JP7557845B2 - Imaging device and imaging method thereof - Google Patents

Description

The present invention relates to an imaging device and an image processing method, and in particular to a high dynamic range imaging device and imaging method that achieves high accuracy and high resolution beyond the optical limit in a noisy environment.

これに対して、ベクトルｙについて逆問題を解くことで対象物体のベクトルｘの情報が復元されるというものである。これらの手法は多数のパターンの照合回数を必要とするものの、測定と関係ないノイズ成分を平均化することができるため、高いダイナミックレンジを持つ観測が可能となる。 The historical background of imaging methods using single-pixel detectors can be broadly divided into two categories: Computational Ghost Imaging (CGI), which has been developed from the viewpoint of statistical optics, and Single-Pixel Imaging (SPI), which has been developed from a spectral measurement method. Imaging techniques using single-pixel detectors called SPI and CGI are known, such as those disclosed in Non-Patent Document 1. SPI and CGI filter an object with a spatial distribution pattern. The filtered light intensity is acquired by a single-pixel detector. The patterns are changed one after another, and the light intensity in each pattern is recorded. CGI uses a random pattern as the spatial distribution pattern, and is a method of restoring information on a column vector x of a two-dimensional structure of a target object by calculating the correlation between the pattern and the obtained light intensity. In addition, SPI uses basis patterns generated from orthogonal matrices such as Hadamard matrices for spatial distribution patterns, and the obtained time series of light intensity is defined as vector y . If the matrix in which each pattern corresponds to each row is defined as H , the relationship between vector y and vector x and matrix H can be expressed by equation (1) .

In contrast, the information on the vector x of the target object is restored by solving the inverse problem for the vector y . Although these methods require a large number of pattern matchings, they enable observations with a high dynamic range because they can average out noise components that are not related to the measurement.

In addition, Patent Literature 1 describes an imaging device and imaging method using a compressed sensing system that uses multiple single-pixel cameras, which makes it possible to correct the observation range of each single-pixel camera even when a fixed-value filter is used.

JP 2018-121273 A

M. F. Duarte, et al, IEEE Signal Proc. Mag. 25, 83 (2008).

As mentioned above, imaging technologies using single-pixel detectors, known as SPI and CGI, are attracting attention. SPI and CGI are techniques that use a single-pixel detector to obtain the dot product value between an object and a pattern as light intensity, and then reconstruct information about the object through digital reconstruction processing. SPI and CGI have the advantages of simplifying the optical system, enabling imaging in wavelength bands outside the visible range, and being noise-resistant. In addition, the use of a single pixel detector makes it possible to ensure a high dynamic range. However, since matching with the pattern takes time, real-time observation is difficult. Therefore, although the imaging method is simple and can be made compact, it lacks real-time capabilities and cannot be used as a normal imaging device. It is also difficult to achieve high accuracy.

The present invention has been made in consideration of the above-mentioned circumstances, and its purpose is to provide a high-speed camera with a high dynamic range by using optical correlation technology that enables high-speed image matching and a single-pixel detector that detects light intensity with high sensitivity. Furthermore, by utilizing the ability to record complex amplitude distributions using holograms, it is also possible to capture changes in the amplitude, phase, wavelength, and polarization information of light. Therefore, it is possible to visualize changes in complex amplitude distributions that cannot be obtained with normal cameras, and even for very small samples that exceed the diffraction limit and do not change with the amplitude information used in conventional imaging methods, it is possible to capture and reconstruct the changes in the complex amplitude information, thereby achieving imaging that exceeds the performance of the imaging optical system.

The present invention has been made in consideration of the above-mentioned circumstances, and its purpose is to provide an imaging apparatus and imaging method that uses optical correlation technology capable of high-speed image matching and a single-pixel detector that detects light intensity with high sensitivity, thereby enabling high-speed imaging with a high dynamic range, imaging that removes fluctuations and scattering, and microscopic imaging that exceeds the performance of imaging optical systems.

As a means for solving the above problems, the technology described in the claims is used.
Therefore, by using optical correlation technology that enables high-speed image matching and a single-pixel detector that detects light intensity with high sensitivity, it is possible to provide a high-speed camera with a high D range.

For example, AI technology is currently being used to develop self-driving cars, and the cameras that act as their eyes need to have a high dynamic range. The reason is that the camera needs to be able to capture sufficient images even if sunlight enters the camera after passing through a tunnel, and it needs to have a high dynamic range that can also handle the headlights of oncoming vehicles. However, with conventional technology, there are no cameras with a sufficiently high dynamic range, so changes in light intensity that cameras cannot handle have been dealt with by combining cameras with infrared sensors, etc. Therefore, if high dynamic range cameras could be provided at low cost, the economic impact would be great.

In addition, in order to use optical correlation technology, a matching pattern can be recorded on a holographic disk, and new data can be further recorded using AI to improve accuracy. Furthermore, the holographic disk does not need to be large in size; it is possible to miniaturize it by using a disk that is at most about 1 inch in size and a single-pixel detector.

In other words, because a CCD camera requires a large number of pixels, the sensor equivalent to each pixel has limited sensitivity. In contrast, with a single-pixel detector, it is easy to make the size of each detector 1,000 times larger than the size of one CCD pixel. If it can be made 1,000 times larger, the sensitivity can be improved by simply 60 dB, and the difference between a tunnel and sunlight can be sufficiently corrected to form an image, for example. This can be confirmed.

Furthermore, because holograms can record complex amplitude distributions, they can also capture changes in the amplitude, phase, wavelength, and polarization information of light. This makes it possible to visualize changes in the amplitude, phase, wavelength, and polarization of light that cannot be captured with normal cameras. Even for tiny samples that exceed the diffraction limit and show no change in amplitude information, capturing and reconstructing the changes in the complex amplitude distribution can achieve imaging that exceeds the performance of imaging optical systems.

According to the present invention, a high-speed camera with a high D range can be provided by using an optical correlation technique that enables high-speed image matching and a single-pixel detector that detects light intensity with high sensitivity. Furthermore, a high-definition camera can be provided because a clear image can be formed by removing spatial fluctuations and scattering that interfere with imaging, such as fog, mirages, and thermal fluctuations that cannot be captured by normal cameras. In addition, the influence of scattering bodies that are distant from the object to be measured, such as skin or scattering bodies that are elements that hide the intended object, can be removed, so a high-definition microscope can be provided. Furthermore, microscopic imaging that exceeds the performance of the imaging optical system can be realized, so a super-resolution microscope and a device for inspecting fine structural defects can be provided.

FIG. 1 is a diagram showing an example of an overall system configuration including an imaging apparatus according to an embodiment. FIG. 1 is a block diagram of an imaging apparatus according to an embodiment. FIG. 2 is a software configuration diagram of the imaging apparatus according to the embodiment. 5A to 5C are diagrams showing an example of a method for recording a matching pattern on a disc in the embodiment. FIG. 13 illustrates an example of how a correlation process can be performed in an embodiment. FIG. 13 is a diagram illustrating an example of a method for performing a reconstruction process in an embodiment. FIG. 1 is a diagram showing an overview of a learning method for evaluating a network constructed according to the present invention. FIG. 1 illustrates the effectiveness of the reconstruction process according to the present invention.

Embodiments of the present invention will be described below with reference to the drawings. Identical components throughout the drawings will be given the same reference numerals, and duplicate descriptions will be omitted. In the following embodiments, an imaging device according to the present invention will be described as an example.

1 is a diagram showing an example of the overall configuration including the imaging device 1 according to an embodiment of the present invention. The figure shows a case where the imaging device 1 of the present invention is used as an on-board camera. For ease of explanation, the imaging device 1 is used as an on-board camera, but the present invention is not limited to this. In FIG. 1(a), when an automobile 20 is traveling in a tunnel 21, for example, a normal camera used as an on-board camera, for example an on-board camera using a CCD, increases the sensitivity or amplifies the image recorded because the surrounding brightness is low. Therefore, the image of the on-board camera can show the inside of the tunnel as shown in FIG. 1(a). After that, the moment the automobile 20 leaves the tunnel 21, the sensitivity of the on-board camera is very high, so that when sunlight from the sun 22 directly enters the on-board camera, for example, halation occurs as shown in FIG. 1(b), and correct shooting cannot be performed. In contrast, when the imaging device 1 of the present invention is used, a very high dynamic range is secured for the light intensity entering the on-board camera, so continuous shooting can be performed without halation. The reason why shooting without halation is possible is achieved by using the optical correlation technology and single image sensor technology shown below, but this will be described in detail later.

Figure 2 is a diagram showing the overall block diagram of the imaging device 1. The imaging device 1 includes a laser 2 that is irradiated to obtain reflected or transmitted light from the target object A, a half mirror A4 that irradiates the target object A with laser light and transmits or reflects the reflected light from the target object A of the irradiated laser light, a holographic disk 8 on which a matching pattern is recorded in advance (described later), a lens 3 for focusing the reflected light from the target object A on the holographic disk 8, a half mirror B5 that reflects the light diffracted by the holographic disk, a detector 6 which is a single pixel detector that converts the intensity of the diffracted light into an electrical signal level and sends it to an imaging processing unit, an optical system consisting of the detector 6, and It is made up of an imaging processing unit 7 that converts the level of the electrical signal from the detector 6 into a time axis, measures the time axis change of that level, and performs image processing to create an image of the original target object A, a servo processing unit 9 that rotates the holographic disk 8 at high speed and controls the matching pattern recorded on the holographic disk 8 and the light from the target object A that is incident on it to match the matching pattern with high precision to form diffracted light with high precision, a display unit 14 that displays the image of the target object, a controller 11 that controls and issues instructions for the entire system, a ROM 10, a RAM 12, an external memory 13 that stores images and image data, and a system bus 16.

The image of the object A may be sent to the automatic control processing unit 15 as data for automatic driving. The imaging processing unit 7 includes a learning processing unit 7a and a restoration processing unit 7b, which are described below as processing units for restoring the input light intensity change over time to the original object image, and a memory 7c that temporarily stores matching patterns recorded in the holographic memory and sends them to the learning processing unit 7a and the restoration processing unit 7b.

FIG. 3 is a diagram showing the software configuration of the imaging device 1 of this embodiment, which is made up of a controller 11, a ROM 10, a RAM 12, an external memory 13, and a system bus 16.
The controller 11 is a microprocessor unit that controls the entire image pickup apparatus 1 in accordance with a predetermined program. The system bus 16 is a data communication path for transmitting and receiving data between the controller 11 and each unit in the image pickup apparatus 1.

The ROM (Read Only Memory) 10 is a memory that stores basic operating programs such as an operating system and other application programs. For example, a rewritable ROM such as an EEPROM (Electrically Erasable Programmable ROM) or a flash ROM is used, and by updating the programs stored in the ROM 10, it is possible to upgrade the basic operating programs and other application programs or to expand their functions.

RAM (Random Access Memory) 12 serves as a work area when the basic operation program and other application programs are executed. Specifically, for example, the basic operation program 10a stored in the ROM 10 is expanded into the RAM 12, and the controller 11 executes the expanded basic operation program to form a basic operation execution unit 12a.

In the following, for the sake of simplicity, the process of the controller 11 loading the basic operation program 10a stored in the ROM 10 into the RAM 12 and executing it to control each part will be described as being performed by the basic operation execution unit 12a to control each part. Note that similar descriptions will be used for other application programs.
The captured image processing execution unit 12b instructs the captured image processing unit 7 to start and stop capturing images, instructs the learning processing unit 7a and the restoration processing unit 7b, and instructs the memory 7c to convert the images into data, improve images, and so on.

The learning processing execution unit 12c is software that detects the light intensity of the reflected light from the object A and the diffracted light produced by the holographic disk 8 in FIG. 1 using the detector 6, quantifies the change on the time axis when the light is input to the image processing unit 7 as an electrical signal, converts it into two dimensions, extracts features, and performs various learning processes described below. The restoration processing execution unit 12d is software that restores the data that has been subjected to the learning process by the learning processing unit 7a in FIG. 1 into image data. Specific examples of the method will be described later. The servo processing execution unit 12e is software that includes an algorithm for controlling the rotation control and position control of the holographic disk 8 by feedback or feedforward. The display execution unit 12f is software for displaying the restored image data. However, the software does not have to be entirely software, and for example, a part of it may be implemented as hardware to increase speed.

The controller 11 incorporates an algorithm for controlling the entire image pickup apparatus 1 .
2, the operation of the imaging device 1 of this embodiment is controlled by the captured image processing execution unit 12b, the learning processing execution unit 12c, the restoration processing execution unit 12d, the servo processing execution unit 12e, and the display execution unit 12f, which are executed by the controller 11 after the imaging processing program 13b, the learning processing execution unit 13c, the restoration processing execution unit 12d, the servo processing execution unit 12e, and the servo processing execution unit 12e, which are stored mainly in the external memory 13, are expanded in the RAM 12. The captured image processing execution unit 12b, the learning processing execution unit 12c, the restoration processing execution unit 12d, the servo processing execution unit 12e, and the display execution unit 12f may be performed by hardware blocks that realize some or all of the operations by hardware.

The ROM 10 and RAM 12 may be integrated with the controller 11. Also, the ROM 10 may not be an independent configuration as shown in FIG. 2, but may use a portion of the storage area in the external memory 13. Also, the RAM 12 is provided with a temporary storage area for temporarily storing data as necessary when various application programs are executed.

The external memory 13 may temporarily store images and videos captured by the captured image processing unit 7. It also stores various operational settings of the imaging device 1, location information and various information of the imaging device 1, some or all of the data of images and video information captured by the imaging device 1, and other programs.

A portion of the external memory 13 may replace all or part of the functions of the ROM 10. Furthermore, the external memory 13 needs to retain the stored information even when power is not being supplied to the imaging device 1. Therefore, for example, a device such as a flash ROM, SSD (Solid State Drive), or HDD (Hard Disc Drive) is used.

Next, a method for recording the above-mentioned matching pattern on the holographic disk 8 will be described with reference to Figure 4. The optical correlation system is a system that applies coaxial holographic memory and is a device that can quickly match large amounts of data recorded on a holographic disk with input data. In the optical correlation single pixel camera, the data recorded on the holographic disk corresponds to the base pattern, and the input data corresponds to the target object. The optical correlation single pixel camera consists of a correlation process that matches the observed object with the pattern, and a reconstruction process that reconstructs object information based on the light intensity acquired in the correlation process. The correlation process is performed on the optical correlation system, and the reconstruction process is performed on a computer.

Figure 4 is a diagram explaining a method for recording a matching pattern on a holographic disk. This is hereafter referred to as the recording process. In the recording process, the data to be recorded is input as page data together with a reference light into an unrecorded holographic disk 8, and the interference fringes are shift-multiplexed and recorded as a hologram to create a matching pattern that serves as a database. In Figure 4(a), in the recording process for recording on the holographic disk 8, a spatial distribution pattern is displayed on the spatial light modulator 17 together with a ring-shaped reference light, and input as optical information into the holographic disk 8 using a laser, and the interference fringes are shift-multiplexed and recorded as a hologram. In this figure, a beam splitter 18 is used instead of the half mirror B5, but a half mirror B5 may also be used. For shift multiplexing, the disk is rotated little by little and recorded on the holographic disk as shown in Figure 4(b). In this recording process, the sensitivity, etc. differs depending on the material of the holographic disk, so the recording speed can be slow in order to record sufficiently and fix the information. In addition, the size of one recording pattern can be about several hundred microns. For example, even if data of 100 microns square is recorded once on a disk with a diameter of about 2 inches (5.08 cm), the capacity is 25.4 mm x 25.4 mm x π/0.1 mm2 = approximately 200,000 pieces of data. In reality, it is possible to multiplex several tens of times, so millions of pieces of data can be recorded. Therefore, a sufficient number of patterns can be recorded for matching. As for the device that records on the holographic disk in Figure 4, for example, one device is installed on a factory production line, and multiple holographic disks on which matching patterns are recorded and placed on each imaging device 1 can be created.

FIG. 5 is a diagram for explaining a method of imaging an actual object A using the holographic disk 8 created in FIG. 4. In this case, the matching pattern shown in FIG. 4 has already been recorded on the holographic disk 8, and the state in which the holographic disk 8 is inserted into each imaging device 1 is shown. The process in FIG. 5 is referred to as a correlation process here. In this correlation process, the object A is irradiated with a laser 2, and the light information passes through a beam splitter 18 and enters the holographic disk 8. The intensity of the light (diffracted light) diffracted from the holographic disk 8 in the reference light direction is detected by a detector 6. In FIG. 5 , a transmitted light cut mask 19 is disposed to remove noise from the transmitted light, but it may not be disposed if it is not necessary to insert it in order to reduce costs. Also, while a beam splitter 18 is used in FIG. 5 , a half mirror 5 having a similar function may be used. Since the single detected light intensity corresponds to the inner product value between the recorded pattern and the object, further speedup of the SPI can be expected due to the high speed of the optical correlation system. Here, the pattern matching speed V of the optical correlation system is expressed by equation (2) .

Here, R is the rotation speed of the disk, r is the recording radius of the disk, and d is the shift multiplex recording interval. For example, if the rotation speed of the holographic disk 8 is 900 rpm, the recording radius of the holographic disk is 2 cm, and the shift multiplex recording interval is 10 μm, V is expressed by formula (3) .

Therefore, even if deep learning is performed to improve accuracy, as described below, more than enough data can be collated, and the same image can be reproduced at the 60 frames per second required for normal video footage. After this, a reconstruction process occurs to return the image to the original image of target object A, as shown in Figure 6. At this time, as shown in the figure, various noises are included in the values of the change in light intensity over time.

Next, the reconstruction process for reconstructing an image of a target object using optical correlation technology will be described. Before describing the reconstruction process in an optical correlation single pixel camera, the reconstruction process in conventional single pixel imaging will be described. Single pixel imaging is a method for analytically reconstructing object information using basis patterns such as a Hadamard pattern or a discrete cosine pattern as a spatial distribution pattern. When vector x is the object to be observed, matrix H is the observation matrix, and vector y is the light intensity to be obtained, the correlation process is expressed by equation (1) .

When reconstructing object information based on the acquired light intensity, the inverse matrix of the observation matrix is applied to the light intensity to reconstruct the object information as shown in equation (4) .

Here, by adopting an orthogonal matrix consisting of base patterns such as Hadamard patterns as the observation matrix , equation (5) can be obtained , making it possible to analytically reconstruct object information under noise-free conditions. In single pixel imaging, a high-resolution reconstructed image was obtained by increasing the number of Hadamard pattern matches. In addition, there is an advantage in that noise is averaged by pattern matching, making it highly resistant to noise. However, there is a drawback in that many pattern matches are required to obtain a high-resolution reconstructed image of an object, which means that it takes a long time to observe. For this reason, this method is not suitable for real-time differential imaging equipment such as that in the present invention. The conventional example in FIG. 6 corresponds to this case.

In order to compensate for the shortcomings of the above conventional techniques, we propose a high-speed and high-precision method using an optical correlation single pixel camera. A network using deep learning is used in the reconstruction process to improve the accuracy of the optical correlation single pixel camera. This is the reconstruction process of the present invention shown in Figure 6. The reconstruction network for the optical correlation single pixel camera of the present invention is a network using deep learning and is used instead of the reconstruction calculation shown in Equation 4. The network consists of a deconvolution layer and a deep convolution layer, and is trained so that when the light intensity obtained in the correlation process for a known object is input, the same value as the object information is output. By using the present invention, it is possible to obtain a clear reconstruction image even at a noise level where a reconstruction image cannot be obtained using conventional techniques. Furthermore, high-definition object reconstruction can be expected even when error factors due to the component technologies of the optical correlation system occur. An experimental evaluation of this is shown below.

Figure 7 shows an overview of the learning method used in the evaluation of the constructed network. Time-varying spatial noise and time-invariant spatial noise were added, and the accuracy of the reconstructed image using the optical correlation single pixel camera in the conventional method shown in Figure 6 was compared with that of the image reconstructed using the constructed network. Here, the fixed spatial noise that does not vary with time is assumed to be an error factor derived from the optical correlation system. For example, this is fixed noise due to disk defects or noise due to aberrations in the lens system. In this evaluation, 1024 Hadamard patterns of 32 x 32 pixels were used in the correlation process. In addition, a part of MNIST, an image dataset used in machine learning, was used as the learning data. MNIST is a 28 x 28 pixel dataset of numbers from 0 to 9, with a total of 70,000 images. Of these, 1024 images were used as learning data by filling the outer frame with zeros to match the size of the Hadamard pattern. When training the network, the simulation light intensity when noise is added to the training data in the correlation process was input, and the network was trained to output the same value as the training data. In addition, MNIST data that was not used in the training data was used as the validation data. Figure 8 shows a comparison of the reconstruction results for the validation data when noise is included in the correlation process. CGI, used as an example for comparison, is a reconstruction method that uses random patterns. Although the reconstruction accuracy is worse than SPI, it has excellent noise resistance by increasing the number of patterns irradiated. It was confirmed that the proposed reconstruction network can reconstruct an object even in a situation where noise is added at a level that makes reconstruction impossible with the conventional methods CGI and SPI, in the case of time-varying noise. In addition, in the case of noise that does not change with each observation and does not vary with time, the reconstruction accuracy of the object does not improve by increasing the number of patterns as with CGI, and this is a reconstruction error factor that cannot be removed with conventional methods. However, the proposed method is able to reconstruct the object, and the effectiveness of the constructed network was confirmed by being able to remove the error factors of the optical correlation system.

The above example shows the use of 1,024 Hadamard patterns, but as shown in the calculation example of formula (1), approximately 188,000 matches can be made per second, which is fast enough to achieve real-time noise removal and visualization.

In the above embodiment, an optical correlation single pixel camera is used, and an embodiment of a high dynamic camera is described, but the present invention is not limited to this. The noise added to the object to be measured at a level that cannot be reconstructed can be considered to be spatial fluctuations and scattering that interfere with the imaging of the object to be measured, such as fog, mirage, and thermal fluctuation. In this imaging method, it is possible to remove noise objects such as fluctuations that occur between the object to be measured and the photodetector, and form a clear image. Therefore, a camera with high sensitivity and high resolution can be provided. In addition, in a microscope, the effect of scattering objects that are separated from the object to be measured, such as skin or scattering objects that are elements that hide the intended object, can be removed by the above-mentioned technology. Furthermore, since the present invention uses hologram technology, it is possible to record complex amplitude distribution. Therefore, it is also possible to capture changes in the amplitude, phase, wavelength, and polarization information of light. Therefore, it is possible to visualize changes in the complex amplitude information of light that cannot be obtained with a normal camera. In other words, even for very small samples that exceed the diffraction limit and do not change in amplitude information, imaging that exceeds the performance of the imaging optical system can be achieved by capturing and reconstructing the changes in the complex amplitude information. This makes it possible to provide a super-resolution microscope or a device for inspecting minute structural defects. The above can also be achieved with the block configuration and software configuration in Figures 2 and 3.

Although the present invention describes the case of correlation calculation, the present invention is not limited to this. Reference light is incident on a hologram medium, and the reconstructed light is used as a high-speed complex amplitude pattern irradiator (hologram reconstructor), and the reconstructed light is irradiated onto a target object, and an interference signal between the reconstructed light obtained and the light modulated by the object to be measured is used to capture an image. In this case, the recording medium functions as a reconstructor that reconstructs (generates) multiple complex amplitude patterns at high speed by irradiating the reference light.

A: target object,
1: imaging device,
2: Laser,
3: Lens,
4: Half mirror A,
5: Half mirror B,
6: detector,
7: Imaging processing unit,
7a: learning processing unit,
7b: restoration processing unit,
7c: memory,
8: Holographic disc,
9: servo processing unit,
10: ROM,
11: controller,
12: RAM,
13: external memory,
14: display unit,
15: automatic control processing unit,
16: system bus,
17: spatial light modulator,
18: Beam splitter,
19: Transmitted light blocking mask,
20: Automobiles,
21: Tunnel,
22: Sun

Claims (11)

In an imaging device for photographing a target object,
a recording unit having a recording medium on which a plurality of spatial distribution patterns are optically recorded;
a correlation unit that compares the observed light from the target object with the distribution pattern and generates an optical correlation signal according to a degree of correlation between the target object and the distribution pattern;
a correlation degree calculation unit that receives the optical correlation signal and calculates a correlation degree between the target object and the distribution pattern;
a restoration unit that restores information of the target object from the correlation degree;
Equipped with
the recording medium is a holographic disk,
when comparing the observation light with the distribution pattern, the correlation unit rotates the holographic disk at high speed, causes reflected light or transmitted light from the target object to be incident on the holographic disk, and generates diffracted light from the holographic disk as the optical correlation signal;
When restoring the information of the target object, the restoration unit acquires a reconstructed image of the target object by inputting information about the light intensity according to the degree of correlation to a reconstruction network for an optical correlation single pixel camera.
Imaging device. In an imaging device for photographing a target object,
a recording unit having a recording medium on which a complex amplitude pattern is optically recorded;
an interference unit that performs interference between observation light from the target object and the complex amplitude pattern and generates an optical interference signal according to the degree of interference between the target object and the complex amplitude pattern;
an interference degree calculation unit that receives the optical interference signal and calculates an interference degree between the target object and the complex amplitude pattern;
a reconstruction unit that reconstructs information about the target object from the degree of interference;
Equipped with
the recording medium is a holographic medium,
When restoring the information of the target object, the restoration unit inputs information about the light intensity according to the degree of interference into a reconstruction network for an optical correlation single pixel camera to obtain a reconstructed image of the target object.
Imaging device. In an imaging device for photographing a target object,
a recording unit having a recording medium on which a complex amplitude pattern is optically recorded;
a reproducing unit that generates a complex amplitude pattern by irradiating the recording medium with a reference beam;
an interference unit that generates an interference signal between the target object and the complex amplitude pattern generated by a reproduction unit;
an interference degree calculation unit that receives the interference signal and calculates an interference degree between the target object and the complex amplitude pattern;
a reconstruction unit that reconstructs information about the target object from the degree of interference;
Equipped with
the recording medium is a holographic medium,
When restoring the information of the target object, the restoration unit inputs information about the light intensity according to the degree of interference into a reconstruction network for an optical correlation single pixel camera to obtain a reconstructed image of the target object.
Imaging device. The restoration unit is
a detector, which is a single pixel detector that converts the intensity of the diffracted light from the holographic disk into an electrical signal level;
an imaging processing unit that measures and processes a time-axis change in the level of the electrical signal from the detector to create an image of the target object;
having
The imaging processing unit includes a learning processing unit that is trained to output a light intensity value the same as that of the target object when a light intensity value acquired by the matching for the target object is input .
The imaging device according to claim 1 . The learning processing unit includes a deconvolution layer and a deep convolution layer ;
In the learning of the learning processing unit, when a light intensity when noise is added to image data as learning data in the collation is input, learning is performed so that a light intensity value the same as that of the image data is output.
The imaging device according to claim 4 . The distribution pattern is optically recorded on the recording medium , and information on the target object is restored by capturing changes in amplitude, phase, wavelength, and polarization information of the light modulated by the target object.
The imaging device according to claim 1 , 4 , or 5 . The complex amplitude pattern is optically recorded on the recording medium , and changes in amplitude, phase, wavelength, and polarization information of the light modulated by the target object are captured to restore information of the target object.
The imaging device according to claim 2 or 3 . the plurality of spatial distribution patterns are generated by inputting the spatial distribution patterns together with a reference light into the holographic disk, and shift-multiplexing and recording interference fringes as holograms .
The imaging device according to claim 5 . An imaging method for photographing a target object, comprising: optically recording a plurality of spatial distribution patterns on a recording medium; comparing observed light from the target object with the distribution patterns; generating an optical correlation signal according to a degree of correlation between the target object and the distribution patterns; receiving the optical correlation signal; calculating a degree of correlation between the target object and the distribution patterns from an intensity of the optical correlation signal ; and restoring information about the target object from the degree of correlation ;
A holographic disk is used as the recording medium,
When comparing the observation light with the distribution pattern, the holographic disk is rotated at high speed, and reflected or transmitted light from the target object is incident on the holographic disk, and diffracted light from the holographic disk is generated as the optical correlation signal.
When restoring information about the target object, information about the light intensity according to the degree of correlation is input to a reconstruction network for an optical correlation single pixel camera to obtain a reconstructed image of the target object.
Imaging method. An imaging method for photographing an object, comprising: optically recording a complex amplitude pattern on a recording medium; causing observation light from the object to interfere with the complex amplitude pattern; generating an optical interference signal according to a degree of interference between the object and the complex amplitude pattern; receiving the optical interference signal; calculating a degree of interference between the object and the complex amplitude pattern from an intensity of the optical interference signal ; and restoring information about the object from the degree of interference ;
A hologram medium is used as the recording medium,
When restoring information about the target object, information about the light intensity according to the degree of interference is input to a reconstruction network for an optical correlation single pixel camera to obtain a reconstructed image of the target object.
Imaging method. In an imaging method for photographing a target object, a complex amplitude pattern is optically recorded on a recording medium, a complex amplitude pattern is generated by irradiating a reference light onto the recording medium, an interference signal between the target object and the complex amplitude pattern generated by a reproduction unit is generated, the interference signal is received, a degree of interference between the target object and the complex amplitude pattern is calculated from an intensity of the interference signal , and information about the target object is restored from the degree of interference;
A hologram medium is used as the recording medium,
When restoring information about the target object, information about the light intensity according to the degree of interference is input to a reconstruction network for an optical correlation single pixel camera to obtain a reconstructed image of the target object.
Imaging method.

Images