is the same as the position of the object 100 in the second image. However, the particles such as cells contained in each pixel in the first image and the second image are different. Therefore, the imaging system 1 can change the position of speckle noise between the first image and the second image.

[0038] If the displacement of the spot SP between the first and second images is greater than or equal to the spot size, the position of the object 100 itself in the images will be shifted. In that case, the entire image, including the object 100, will change between the first and second images. Therefore, it will be impossible to generate an appropriate learning model using the first and second images. On the other hand, as shown in FIG. 2, by setting the displacement of the spot SP to less than the spot size, the position of the speckle noise can be made different between the first and second images without displacing the position of the object 100 itself. This makes it possible to generate an appropriate learning model using the first and second images.

[0039] The imaging device 60 captures the first and second images within, for example, 0.004 seconds. In this case, the time required to capture one image is 0.002 seconds. For example, if the sampling rate is 250 kHz and the number of pixels in the OCT image is 512, the displacement frequency of the displacement mirror 32 is 488.28125 Hz. By capturing the first and second images within such a short time, changes in the object 100 that occur over time between capturing the first and second images can be minimized. In other words, even if the object 100 is a living organism, the first and second images can be captured in a shorter time than the time it takes for the object 100 to change due to biological activity.

[0040] When the object 100 is a living body, the imaging system 1 can be used for structural analysis of the living body. Other applications of the imaging system 1 include preservation of artworks, medical care, and quality checks of industrial products. In these cases, the object 100 is not limited to a living body.

[0041] (Noise2Noise method) FIG. 3 is a diagram for explaining the Noise2Noise method. In FIG. 3, reference numeral 301 is a diagram illustrating the Noise2Noise method using images 301a and 301b captured by a conventional imaging system. Reference numeral 302 is a diagram illustrating the Noise2Noise method using a first image 302a and a second image 302b captured by the imaging system 1. The irradiation positions of the signal light are the same in the images 301a and 301b. As described above, the irradiation positions of the signal light are different in the first image 302a and the second image 302b.

[0042] The images 301a and 301b, and the first and second images 302a and 302b each contain normal noise NN and speckle noise SN. For visibility, only some of the normal noise NN is labeled.

[0043] In the example indicated by the reference numeral 301, the position of normal noise NN is different between images 301a and 301b. On the other hand, the position of speckle noise SN is the same between images 301a and 301b. For this reason, the speckle noise SN is not recognized as noise and cannot be removed by a learning model generated by the Noise2Noise method using images 301a and 301b as input.

[0044] In contrast, in the example shown by reference numeral 302, the positions of both the normal noise NN and the speckle noise SN are different between the first image 302a and the second image 302b. Therefore, according to a learning model generated by the Noise2Noise method using the first image 302a and the second image 302b as input, both the normal noise NN and the speckle noise SN can be removed.

[0045] As described above, the position of the speckle noise in the first image and the position of the speckle noise in the second image are different from each other. Therefore, by generating a learning model using the first image and the second image, it is possible to generate a learning model that can remove speckle noise.

[0046] In the imaging system 1, the displacement mirror 32 may randomly change the amount and direction of displacement of the irradiation position where the signal light is irradiated on the object 100. This allows multiple first and second images to be captured. When generating a learning model using the Noise2Noise method, using multiple pairs of the first and second images allows speckle noise to be removed with higher accuracy.

[0047] Speckle noise depends on the resolution and spatial frequency of the image. Therefore, a learning model generated using a first image and a second image can also remove speckle noise from other images that have the same resolution and spatial frequency as the first image and the second image. The objects in the other images may be different from the object 100 in the first image and the second image.

[0048] The spatial frequency of an image can be derived from the image itself, so information not contained in the image, such as information about the circumstances under which the image was captured, is not required to determine whether or not a learning model can be applied to the image, or which learning model should be applied.

[0049] (Server configuration) 4 is a block diagram illustrating the configuration of the main parts of the server 200. The server 200 functions as a noise removal device that removes noise from images captured by the OCT method. As shown in FIG. 4, the server 200 includes a first acquisition unit 210, a learning unit 220, a second acquisition unit 230, an image processing unit 240, and an output unit 250.

[0050] The first acquisition section 210 acquires a pair of a first image and a second image captured using the imaging system 1. At this time, the first acquisition section 210 may acquire a plurality of pairs of a first image and a second image.

[0051] The learning unit 220 generates a learning model 80. The learning model 80 is a learning model generated using the first image and the second image as input. The learning model 80 has an algorithm that determines whether a certain region in an image captured by the OCT method is noise and minimizes the noise. The algorithm for generating the learning model 80 may be the Noise2Noise method described above, but may also be another algorithm.

[0052] The generated trained learning model 80 is stored in the server 200. To be precise, the server 200 is provided with a storage device (not shown) and stores the learning model 80 in the storage device. For simplicity, in FIG. 4, the training model 80 is included in the image processing unit 240. The trained learning model 80 has learned the feature amounts of noise parts, and outputs an image in which noise has been removed from an input image.

[0053] The second acquisition unit 230 acquires an input image to be subjected to noise removal from an external device or the like. The input image may be a single image, not multiple images in which the irradiation positions of the signal light are displaced from each other. The input image may be an image captured by the imaging system 1 or another device, as long as it is an image captured by the OCT method, i.e., an image in which the signal light and the reference light interfere with each other.

[0054] The image processing unit 240 uses the learning model 80 to generate an output image that has been image-processed to remove noise from the input image. The output unit 250 outputs the output image to a storage device included in the server 200, or to an external storage device or information terminal. Therefore, the server 200 can use the learning model 80 to remove noise contained in the input image.

[0055] (Image capture method) FIG. 5 is a flowchart illustrating an imaging method in the imaging system 1. In the imaging system 1, first, the imaging device 60 captures a first image in which the irradiation position where the signal light is irradiated on the object 100 is a first position (S1, first image capturing step). The position on the object 100 where the signal light is actually irradiated is scanned by the scanning mirror 31 within a range included in the first image during the process of capturing the first image. The "irradiation position" refers to the position within the range scanned by the scanning mirror 31 with the signal light. Next, the displacement mirror 32 displaces the irradiation position of the signal light from the first position to a second position (S2, displacement step). Thereafter, the imaging device 60 captures a second image in which the irradiation position is at the second position (S3, second image capturing step).

[0056] (Learning model generation method) The method for generating a learning model according to the present invention is as follows: First, a first image and a second image are captured by the above-described imaging method (learning image capturing step). Then, a learning model is generated by the Noise2Noise method using the first image and the second image as input (learning step).

[0057] As described above, the first image and the second image have different irradiation positions of the signal light on the object 100. Therefore, the speckle noise that does not change between the first image and the second image is less than when the irradiation position of the signal light on the object 100 is the same between the first image and the second image. Furthermore, even if the object is a living body, the first image and the second image are captured in a time shorter than the time it takes for the object 100 to change. Therefore, the learning model generated by the above method can remove noise, including speckle noise, from images captured by the OCT method with high accuracy.

[0058] (Example) FIG. 6 shows an example of an image in which the signal light and the reference light interfere with each other. The image shown in FIG. 6 is an image captured by OCT using the object 100, which is the cochlea of ​​a mouse. In FIG. 6, reference numeral 601 denotes an image in which noise removal was not performed. Reference numeral 602 denotes an image in which noise removal was performed using a learning model generated by the Noise2Noise method using images in which the irradiation position was not shifted between the first image and the second image. Reference numeral 603 denotes an image in which noise removal was performed using a learning model generated by the Noise2Noise method using the first image and the second image captured by the imaging system 1.

[0059] The image shown by reference numeral 601 clearly contains noise. On the other hand, the images shown by reference numerals 602 and 603 both have reduced noise compared to the image shown by reference numeral 601. In particular, the image shown by reference numeral 603 has even less noise than the image shown by reference numeral 602.

[0060] The difference between the image indicated by reference numeral 602 and the image indicated by reference numeral 603 is thought to be due to the fact that speckle noise was not reduced in the image indicated by reference numeral 602, whereas speckle noise was reduced in the image indicated by reference numeral 603. Therefore, it can be said that by inputting the first and second images captured by the imaging system 1, a learning model capable of removing speckle noise contained in images captured by the OCT method with high accuracy was generated.

[0061] Furthermore, the imaging system 1 may be used to remove noise from OCT images, rather than capturing the first and second images used to generate a learning model. For example, if the object 100 is not a living organism but a completely stationary object, multiple OCT images of the object 100 are captured and simply added together to output an OCT image from which speckle noise has been removed as an average value.

[0062] 〔summary〕 The present invention can also be expressed as follows.

[0063] An imaging system according to aspect 1 of the present invention comprises a light source, a beam splitter that divides light emitted from the light source into signal light and reference light, an imaging device that captures an image in which the signal light irradiated onto an object and the reference light that is not irradiated onto the object interfere with each other, and a displacement mirror that displaces the irradiation position at which the signal light is irradiated onto the object by an amount less than the spot size of the signal light on the surface of the object, and the imaging device captures as the images a first image in which the irradiation position is at a first position, and a second image in which the irradiation position is at a second position displaced from the first position by an amount less than the spot size.

[0064] Aspect 2 of the present invention provides an imaging system according to aspect 1, wherein the displacement mirror includes a plurality of MEMS (Micro Electro Mechanical System) mirrors.

[0065] A third aspect of the present invention provides an imaging system in accordance with the second aspect, wherein the displacement amounts of the plurality of MEMS mirrors are the same.

[0066] A fourth aspect of the present invention relates to the imaging system of the second or third aspect, wherein the frequency of displacement of the plurality of MEMS mirrors is 100 Hz or higher.

[0067] According to a fifth aspect of the present invention, in the imaging system of any one of the first to fourth aspects, the imaging system further includes a scanning mirror that scans the signal light on the object when capturing each of the first image and the second image.

[0068] A sixth aspect of the present invention relates to an imaging system in any one of the first to fifth aspects, wherein the position of speckle noise in the first image and the position of speckle noise in the second image are different from each other.

[0069] The learning model according to the seventh aspect of the present invention is generated by inputting the first image and the second image captured using the imaging system according to any one of the first to sixth aspects.

[0070] A server according to an eighth aspect of the present invention stores the learning model of the seventh aspect.

[0071] An imaging method according to aspect 9 of the present invention includes a first image capturing step of capturing a first image, which is an image obtained by mutual interference between signal light irradiated onto an object and reference light not irradiated onto the object, wherein the irradiation position at which the signal light is irradiated onto the object is a first position; a displacement step of displacing the irradiation position by an amount less than the spot size of the signal light on the surface of the object; and a second image capturing step of capturing a second image, wherein the irradiation position is a second position displaced from the first position by an amount less than the spot size.

[0072] A learning model generation method according to aspect 10 of the present invention is a learning model generation method for generating a learning model for removing speckle noise in an image, and includes a learning image capturing step of capturing the first image and the second image using the imaging method of aspect 9, and a learning step of generating a learning model using the first image and the second image as input.

[0073] The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope of the claims. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of the present invention. [Explanation of symbols]

[0074] 1. Imaging system 10 light source 20 Beam Splitter 31 Scanning Mirror 32 Displacement Mirror 60 Imaging device 80 Learning Model 200 servers

Claims

1. A light source and a beam splitter that splits the light emitted from the light source into a signal light and a reference light; an imaging device that captures an image in which the signal light irradiated onto an object and the reference light not irradiated onto the object interfere with each other; a displacement mirror that displaces an irradiation position where the signal light is irradiated on the object by an amount less than a spot size of the signal light on the surface of the object, An imaging system in which the imaging device captures, as the images, a first image in which the irradiation position is at a first position and a second image in which the irradiation position is at a second position displaced from the first position by less than the spot size.

2. The imaging system of claim 1 , wherein the displacement mirror comprises a plurality of MEMS (Micro Electro Mechanical System) mirrors.

3. The imaging system according to claim 2 , wherein the displacement amounts of the plurality of MEMS mirrors are the same.

4. The imaging system according to claim 2 , wherein the frequency of displacement of the plurality of MEMS mirrors is 100 Hz or higher.

5. The imaging system according to claim 1 , further comprising a scanning mirror that scans the signal light on the object when capturing each of the first image and the second image.

6. The imaging system according to claim 1 , wherein a position of the speckle noise in the first image and a position of the speckle noise in the second image are different from each other.

7. A learning model generated using the first image and the second image captured using the imaging system according to claim 1 as input.

8. A server that stores the learning model according to claim 7.

9. a first image capturing step of capturing a first image, the first image being an image in which a signal light irradiated onto an object and a reference light not irradiated onto the object interfere with each other, and the irradiation position at which the signal light is irradiated onto the object is a first position; a displacement step of displacing the irradiation position by an amount less than a spot size of the signal light on the surface of the object; a second image capturing step of capturing a second image in which the irradiation position is at a second position displaced from the first position by less than the spot size.

10. A learning model generation method for generating a learning model for removing speckle noise in an image, comprising: a learning image capturing step of capturing the first image and the second image by the imaging method according to claim 9; A learning step of generating a learning model using the first image and the second image as input.

Images