X-ray laminated coherent diffraction imaging noise reduction method and system

Abstract

The invention relates to the technical field of coherent diffraction imaging, in particular to an X-ray laminated coherent diffraction imaging noise reduction method and system, and the method comprises the steps: collecting a plurality of overlapped diffraction signals of a to-be-detected object through a laminated imaging system, reconstructing an image of the to-be-detected object through a periodic artifact suppression algorithm, and removing periodic low-frequency noise in the image; and inputting the reconstructed image of the to-be-measured object into the depth image prior network, and suppressing high-frequency noise in the image of the to-be-measured object by using the characteristics of high impedance of the depth image prior network to low-frequency information and low impedance of high-frequency information to obtain final image output of the to-be-measured object. According to the method, low-frequency and high-frequency noise in X-ray laminated coherent diffraction imaging can be removed at the same time, and the robustness and precision of image reconstruction are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of coherent diffraction imaging, and in particular to a method and system for reducing noise in X-ray stacked coherent diffraction imaging. Background Art

[0002] As one of the important developments of coherent diffraction imaging (CDI), ptychography combines the concepts of CDI and scanning transmission X-ray microscope (STXM), uses local illumination to overlap and illuminate the sample to obtain spatial redundancy, and reconstructs the complex distribution of the extended object through a series of diffraction patterns to achieve nanometer-level imaging resolution. However, in the X-ray ptychography experiment with nanometer-level resolution, the key to success is the stability of the system and the high signal-to-noise ratio of the diffraction signal, otherwise the reconstruction algorithm may fall into a local optimum or even fail to converge. Therefore, an image reconstruction scheme for noisy data is urgently needed to improve the quality of the object distribution reconstructed by the diffraction pattern. Summary of the invention

[0003] To this end, the present invention provides a method and system for reducing noise in X-ray stack coherent diffraction imaging to solve the problem that the existing reconstructed image effect is not ideal.

[0004] According to the design scheme provided by the present invention, on the one hand, a method for denoising X-ray stacked coherent diffraction imaging is provided, comprising: using a stacked imaging system to collect multiple overlapping diffraction signals of an object to be measured, reconstructing the image of the object to be measured and removing the periodic low-frequency noise therein through a periodic artifact suppression algorithm; inputting the reconstructed image of the object to be measured into a deep image prior network, and using the high impedance characteristics of the deep image prior network for low-frequency information and low impedance characteristics for high-frequency information to suppress the high-frequency noise in the image of the object to be measured, so as to obtain a final output of the image of the object to be measured.

[0005] As the X-ray stack coherent diffraction imaging noise reduction method of the present invention, further, the image of the object to be measured is reconstructed and the periodic low-frequency noise therein is removed by a periodic artifact suppression algorithm, comprising:

[0006] Set the total number of iterations of the periodic artifact suppression algorithm;

[0007] Initialize the distribution of the object to be measured, the probe and the noise distribution estimation function;

[0008] In each round of iteration, the outgoing wave corresponding to each probe position in the current round is obtained according to the object to be measured, the probe and the noise distribution estimation function, and the outgoing wave includes: the object to be measured and the probe distribution; after the outgoing wave is diffracted and propagated, it is superimposed with the noise distribution estimation function on the detector plane to form a guessed value of the diffraction field, and the diffraction signal set measured by the detector is used to constrain the guessed value of the diffraction field and obtain a new diffraction field; the new diffraction field is back-propagated to the object plane and the object to be measured and the detection distribution are updated, and the updated object to be measured and the detection distribution are used to update the outgoing wave in the next round of iteration, until the distribution of the object to be measured and the probe distribution at all positions are updated, and the next round of iteration is entered;

[0009] Until the total iteration rounds of the periodic artifact suppression algorithm are reached, a reconstructed image of the object to be measured is obtained.

[0010] As the X-ray stack coherent diffraction imaging noise reduction method of the present invention, further, the guessed value of the diffraction field formed on the detector plane is expressed as:

[0011]

[0012] Where n is the current iteration round, F represents the diffraction propagation process, r represents the object plane coordinate, u represents the diffraction plane coordinate, represents the outgoing wave at the jth probe position in the current n iteration rounds, represents the diffraction signal measured at the jth probe position, represents the noise distribution estimation function in the first n iterations, It is represented as the guess value of the diffraction field at the jth probe position in the current n iteration rounds.

[0013] As the X-ray stack coherent diffraction imaging noise reduction method of the present invention, further, the process of using the detector measured diffraction signal set to constrain the guessed value of the diffraction field and obtain a new diffraction field is expressed as:

[0014]

[0015] As the X-ray stack coherent diffraction imaging denoising method of the present invention, further, the deep image prior network includes: constructing a random down-sampled input and extracting the input image features through convolution, up-sampling, convolution, normalization and pooling the image features to restore the size of the image features to the reconstructed image size of the object to be measured.

[0016] As the X-ray stack coherent diffraction imaging denoising method of the present invention, further, the reconstructed image of the object to be measured is input into a deep image prior network, comprising:

[0017] Construct a reconstructed image of size The 32-channel random noise encoding is used as the input of the neural network; the reconstructed image of the object to be tested is used as a label to supervise the network parameter tuning.

[0018] As the X-ray stack coherent diffraction imaging noise reduction method of the present invention, further, the optimized neural network parameters are expressed as: The final output of the neural network is Among them, f θ (·) is the deep image prior network, O ↓ (r) is the randomly encoded image input, O P (r) is the image of the object to be tested reconstructed by the periodic artifact suppression algorithm, which serves as the label for supervising network tuning, θ * is the optimal network parameter.

[0019] In another aspect, the present invention further provides an X-ray stack coherent diffraction imaging noise reduction system, comprising: an image reconstruction module and an image optimization module, wherein:

[0020] An image reconstruction module, used to reconstruct multiple overlapping diffraction signals of the object to be measured by using a periodic artifact suppression algorithm and obtain a reconstructed image of the object to be measured;

[0021] The image optimization module is used to input the reconstructed image of the object to be tested into the deep image prior network, and use the deep image prior network to optimize the image quality of the reconstructed image of the object to be tested, suppress the high-frequency noise therein, and obtain the final output of the image of the object to be tested. The deep image prior network suppresses the high-frequency noise in the image based on the high impedance of the neural network to the low-frequency information of the image and the low impedance of the high-frequency information, so as to further improve the quality of the reconstructed object image.

[0022] Beneficial effects of the present invention:

[0023] The present invention combines the physical process of diffraction imaging with deep learning, uses a periodic artifact suppression algorithm to eliminate low-frequency noise in the reconstructed image, and introduces the implicit prior of the neural network into the reconstructed image of the object in combination with the high impedance of the low-frequency information of the image and the low impedance of the high-frequency information of the image, so as to deal with the image quality degradation caused by high-frequency noise. The experimental test was further carried out in the actual measured X-ray stack coherent diffraction imaging data. The results show that the scheme of this case can achieve the simultaneous removal of low-frequency noise and high-frequency noise in X-ray stack coherent diffraction imaging, and improve the robustness and accuracy of image reconstruction, and has a good application effect in the field of image reconstruction. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 The schematic diagram of the noise reduction process of X-ray stack coherent diffraction imaging in the embodiment;

[0025] Figure 2It is a schematic diagram of the working principle of the deep image prior network in the embodiment;

[0026] Figure 3 Schematic diagram of the structure distribution of the reconstructed image before and after noise reduction in the embodiment. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solutions and advantages of the present invention clearer and more understandable, the present invention is further described in detail below in conjunction with the accompanying drawings and technical solutions.

[0028] In view of the problem that the existing noise causes the quality of X-ray stack coherent diffraction imaging to deteriorate, the embodiments of the present invention refer to Figure 1 As shown, a method for reducing noise in X-ray stack coherent diffraction imaging is provided, comprising:

[0029] S101, using a stacked imaging system to collect multiple overlapping diffraction signals of the object to be measured, reconstructing the image of the object to be measured and removing the periodic low-frequency noise therein through a periodic artifact suppression algorithm.

[0030] Specifically, the image of the object to be measured is reconstructed and the periodic low-frequency noise therein is removed by a periodic artifact suppression algorithm, which can be designed to include:

[0031] Set the total number of iterations of the periodic artifact suppression algorithm;

[0032] Initialize the distribution of the object to be measured, the probe and the noise distribution estimation function;

[0033] In each round of iteration, the outgoing wave corresponding to each probe position in the current round is obtained according to the object to be measured, the probe and the noise distribution estimation function, and the outgoing wave includes: the object to be measured and the probe distribution; after the outgoing wave is diffracted and propagated, it is superimposed with the noise distribution estimation function on the detector plane to form a guessed value of the diffraction field, and the diffraction signal set measured by the detector is used to constrain the guessed value of the diffraction field and obtain a new diffraction field; the new diffraction field is back-propagated to the object plane and the object to be measured and the detection distribution are updated, and the updated object to be measured and the detection distribution are used to update the outgoing wave in the next round of iteration, until the distribution of the object to be measured and the probe distribution at all positions are updated, and the next round of iteration is entered;

[0034] Until the total iteration rounds of the periodic artifact suppression algorithm are reached, a reconstructed image of the object to be measured is obtained.

[0035] A noise distribution estimation function is introduced into the extended ptychographical iterative engine (ePIE), and the noise in the diffraction signal is adaptively and gradually suppressed by alternating projections. This algorithm is named the periodic artifact suppression algorithm. In the specific implementation, the coordinates of the object plane and the diffraction plane can be denoted by r and u, respectively. The imaging object is marked as O(r), the probe is marked as P(r), the total number of imaging positions is set to J, and the diffraction pattern set measured by the detector is set to The total number of iterations is N. The steps of the ePIE algorithm can be summarized as follows:

[0036] (1) Randomly guess the initial distribution of objects and probes, denoted as O 0 (r) and P 0 (r);

[0037] (2) For the jth probe position (j≤J) in the nth iteration (n≤N) of the periodic artifact suppression algorithm, the outgoing wave is recorded as:

[0038]

[0039] (3) Using the measured diffraction pattern Constrain the guess value of the diffraction field and obtain a new diffraction field Where F is the diffraction propagation function:

[0040]

[0041] (4) will pass The diffraction field after constraint Back propagates to the object plane to obtain the updated outgoing wave And update the distribution of objects and probes accordingly:

[0042]

[0043] Where Γ represents the support area of ​​the probe, that is, it is 1 within the illumination range of the probe and 0 outside the range. * represents conjugation. α1 and α2 are constants

[0044] (5) Repeat (2)-(4) until the object distribution and probe distribution update of all positions are completed, that is, one round of iteration is completed. After reaching the set total number of iterations, the object reconstruction is achieved and the reconstructed object O is obtained. P (r).

[0045] S102, inputting the reconstructed image of the object to be tested into a deep image prior network, and using the high impedance of the deep image prior network to low-frequency information and the low impedance of high-frequency information to suppress high-frequency noise in the image of the object to be tested, to obtain a final output of the image of the object to be tested.

[0046] Specifically, the deep image prior network includes: constructing a random downsampled input and extracting input image features through convolution, upsampling, convolution, normalization and pooling the image features to restore the size of the image features to the reconstructed image size of the object to be measured.

[0047] The reconstructed image of the object to be tested is input into the deep image prior network, which may include:

[0048] Construct a reconstructed image of size The 32-channel random noise encoding is used as the input of the neural network; the reconstructed image of the object to be tested is used as a label to supervise the network parameter tuning.

[0049] The complete steps of network parameter tuning are expressed as:

[0050]

[0051] O N (r) = f θ* [O ↓ (r)]

[0052] Among them, f θ (·) is the deep image prior network, O ↓ (r) is the randomly encoded image input, O P (r) is the image of the object to be tested reconstructed by the periodic artifact suppression algorithm, which serves as the label for supervising network tuning, θ * is the optimal network parameter.

[0053] like Figure 2 As shown, the deep image prior network f θ (·) structure, H and W are the height and width of O respectively. When generating O(r), O ↓ (r) Input network, O P (r) serves as a label to guide the update of the neural network parameters θ.

[0054] Furthermore, based on the above method, an embodiment of the present invention also provides an X-ray stack coherent diffraction imaging noise reduction system, comprising: an image reconstruction module and an image optimization module, wherein:

[0055] An image reconstruction module, used to reconstruct multiple overlapping diffraction signals of the object to be measured by using a periodic artifact suppression algorithm and obtain a reconstructed image of the object to be measured;

[0056] The image optimization module is used to input the reconstructed image of the object to be tested into the deep image prior network, and use the deep image prior network to optimize the image quality of the reconstructed image of the object to be tested, suppress the high-frequency noise therein, and obtain the final output of the image of the object to be tested. The deep image prior network suppresses the high-frequency noise in the image based on the high impedance of the neural network to the low-frequency information of the image and the low impedance of the high-frequency information, so as to further improve the quality of the reconstructed object image.

[0057] In order to verify the effectiveness of this solution, the following is a further explanation of this solution combined with experimental data:

[0058] The test was carried out on the actual measured X-ray stack coherent diffraction imaging data, and the test results are as follows Figure 3 As shown, the corrected image is the image reconstructed result after removal using the scheme of this case. The results show that the scheme of this case utilizes the noise distribution estimation function in the traditional iterative reconstruction framework and the high impedance characteristics of the neural network to low-frequency noise and high-frequency noise to achieve the simultaneous removal of low-frequency noise and high-frequency noise in the reconstruction results of X-ray stack coherent diffraction imaging, and improve the reconstruction robustness and accuracy.

[0059] Unless otherwise specifically stated, the relative steps, numerical expressions and values ​​of the components and steps set forth in these embodiments do not limit the scope of the present invention.

[0060] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0061] The units and method steps of each example described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. A person of ordinary skill in the art may use different methods to implement the described functions for each specific application, but such implementation is not considered to be beyond the scope of the present invention.

[0062] Those skilled in the art will appreciate that all or part of the steps in the above method can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, such as a read-only memory, a disk or an optical disk. Optionally, all or part of the steps in the above embodiment can also be implemented using one or more integrated circuits, and accordingly, each module / unit in the above embodiment can be implemented in the form of hardware or in the form of software function modules. The present invention is not limited to any specific form of combination of hardware and software.

[0063] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A method for reducing noise in X-ray stack coherent diffraction imaging, characterized in that: Include: A stacked imaging system is used to collect multiple overlapping diffraction signals of the object to be measured, and the image of the object to be measured is reconstructed and the periodic low-frequency noise therein is removed through a periodic artifact suppression algorithm; The reconstructed image of the object to be tested is input into the deep image prior network, which uses the low impedance of low-frequency information and high impedance of high-frequency information of the deep image prior network to suppress the high-frequency noise in the image of the object to be tested, and obtains the final image output of the object to be tested.

2. The X-ray stack coherent diffraction imaging noise reduction method according to claim 1, characterized in that: The image of the object to be measured is reconstructed and the periodic low-frequency noise is removed through the periodic artifact suppression algorithm, including: Set the total number of iterations of the periodic artifact suppression algorithm; Initialize the distribution of the object to be measured, the probe and the noise distribution estimation function; In each round of iteration, the outgoing wave corresponding to each probe position in the current round is obtained according to the object to be measured, the probe and the noise distribution estimation function, and the outgoing wave includes: the object to be measured and the probe distribution; after the outgoing wave is diffracted and propagated, it is superimposed with the noise distribution estimation function on the detector plane to form a guessed value of the diffraction field, and the diffraction signal set measured by the detector is used to constrain the guessed value of the diffraction field and obtain a new diffraction field; the new diffraction field is back-propagated to the object plane and the object to be measured and the detection distribution are updated, and the updated object to be measured and the detection distribution are used to update the outgoing wave in the next round of iteration, until the distribution of the object to be measured and the probe distribution at all positions are updated, and the next round of iteration is entered; Until the total iteration rounds of the periodic artifact suppression algorithm are reached, a reconstructed image of the object to be measured is obtained.

3. The X-ray stack coherent diffraction imaging noise reduction method according to claim 2, characterized in that: The guess value of the diffraction field formed in the detector plane is expressed as: Where n is the current iteration round, F represents the diffraction propagation process, r represents the object plane coordinate, u represents the diffraction plane coordinate, represents the outgoing wave at the jth probe position in the current n iteration rounds, represents the diffraction signal measured at the jth probe position, Represents the noise distribution estimation function in the first n iterations.

4. The X-ray stack coherent diffraction imaging noise reduction method according to claim 3, characterized in that: The process of using the detector's measured diffraction signal set to constrain the guessed value of the diffraction field and obtain a new diffraction field is as follows:

5. The X-ray stack coherent diffraction imaging noise reduction method according to claim 1, characterized in that: The deep image prior network includes: constructing a random down-sampled input and extracting input image features through convolution, and performing up-sampling, convolution, normalization and pooling processing on the image features to restore the size of the image features to the reconstructed image size of the object to be measured.

6. The X-ray stack coherent diffraction imaging noise reduction method according to claim 1, 2 or 5, characterized in that: The reconstructed image of the object to be tested is input into the deep image prior network, which includes: Construct a reconstructed image of size The 32-channel random noise encoding is used as the input of the neural network; the reconstructed image of the object to be tested is used as a label to supervise the network parameter tuning.

7. The X-ray stack coherent diffraction imaging noise reduction method according to claim 6, characterized in that: The optimized neural network parameters are expressed as: The final output of the neural network is Among them, f θ (·) is the deep image prior network, O ↓ (r) is the randomly encoded image input, O P (r) is the image of the object to be tested reconstructed by the periodic artifact suppression algorithm, which serves as the label for supervising network tuning, θ * is the optimal network parameter.

8. An X-ray stack coherent diffraction imaging noise reduction system, characterized in that: Contains: image reconstruction module and image optimization module, among which, An image reconstruction module, used to reconstruct multiple overlapping diffraction signals of the object to be measured by using a periodic artifact suppression algorithm and obtain a reconstructed image of the object to be measured; The image optimization module is used to input the reconstructed image of the object to be tested into the deep image prior network, and use the deep image prior network to optimize the image quality of the reconstructed image of the object to be tested, suppress the high-frequency noise therein, and obtain the final output of the image of the object to be tested. The deep image prior network suppresses the high-frequency noise in the image based on the high impedance of the neural network to the low-frequency information of the image and the low impedance of the high-frequency information, so as to further improve the quality of the reconstructed object image.

9. An electronic device, characterized in that: include: at least one processor, and a memory coupled to the at least one processor; The memory stores a computer program, and the computer program can be executed by the at least one processor to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed, the method according to any one of claims 1 to 7 can be implemented.

Images