A method for recovering and correcting diffraction degraded data and an encoding spectral imaging device

Abstract

The application discloses a kind of diffractive degradation data recovery correction method and coded spectral imaging equipment, to solve the technical problem of information crosstalk and image blur between each channel of spectrum caused by the diffraction of the microstructure of MEMS device.The specific steps include: step 1, obtain the coded spectral imaging data and single-spectral diffraction calibration data of coded imaging instrument;Step 2, construct the degradation model of single-spectral coded data;Step 3, calculate the diffraction degradation factor in frequency domain;Step 4, construct the forward single-spectral image coding model and single-spectral coded image decomposition model of Hadamard multiplication;Step 5, obtain the single-spectral spectral image of diffraction degradation;Step 6, obtain the corrected single-spectral coded image;Step 7, obtain multi-band spectral coded image;Step 8, based on linear coding algorithm and multi-band spectral coded image, the corrected spectral data cube is obtained by inversion calculation, and the recovery correction of diffractive degradation data in spatial dimension and spectral dimension is realized.

Description

Technical Field

[0001] This invention relates to coded spectral imaging methods, specifically to a method for restoring and correcting diffraction degradation data and a coded spectral imaging device. Background Technology

[0002] Encoded aperture spectral imaging technology originated in the field of astronomy in the 1960s. In the imaging of X-rays and gamma rays, ordinary optical lenses are no longer effective for imaging high-energy rays, so it is necessary to encode image information through a mask and indirectly calculate the original image information through mathematical algorithms.

[0003] Since the 1980s, liquid crystal light valves have replaced traditional mechanical templates, significantly improving spatial, spectral, and temporal resolution in coded imaging and coded imaging spectrometers. In the 1990s, companies developed Digital Micro-Mirror Devices (DMDs). Compared to traditional spatial modulation templates, DMDs, based on MEMS technology, offer numerous advantages, including smaller size, faster refresh rates, higher resolution, and more flexible encoding methods. From their inception, they have been applied to various coded imaging scenarios, enabling the control of different optical physical quantities and significantly improving the controllability of coded computational imaging in terms of frame rate, resolution, and information dimension.

[0004] The development of spatial light modulation devices based on MEMS technology has greatly promoted the development of coded imaging technology. However, for optical systems, the diffraction effect brought about by the reduction in the size of the modulation device also has a significant impact on image quality. According to the diffraction law, the closer the wavelength of light is to the size of the aperture through which it passes, the more pronounced the diffraction effect becomes. From the Airy disk calculation formula: Where θ represents the diffraction half-angle. It can be seen that when the aperture diameter D decreases or the wavelength λ of the incident light increases, the field of view of the Airy disk increases. Ultimately, on the imaging plane, the same pixel will simultaneously sense information from multiple modulation channels, greatly affecting the imaging quality of the optical system. This impact is even more severe for multidimensional information encoding imaging. Taking Hadamard-coded spectral imaging technology as an example, DMD not only encodes and modulates spatial information but also includes information from each spectral channel. Crosstalk between channels caused by diffraction has a significant impact on spectral information.

[0005] There are currently two solutions to this problem. The first is to minimize the impact of diffraction on image quality through structural optimization during the optical design process. The second is to directly apply filtering or other optimization processes to the image to improve image quality. However, the first method has limited effectiveness in reducing the impact of diffraction, while the second method, although effective in improving the data quality of the image dimension, may not be effective for the spectral dimension, and may even have a more severe impact on the data quality of the spectral dimension. Summary of the Invention

[0006] The purpose of this invention is to provide a method for restoring and correcting diffraction-degraded data and an encoded spectral imaging device to solve the technical problems of crosstalk between spectral channels and image blurring caused by diffraction generated by the microstructure of MEMS devices.

[0007] To achieve the above objectives, the present invention provides a method for restoring and correcting diffraction degradation data, characterized by the following steps:

[0008] Step 1: Acquire the coded spectral imaging data and single-band diffraction calibration data of the coded imaging instrument;

[0009] Step 2: Construct a degradation model for single-band coded data;

[0010] Step 3: Calculate the diffraction degradation factor in the frequency domain based on the degradation model of the coded template image, single-band diffraction calibration data, and single-band coded data;

[0011] Step 4: Construct the forward single-band image coding model and single-band coded image decomposition model of Hadamard multiplication;

[0012] Step 5: Perform single-band image decomposition based on the Hadamard multiplication single-band coded image decomposition model and coded spectral imaging data to obtain diffraction-degraded single-band spectral images;

[0013] Step 6: Based on the degradation model of single-band coded data and the diffraction degradation factor, perform frequency domain correction on the diffraction-degraded single-band spectral image to obtain the frequency domain signal of the single-band coded image. Then, convert the frequency domain signal into a spatial image signal using the frequency domain conversion method to obtain the corrected single-band coded image.

[0014] Step 7: Obtain multi-band spectral coded images from the forward single-band image coding model based on Hadamard multiplication and the corrected single-band coded image;

[0015] Step 8: Based on the linear coding algorithm and multi-band spectral coding image inversion calculation, the corrected spectral data cube is obtained, realizing the restoration and correction of diffraction degradation data in the spatial and spectral dimensions.

[0016] Furthermore, step 1 specifically includes:

[0017] 1.3 Acquire coded spectral imaging data;

[0018] 1.4. Turn on the light source, which is monochromatic light output by a monochromator or other light homogenizing device;

[0019] 1.3 Adjust the DMD template to the working state and load the calibration template; different calibration templates correspond to different spectral bands, and adjust the light source to the corresponding band;

[0020] 1.4 Acquire single-band diffraction calibration data using a detector.

[0021] Furthermore, in step 2, the degradation model includes a calibration degradation model and an imaging degradation model;

[0022] Step 2 includes:

[0023] 2.1 Constructing a scaling and degradation model for the spatial domain of single-spectral-band coded data:

[0024] M'(x,y)=M(x,y)*h(x,y);

[0025] Where M′ represents the actual diffraction image signal of the coding template; M represents the ideal image signal of the coding template; h is the diffraction degradation factor; (x, y) represents the two-dimensional coordinates of the spatial image;

[0026] 2.2. By performing a Fourier transform on the spatial domain scaling degradation model of single-spectral-band coded data, a scaling degradation model in the frequency domain is obtained:

[0027]

[0028] in, This represents the frequency domain signal of the actual diffraction image of the encoded template; H represents the frequency domain signal of the ideal image of the coding template; H is the frequency domain signal of the diffraction degradation factor; (u, v) represents the two-dimensional coordinates of the corresponding frequency domain of the spatial image;

[0029] 2.3 Constructing an imaging degradation model for single-band coded data in the spatial domain:

[0030]

[0031] Where I represents the input image; η represents noise;

[0032] 2.4 Ignoring the influence of noise, the imaging degradation model in the frequency domain is obtained by performing a Fourier transform on the imaging degradation model in the spatial domain of the single-spectral-band coded data:

[0033]

[0034] Where F represents the Fourier transform.

[0035] Furthermore, step 3 specifically includes:

[0036] 3.1. Based on Fourier transform, the coded template image and single-band diffraction calibration data are transformed into corresponding frequency domain signals;

[0037] 3.2. Based on the calibration degradation model and imaging degradation model in the frequency domain, the diffraction degradation factor in the frequency domain is calculated from the frequency domain signal obtained in step 3.1.

[0038] Furthermore, step 4 specifically includes:

[0039] 4.1 Constructing a forward single-band image coding model for Hadamard multiplication:

[0040]

[0041] Where P′ is the actual multi-band diffraction coded image, T is the process matrix constructed according to the set coding template, and M i ′ represents the degraded encoding template, I i The image represents a single-spectral-band image with an original size of m×n, where m and n represent the rows and columns of the detector, respectively, and p is the order of the coding template; 4.2, the forward single-spectral-band image coding model is converted into a single-spectral-band coded image decomposition model through inverse matrix calculation:

[0042]

[0043] Among them, T -1 This represents the inverse of the process matrix.

[0044] Furthermore, step 6 specifically includes:

[0045] 6.1 Perform Fourier transform on the diffraction-degraded single-segment spectral image obtained in step 5 to convert it into the frequency domain signal of the diffraction-degraded single-segment coded spectral image.

[0046] 6.2. Based on the image degradation model in the frequency domain of single-band coded data and the diffraction degradation factor, frequency domain correction is performed on the frequency domain signal of the diffraction-degraded single-band coded spectral image to obtain the frequency domain signal of the corrected single-band coded image:

[0047]

[0048] 6.3. The frequency domain signal of the corrected single-band coded image is converted into a spatial image signal by the frequency domain conversion method to obtain the corrected single-band coded image.

[0049] Furthermore, step 8 specifically includes:

[0050] 8.1 Establishing a spatial coding imaging linear model:

[0051] Y = AX + e;

[0052] Where A represents the coding matrix, which generally contains only "0" and "1" elements, X represents the spectral signal to be solved, Y represents the channel signal of the multi-band spectral coded image, and e represents noise.

[0053] 8.2 For complete encoding, the spectral signal is solved by inversion to obtain the corrected spectral data cube;

[0054] For Hadamard-coded spectral imaging, the corrected spectral data cube is obtained by solving for the spectral dimension using the following formula:

[0055]

[0056] Where N represents the coding order (equivalent spectral band number), A -1 Let A' denote the inverse matrix of A, and let A' denote the transpose matrix of A.

[0057] For sparse coding or compressed sensing coding, the corrected spectral data cube is obtained through convex optimization algorithms or through neural networks and deep learning algorithms. The convex optimization algorithms include greedy algorithms, greedy pursuit, thresholding algorithms, matching pursuit, and subspace pursuit.

[0058] Meanwhile, the present invention also provides a coded spectral imaging device for the above-mentioned method for restoring and correcting diffraction degradation data, so as to obtain coded spectral imaging data and single-band diffraction calibration data of the coded imaging instrument; its special feature is that it includes a coded spectral imager and a monochromatic surface light source.

[0059] The monochromatic surface light source is used to emit outgoing light;

[0060] The coded spectral imager includes a front imaging mirror, a coding template, a beam splitter, an imaging mirror, and a detector arranged sequentially along the optical path of the emitted light, as well as a host computer for control; the coding template and the detector are respectively connected to the host computer for control.

[0061] Furthermore, it also includes a collimating lens disposed between the coding template and the beam splitter and located on the optical path of the outgoing light.

[0062] Furthermore, the monochromatic surface light source includes a blackbody light source and a filter; the blackbody light source emits outgoing light; the filter is disposed in the optical path between the blackbody light source and the front imaging mirror;

[0063] Alternatively, the monochromatic surface light source includes a monochromator and a homogenizing lens; the monochromator emits outgoing light rays; the homogenizing lens is disposed in the optical path between the monochromator and the front imaging mirror;

[0064] Alternatively, the monochromatic surface light source includes a tunable laser and an integrating sphere; the tunable laser emits outgoing light rays; and the integrating sphere is disposed in the optical path between the tunable laser and the front imaging mirror.

[0065] Furthermore, the front imaging mirror is a transmission lens or a reflection lens;

[0066] The beam splitter is a dispersive prism or a grating.

[0067] The beneficial effects of this invention are:

[0068] 1. The method for restoring and correcting diffraction-degraded data provided by this invention can restore and correct diffraction-degraded images, obtaining high-quality imaging data. This method is particularly suitable for coded imaging devices that use multi-dimensional information compression imaging.

[0069] 2. The method for restoring and correcting degraded diffraction data provided by the present invention does not require any changes to the hardware of the coded imaging spectroscopy equipment. The degraded data can be corrected only through camera calibration and algorithm correction in the later stage.

[0070] 3. Compared with other correction methods, the method for restoring and correcting diffraction degradation data provided by this invention combines the principles of coded spectral imaging and encoding / decoding during the correction process, which can achieve synchronous correction and optimization of image information and spectral information.

[0071] 4. The diffraction degradation factor and degradation matrix of the present invention are obtained through laboratory calibration with high accuracy, and the data correction accuracy of the restored data cube is high.

[0072] 5. This invention achieves high-precision correction of spatial and spectral information through data transformation and calculation in three dimensions: spatial domain, frequency domain, and spectral domain.

[0073] 6. The method for restoring and correcting diffraction degradation data provided by this invention is universal and can be applied to the data quality improvement work of various computational coding imaging devices.

[0074] 7. This invention can effectively improve the data quality of coded spectral imaging equipment through algorithms, which is conducive to the low-cost application of coded spectral imaging equipment. Attached Figure Description

[0075] Figure 1 This is a schematic diagram of the structure of an embodiment of the coded spectral imaging device of the present invention;

[0076] Figure 2This is a flowchart of an embodiment of a method for restoring and correcting degraded diffraction data according to the present invention.

[0077] Icon labels:

[0078] 1-Spectral imager, 1-1-Front imaging mirror, 1-2-Encoding template, 1-3-Collimating mirror, 1-4-Spectrometer, 1-5-Imaging mirror, 1-6-Detector, 1-7-Control host computer;

[0079] 2- Monochromatic surface light source. Detailed Implementation

[0080] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0081] This invention proposes a method for restoring and correcting diffraction-degraded data in a MEMS-based coded spectral imaging device. It mainly addresses the problems of crosstalk and image blurring between different spectral channels caused by diffraction generated by the microstructure of MEMS devices. By combining camera diffraction effect calibration and data restoration and correction models and algorithms, the method can jointly achieve the restoration, correction, and inversion of the acquired data.

[0082] This method obtains the diffraction degradation factor in the frequency domain for different spectral channels and spatial positions during modulation by calibrating the diffraction degradation data formed by the MEMS coded template during the imaging process and then calculating the frequency domain diffraction degradation factor through frequency domain transformation. Secondly, a model for restoring and correcting coded spectral imaging data is established. A decomposition method for single-segment coding is constructed based on Hadamard multiplication to solve for the single-segment coded degradation image. Further correction is performed by combining camera-acquired data, calibration factors, and matrices after frequency domain transformation. The corrected data is then reconstructed using Hadamard multiplication. Based on this, the spectral dimensions of the data are decoded and reconstructed, resulting in a corrected and restored spectral data cube.

[0083] Step 1: Build a calibration device platform for the coded spectral imaging equipment to obtain coded spectral imaging data and single-band diffraction calibration data of the coded imaging instrument;

[0084] The diffraction effect of a coded template-based coded spectral imaging device is calibrated, and the calibration device is as follows: Figure 1As shown, it includes an coded spectral imager 1 and a monochromatic surface light source 2. The monochromatic surface light source 2 is used to output the emitted light to the coded spectral imager 1. The coded spectral imager 1 includes a front imaging mirror 1-1, an encoding template 1-2, a collimating mirror 1-3, a beam splitter 1-4, an imaging mirror 1-5, and a detector 1-6 arranged sequentially along the optical path of the emitted light, as well as a control host computer 1-7. The encoding template 1-2 and the detector 1-6 are respectively connected to the control host computer 1-7. The front imaging mirror 1-1 can be a transmission or reflection lens, the encoding template 1-2 can be a digital micromirror array or a liquid crystal spatial light modulator, and the beam splitter 1-4 is a dispersive prism or a grating. The monochromatic surface light source 2 is configured as a blackbody light source with a filter, a monochromator and a homogenizing lens, a tunable laser, and an integrating sphere, etc.

[0085] like Figure 2 As shown, the calibration method is as follows: First, open the digital micromirror array in encoding template 1-2 and load the calibration template. The calibration template only opens the DMD encoding template corresponding to different spectral bands each time. Simultaneously, during the calibration process, sequentially adjust the wavelength of the monochromator or other uniform light source to obtain the diffraction matrix and diffraction degradation factor for different encoding templates corresponding to different spectral bands.

[0086] The detailed process and calculation steps for acquiring coded spectral imaging data and single-band diffraction calibration data of the coded imaging instrument are as follows:

[0087] First, turn on the monochromatic surface light source 2, adjust it to the specified single-spectral channel position, adjust the DMD to be fully on and fully off, and collect and record imaging data through detectors 1-6.

[0088] Next, adjust the DMD to the working state of the coding template 1-2, and collect coded imaging data through detectors 1-6.

[0089] Finally, the monochromator outputs monochromatic light signals of different wavelengths, which are collected by detectors 1-6, and the wavelengths of the monochromator signals are recorded.

[0090] Here, the ideal image signal of the coding template is denoted as M, and the actual diffraction image signal of the coding template is denoted as M′.

[0091] Step 2: Construct a degradation model for single-band coded data, which includes a calibration degradation model and an imaging degradation model.

[0092] Taking a spatially modulated Hadamard-coded spectral imager as an example, its coding can be modeled from both spatial and spectral modulation perspectives. From the perspective of spatial coding, coded spectral imaging technology can be considered as modulating the spatial signals of each spectral band using a coding template, integrating each band, and finally obtaining a coded image. The technology then decodes the image and spectral signals using the coded image.

[0093] Ideally, the single-band coded spectral imaging process can be represented by the following formula:

[0094]

[0095] Among the symbols Let I represent the Hadamard product, I represent the input image, and S represent the ideally encoded single-spectral image.

[0096] However, in actual imaging, image degradation is unavoidable for single-band coded images due to diffraction effects. This degradation occurs because signals with a "1" coded position distribute some energy to imaging positions with a "0" coded position through diffraction. This crosstalk is similar to the convolution smoothing effect of a diffraction degradation factor on a single-band coded image, reducing the system's coding contrast.

[0097] Let the degraded image be S′. Considering detector noise in the actual imaging process, we construct an imaging degradation model for the system, which can be expressed as:

[0098]

[0099] Where h is the diffraction degradation factor and η(x,y) is the degradation noise.

[0100] Step 3: Calculate the diffraction degradation factor in the frequency domain based on the degradation model of the coded template image, single-band diffraction calibration data, and single-band coded data;

[0101] For the coded imaging degradation process, the degradation factor generated by the coded template acts on the final image. When the input image I=1, this degradation factor is superimposed on the ideal image signal M of the coded template, making the actual diffraction image signal of the coded template M′. Therefore, the degradation model of the coded template during calibration can be expressed as:

[0102] M′=M*h (3);

[0103] Considering the detector electronic noise and system photon noise, the degradation process of the coding template can be further expressed as:

[0104] M′=M*h+η(4);

[0105] Theoretically, when no diffraction occurs during imaging, h(x,y) = 1, and M′ = M. However, in actual imaging, the diffraction effect produced by some micromirrors in the digital micromirror array (DMD) degrades the surrounding signal. Therefore, it can be assumed that the spatial coding template matrix has changed, and its relationship to the spatial dimension is as follows:

[0106] M'(x,y)=M(x,y)*h(x,y) (5);

[0107] In the above equation, M is a known ideal matrix containing only "0" and "1". M′ is obtained through actual calibration. After normalization, the signal is no longer an integer, but includes the small component changes caused by diffraction. Performing a Fourier transform on equation (5) yields the frequency domain operations:

[0108]

[0109] exist and Given the information, the value of H(u,v) is calculated in the frequency domain to obtain the degradation factor of the single-band coding template in the frequency domain.

[0110] Step 4: Construct the forward single-band image coding model and single-band coded image decomposition model of Hadamard multiplication;

[0111] The spatial coding principle of a coded spectral imager can be explained as follows: light in different spectral bands is modulated with different spatial signals, and the coded image is obtained by superimposing the modulated signals of each spectral band. This mathematical process can be represented by the Hadamard product.

[0112] Ideally, from a spatial coding perspective, aperture coding involves integrating the coded images of different spectral bands using an integration process matrix T to ultimately obtain the coded image of the entire spectral band.

[0113]

[0114] Where S is the ideal encoded image, and S represents the sequence [S1, S2, S3, ... S2]. p A 2D matrix composed of ], where S i This represents a 1D vector transformed from a 2D image. T represents the integration process matrix consisting of "0"s and "1"s, and T′ represents the transpose of the matrix. P represents a 1D vector transformed from an encoded frame of the image.

[0115] Substituting equation (1) into equation (7), we obtain the following result:

[0116]

[0117] Where p represents the number of spectral segments of the signal (i.e., the order of the coding template), M i and I iLet represent the coding template and the single-band image, respectively, with an original size of m×n, where m and n represent the row and column of the detector, respectively. These are then represented as column vectors of size [m×n, 1]. For most computational coded spectral imaging techniques, the coding templates for each channel are known, where T represents the superposition matrix after coding different spectral bands, and this matrix is ​​sparse.

[0118] For a degraded image, it can be represented as:

[0119]

[0120] In equation (9) above, P′ is the result of actual coded imaging, which can be obtained by reading out the data collected by the detector, and T is the process matrix, which is a known quantity. Then, through matrix inverse transformation, equation (9) above can be transformed into:

[0121]

[0122] Step 5: Perform single-band image decomposition based on the Hadamard multiplication single-band coded image decomposition model and coded spectral imaging data to obtain diffraction-degraded single-band spectral images;

[0123] Given the actual coded imaging P′ and the process matrix T, the actual imaging for each band can be calculated using equation (9). That is, to obtain... The value of , where i = 1, 2, ..., p.

[0124] Step 6: Based on the degradation model of single-band coded data and the diffraction degradation factor, perform frequency domain correction on the diffraction-degraded single-band spectral image to obtain the frequency domain signal of the single-band coded image. Then, convert the frequency domain signal into a spatial image signal using the frequency domain conversion method to obtain the corrected single-band coded image.

[0125] In step 3 above, the frequency domain function H(u,v) of the degradation factor was calculated. In step 5, the following was calculated: Assuming the system noise can be eliminated, it will be reduced to a factor. Substituting this into equation (6), we get the following formula:

[0126]

[0127] Where "F" represents Fourier transform. Given H(u,v), and I representing the input single-spectral-band image to be solved, then:

[0128]

[0129] Furthermore, by using inverse Fourier transform, the obtained frequency domain information is... Transforming to the spatial domain yields the corrected single-band coded image.

[0130] Step 7: Obtain multi-band spectral coded images from the forward single-band image coding model based on Hadamard multiplication and the corrected single-band coded image;

[0131] For an image with an N-order coding template, there are N sets of coded images in the complete coding state. By correcting them one by one through the above steps, N corrected full-spectrum coded images are obtained.

[0132] Step 8: Based on the linear coding algorithm and multi-band spectral coding image inversion calculation, the corrected spectral data cube is obtained, realizing the restoration and correction of diffraction degradation data in the spatial and spectral dimensions.

[0133] For coded spectral imaging systems, the coding process is actually a linear superposition of signals from different bands, but its expression is as follows:

[0134] Y = AX + e;

[0135] Where Y represents the sensor signal output by the detector, A represents the encoding matrix, which typically contains only "0" and "1" elements, X represents the spectral signal to be solved, and e represents noise, which can usually be eliminated or ignored through denoising. However, in practice, different encoding templates A can lead to differences in data inversion algorithms.

[0136] For complete encoding, since matrix A is full rank, the spectral signal X can be solved by inverting it to obtain the corrected spectral data cube.

[0137] For spectral imaging technology based on Hadamard coding, it can achieve coded imaging through a cyclic S-coding matrix. Due to its special coding matrix construction, the method for solving its spectral dimensions is as follows:

[0138]

[0139] Where A′ represents replacing “0” with “-1” in matrix A, then:

[0140]

[0141] The above formula can be used to solve for the restored and corrected spectral data cube.

[0142] For encoding methods such as compression coding and sparse coding, the data cube can be obtained by iterative optimization using algorithms such as convex optimization, neural networks, and deep learning. Convex optimization algorithms include greedy algorithms, greedy tracking, thresholding algorithms, matching tracking, and subspace tracking. The restored and corrected data cube can be calculated using the above methods.

[0143] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions within the technical scope disclosed in the present invention should be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for restoring and correcting degraded diffraction data, characterized in that, Includes the following steps: Step 1: Acquire the coded spectral imaging data and single-band diffraction calibration data of the coded imaging instrument; Step 2: Construct a degradation model for single-spectral-band encoded data; the degradation model includes a calibration degradation model and an imaging degradation model; specifically including: 2.

1. Construct a scaling and degradation model for the spatial domain of single-spectral-band coded data: ； in, The actual diffraction image signal of the coding template is represented by ; M represents the ideal image signal of the coding template; h is the diffraction degradation factor; (x, y) represents the two-dimensional coordinates of the spatial image. 2.

2. By performing a Fourier transform on the spatial domain scaling degradation model of single-spectral-band coded data, a frequency domain scaling degradation model is obtained: ； in, This represents the frequency domain signal of the actual diffraction image of the encoded template; H represents the frequency domain signal of the ideal image of the coding template; H is the frequency domain signal of the diffraction degradation factor; (u, v) represents the two-dimensional coordinates of the corresponding frequency domain of the spatial image; 2.3 Constructing an imaging degradation model for single-band coded data in the spatial domain: ； Where I represents the input image; η represents noise; 2.4 Ignoring the influence of noise, the imaging degradation model in the frequency domain is obtained by performing a Fourier transform on the imaging degradation model in the spatial domain of the single-spectral-band coded data: ； Where F represents the Fourier transform; Step 3: Calculate the diffraction degradation factor in the frequency domain based on the degradation model of the coded template image, single-band diffraction calibration data, and single-band coded data; Step 4: Construct the forward single-band image coding model and single-band coded image decomposition model of Hadamard multiplication; Step 5: Perform single-band image decomposition based on the Hadamard multiplication single-band coded image decomposition model and coded spectral imaging data to obtain diffraction-degraded single-band spectral images; Step 6: Based on the degradation model of single-band coded data and the diffraction degradation factor, perform frequency domain correction on the diffraction-degraded single-band spectral image to obtain the frequency domain signal of the single-band coded image. Then, convert the frequency domain signal into a spatial image signal using the frequency domain conversion method to obtain the corrected single-band coded image. Step 7: Obtain multi-band spectral coded images from the forward single-band image coding model based on Hadamard multiplication and the corrected single-band coded image; Step 8: Based on the linear coding algorithm and multi-band spectral coding image inversion calculation, the corrected spectral data cube is obtained, realizing the restoration and correction of diffraction degradation data in the spatial and spectral dimensions.

2. The method for restoring and correcting diffraction degradation data according to claim 1, characterized in that, Step 1 specifically includes: 1.1 Acquire coded spectral imaging data; 1.

2. Turn on the light source, which is monochromatic light output by a monochromator or other light homogenizing device; 1.3 Adjust the DMD template to the working state and load the calibration template; different calibration templates correspond to different spectral bands, and adjust the light source to the corresponding band; 1.4 Acquire single-band diffraction calibration data using a detector.

3. The method for restoring and correcting diffraction degradation data according to claim 2, characterized in that, Step 3 specifically includes: 3.

1. Based on Fourier transform, the coded template image and single-band diffraction calibration data are transformed into corresponding frequency domain signals; 3.

2. Based on the calibration degradation model and imaging degradation model in the frequency domain, the diffraction degradation factor in the frequency domain is calculated from the frequency domain signal obtained in step 3.

1.

4. The method for restoring and correcting diffraction degradation data according to claim 3, characterized in that, Step 4 specifically includes: 4.1 Constructing a forward single-band image coding model for Hadamard multiplication: i =1,2,…,p; in, This is the actual multi-band diffraction-coded image, where T is the process matrix, constructed according to the set coding template. This represents the degraded encoding template. The image represents a single-spectral-band image with an original size of m×n, where m and n represent the rows and columns of the detector, respectively, and p is the order of the coding template; 4.2, the forward single-spectral-band image coding model is converted into a single-spectral-band coded image decomposition model through inverse matrix calculation: ； Among them, T -1 This represents the inverse of the process matrix.

5. The method for restoring and correcting diffraction degradation data according to claim 4, characterized in that, Step 6 specifically includes: 6.1 Perform Fourier transform on the diffraction-degraded single-segment spectral image obtained in step 5 to convert it into the frequency domain signal of the diffraction-degraded single-segment coded spectral image. 6.

2. Based on the image degradation model in the frequency domain of single-band coded data and the diffraction degradation factor, frequency domain correction is performed on the frequency domain signal of the diffraction-degraded single-band coded spectral image to obtain the frequency domain signal of the corrected single-band coded image: ； 6.

3. The frequency domain signal of the corrected single-band coded image is converted into a spatial image signal by the frequency domain conversion method to obtain the corrected single-band coded image.

6. The method for restoring and correcting diffraction degradation data according to claim 5, characterized in that, Step 8 specifically includes: 8.1 Establishing a spatial coding imaging linear model: ； Where A represents the coding matrix, which contains only "0" and "1" elements, X represents the spectral signal to be solved, Y represents the channel signal of the multi-band spectral coded image, and e represents noise; 8.2 For complete encoding, the spectral signal is solved by inversion to obtain the corrected spectral data cube; For Hadamard-coded spectral imaging, the corrected spectral data cube is obtained by solving for the spectral dimension using the following formula: ； Where N represents the coding order, A -1 Let A' denote the inverse matrix of A, and let A' denote the transpose matrix of A. For sparse coding or compressed sensing coding, the corrected spectral data cube is obtained through convex optimization algorithms or through neural networks and deep learning algorithms. The convex optimization algorithms include greedy algorithms, greedy pursuit, thresholding algorithms, matching pursuit, and subspace pursuit.

7. A coded spectral imaging device, used for the method of restoring and correcting diffraction degradation data as described in any one of claims 1-6, to obtain coded spectral imaging data and single-band diffraction calibration data of the coded imaging instrument; characterized in that: It includes a coded spectral imager (1) and a monochromatic surface light source (2); The monochromatic surface light source (2) is used to emit outgoing light; The coded spectral imager (1) includes a front imaging mirror (1-1), a coding template (1-2), a beam splitter (1-4), an imaging mirror (1-5), and a detector (1-6) arranged sequentially along the optical path of the emitted light, as well as a control host computer (1-7); the coding template (1-2) and the detector (1-6) are respectively connected to the control host computer (1-7).

8. The coded spectral imaging device according to claim 7, characterized in that: It also includes a collimating lens (1-3) disposed between the coding template (1-2) and the beam splitter (1-4) and located in the optical path of the outgoing light.

9. The coded spectral imaging device according to claim 7 or 8, characterized in that: The monochromatic surface light source (2) includes a blackbody light source and a filter; the blackbody light source emits outgoing light rays; the filter is disposed in the optical path between the blackbody light source and the front imaging mirror (1-1); Alternatively, the monochromatic surface light source (2) includes a monochromator and a homogenizing lens; the monochromator emits outgoing light rays; the homogenizing lens is disposed in the optical path between the monochromator and the front imaging mirror (1-1); Alternatively, the monochromatic surface light source (2) includes a tunable laser and an integrating sphere; the tunable laser emits outgoing light rays; the integrating sphere is disposed in the optical path between the tunable laser and the front imaging mirror (1-1); The front imaging mirror (1-1) is a transmission lens or a reflection lens; The beam splitting device (1-4) is a dispersive prism or a grating.

Images