WO2025252091A1 - 一种相干衍射成像方法及装置 - Google Patents

一种相干衍射成像方法及装置

Info

Publication number
WO2025252091A1
WO2025252091A1 PCT/CN2025/098904 CN2025098904W WO2025252091A1 WO 2025252091 A1 WO2025252091 A1 WO 2025252091A1 CN 2025098904 W CN2025098904 W CN 2025098904W WO 2025252091 A1 WO2025252091 A1 WO 2025252091A1
Authority
WO
WIPO (PCT)
Prior art keywords
distribution
estimate
frames
frame
sub
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
PCT/CN2025/098904
Other languages
English (en)
French (fr)
Inventor
张福才
盛鹏举
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Southern University of Science and Technology
Original Assignee
Southern University of Science and Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Southern University of Science and Technology filed Critical Southern University of Science and Technology
Publication of WO2025252091A1 publication Critical patent/WO2025252091A1/zh
Pending legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N21/47—Scattering, i.e. diffuse reflection
    • G01N21/4788—Diffraction
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N22/00—Investigating or analysing materials by the use of microwaves or radio waves, i.e. electromagnetic waves with a wavelength of one millimetre or more
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N23/00—Investigating or analysing materials by the use of wave or particle radiation, e.g. X-rays or neutrons, not covered by groups G01N3/00 – G01N17/00, G01N21/00 or G01N22/00
    • G01N23/20—Investigating or analysing materials by the use of wave or particle radiation, e.g. X-rays or neutrons, not covered by groups G01N3/00 – G01N17/00, G01N21/00 or G01N22/00 by using diffraction of the radiation by the materials, e.g. for investigating crystal structure; by using scattering of the radiation by the materials, e.g. for investigating non-crystalline materials; by using reflection of the radiation by the materials
    • G01N23/207—Diffractometry using detectors, e.g. using a probe in a central position and one or more displaceable detectors in circumferential positions

Definitions

  • This disclosure relates to the field of optical diffraction imaging, and more particularly to a coherent diffraction imaging method and apparatus.
  • Coherent diffractive imaging is a lensless imaging technique that uses the coherence of electromagnetic waves such as light waves or X-rays to reconstruct the structural information of an object. This technique breaks through the diffraction limit of traditional optical microscopes, enabling the resolution of the structure of matter at the atomic scale, and has significant implications for materials science, biomedicine, physics, and other fields.
  • diffraction information is typically received by the detector in the form of diffraction patterns. Diffraction patterns only record amplitude information and lose phase information; therefore, phase recovery is required through algorithms to obtain the structural information of the sample.
  • This disclosure provides a coherent diffraction imaging method that improves the iterative convergence speed while obtaining better sample imaging results.
  • This disclosure provides a coherent diffraction imaging method and apparatus that can improve the iterative convergence speed of phase retrieval.
  • this disclosure provides a coherent diffraction imaging method, comprising: obtaining N frames of diffraction patterns ( ⁇ I ⁇ N ) of a target object, wherein a coherent wave diffracts to form an outgoing wave when passing through the target object, and at N occurrence times, the outgoing wave forms N frames of first distribution ( ⁇ N) at the target object, and when the outgoing wave propagates along the outgoing direction to a detector, the N frames of first distribution ( ⁇ N) become N frames of second distribution ( ⁇ N ), and the N frames of diffraction patterns ( ⁇ I ⁇ N) are intensity information of the N frames of second distribution ( ⁇ N) obtained by the detector, wherein N is a positive integer; based on the inter-frame constraints of the N frames of first distribution ( ⁇ N ) and the N frames of diffraction patterns ( ⁇ I ⁇ N), the method further comprises: obtaining N frames of diffraction patterns ( ⁇ I ⁇ N ) of a target object, wherein a coherent wave diffracts to form an out
  • the target complex amplitude distribution (O) of the target object is obtained by performing target iteration until the target complex amplitude distribution (O) satisfies the preset convergence condition.
  • the inter-frame constraint includes updating the similarity distribution ( ⁇ s) between the N frame first distributions ( ⁇ N ) or between the N frame first distributions corresponding to the N frame distributions in the transform domain in each target iteration to obtain the first estimate ( ⁇ ′ ⁇ N ) of the N frame first distribution; and outputting the target complex amplitude distribution (O).
  • updating the similarity distribution ( ⁇ s,m ) among the N-frame first distribution ( ⁇ N ) to obtain a first estimate ( ⁇ ′m ⁇ N ) of the N-frame first distribution includes: determining the similarity ( Sm) of the initial estimate ( ⁇ 0m ⁇ N ) of the N-frame first distribution in the current iteration; based on the similarity ( Sm ), determining the similarity distribution ( ⁇ s ,m ) and the difference distribution ( ⁇ ⁇ e ,m ⁇ N) among the initial estimates ( ⁇ 0m ⁇ N ) of the N-frame first distribution in the current iteration; and based on the similarity distribution ( ⁇ s,m ) and the difference distribution ( ⁇ ⁇ e,m ⁇ N ), determining the first estimate ( ⁇ ⁇ ′m ⁇ N ) of the N-frame first distribution in the current iteration.
  • ⁇ m (r) is the average value of the initial estimate ( ⁇ ⁇ 0m ⁇ N ) of the first distribution of the N frames
  • ⁇ m(r) is the standard deviation distribution of the initial estimate ( ⁇ ⁇ 0m ⁇ N ) of the first distribution of the N frames
  • Cm is the transformation function used to transform ⁇ m (r) into the similarity Sm (r)
  • ⁇ , ⁇ m and ⁇ m are used to adjust the shape of the transformation function.
  • the target iteration includes: obtaining a second estimate ( ⁇ ′′ m-1 ⁇ N ) of the N-frame first distribution of the previous iteration; obtaining an initial estimate ( ⁇ 0m ⁇ N ) of the N-frame first distribution of the current iteration based on the second estimate ( ⁇ ′′ m-1 ⁇ N ) of the N-frame first distribution of the previous iteration; applying the inter-frame constraint to the initial estimate ( ⁇ 0m ⁇ N ) of the N-frame first distribution to obtain a first estimate ( ⁇ ′ m ⁇ N ) of the N-frame first distribution of the current iteration; and obtaining a second estimate ( ⁇ ′′ m ⁇ N ) of the N-frame first distribution of the current iteration based on the first estimate ( ⁇ ′ m ⁇ N ) of the N-frame first distribution of the current iteration, the second estimate ( ⁇ ′′ m -1 ⁇ N ) of
  • obtaining the initial estimate ( ⁇ ′′ m-1 ⁇ N ) of the N-frame first distribution based on the second estimate ( ⁇ ′′ m -1 ⁇ N) of the previous iteration includes: when the second estimate ( ⁇ ′′ m-1 ⁇ N ) of the N-frame first distribution is located within the support region, using the second estimate ( ⁇ ′′ m-1 ⁇ N ) of the N-frame first distribution as the initial estimate ( ⁇ 0m ⁇ N ) of the N-frame first distribution; or when the second estimate ( ⁇ ′′ m -1 ⁇ N ) of the N-frame first distribution is located outside the support region, the value of the second estimate ( ⁇ ′′ m-1 ⁇ N ) of the N-frame first distribution in that region is zero.
  • obtaining the second estimate ( ⁇ ′′ m ⁇ N ) of the N-frame first distribution in the current iteration includes: obtaining the second estimate ( ⁇ ′ m ⁇ N ) of the N-frame diffraction pattern ( ⁇ I ⁇ N ) of the current iteration, wherein the second estimate ( ⁇ ′ m ⁇ N ) of the N-frame second distribution in the current iteration is an estimate of the complex amplitude distribution of the outgoing wave corresponding to the N-frame diffraction pattern ( ⁇ I ⁇ N ) at the detector; and determining the second estimate ( ⁇ ′′ m ⁇ N ) of the N-frame first distribution in the current iteration corresponding to the backpropagation of the second estimate ( ⁇ ′ m ⁇ N ) of the N-frame first distribution in the current iteration to the target object based on the first estimate ( ⁇ ⁇ ′ m ⁇ N ) of the N-frame first distribution in the current iteration and the second estimate ( ⁇ ′′ m - 1 ⁇ N
  • the second estimate ( ⁇ ′′ m ⁇ N ) of the first distribution of the N frames in the current iteration satisfies: in, ⁇ is the inverse Fourier transform function, k is the coordinate vector of the plane where the detector is located, n is a positive integer less than or equal to N, and ⁇ is linearly correlated with the first estimate ( ⁇ ′ m ⁇ N ) of the first distribution of the N frames and the second estimate ( ⁇ ′′ m-1 ⁇ N ) of the first distribution of the N frames in the previous iteration.
  • n is a positive integer less than or equal to N
  • OK (r;n) is the initial estimate of the target complex amplitude distribution (O)
  • PK (r) is the initial estimate of the probe complex amplitude distribution
  • ⁇ is a preset constant.
  • the target iteration in the m-th round further includes: when m ⁇ K, K is an integer and K ⁇ 0, updating the target complex amplitude distribution (O) based on the second estimate ( ⁇ ′′ m-1 ⁇ N ) of the first distribution of the N frames of the current iteration.
  • the estimate of the target complex amplitude distribution (O) after the current iteration update satisfies:
  • r is the coordinate vector of the plane where the target object is located
  • n is a positive integer less than or equal to N
  • O ⁇ sub>m-1 ⁇ /sub> (r;n) is the estimate of the complex amplitude distribution of the target after the previous iteration update
  • P ⁇ sub> m-1 ⁇ /sub> (r) is the estimate of the complex amplitude distribution of the probe after the previous iteration update
  • P ⁇ sub>m ⁇ /sub>(r) is the estimate of the complex amplitude distribution of the probe after the current iteration update
  • * represents the conjugate complex number
  • ⁇ sub> max ⁇ /sub> represents the non-negativity constraint
  • ⁇ sub> 1 ⁇ /sub>, ⁇ ⁇ sub>2 ⁇ /sub> , ⁇ ⁇ sub>3 ⁇ /sub> and ⁇ ⁇ sub>4 ⁇ /sub> are preset constants between 0 and 1.
  • this disclosure provides an imaging apparatus, comprising: at least one storage medium storing at least one set of instructions for implementing coherent diffraction imaging; and at least one processor communicatively connected to the at least one storage medium, wherein when the imaging apparatus is in operation, the at least one processor reads the at least one set of instructions and implements the method described in the first aspect.
  • the device further includes: a coherent light source configured to emit a coherent wave, wherein, during operation of the device, the coherent wave is constrained to illuminate a target object by a probe and diffracts; and a detector configured to receive the outgoing wave generated by the diffraction and generate N frames of diffraction patterns.
  • a coherent light source configured to emit a coherent wave, wherein, during operation of the device, the coherent wave is constrained to illuminate a target object by a probe and diffracts
  • a detector configured to receive the outgoing wave generated by the diffraction and generate N frames of diffraction patterns.
  • the coherent wave can be at least one of electromagnetic radiation, photons, X-rays, electrons, neutrons, and protons.
  • the imaging method P100 estimates the similarity distribution ( ⁇ s ) among the first distribution ( ⁇ N ) of N frames during the target iteration process, and then updates the first estimate ( ⁇ ′ ⁇ N ) of the first distribution of N frames based on the similarity distribution ( ⁇ s ). Since the similarity distribution ( ⁇ s ) is adjusted in each iteration as the estimate of the first distribution of N frames is continuously updated, the imaging method P100 can adaptively utilize the similarity information between frames in the iteration.
  • Using the similarity distribution ( ⁇ s ) as a constraint condition for iteratively updating the first estimate ( ⁇ ′ ⁇ N ) of the first distribution of N frames in the target iteration can improve the speed at which the estimate ( ⁇ ′ ⁇ N ) of the first distribution of N frames gradually approaches the true distribution ( ⁇ N ), thereby improving the iterative convergence speed of the target complex amplitude distribution (O).
  • Figure 1 shows a schematic diagram of an imaging apparatus provided according to some embodiments of the present disclosure
  • Figure 2 shows a schematic diagram of the hardware structure of a computing device according to some embodiments of the present disclosure
  • Figure 3 shows a flowchart of a coherent diffraction imaging method provided according to some embodiments of the present disclosure
  • Figure 4 illustrates a flowchart of applying inter-frame constraints according to some embodiments of the present disclosure
  • Figure 5 illustrates a flowchart of a target iteration provided according to some embodiments of the present disclosure
  • Figure 6 illustrates a flowchart of a target iteration provided according to some embodiments of the present disclosure.
  • Figures 7A-7H show experimental comparison results provided according to some embodiments of the present disclosure.
  • X includes at least one of A, B, or C
  • X includes at least A, or X includes at least B, or X includes at least C. That is, X can include only one of A, B, and C, or any combination of A, B, and C, as well as other possible content/elements.
  • the arbitrary combination of A, B, and C can be A, B, C, AB, AC, BC, or ABC.
  • the relationships between structures can be direct or indirect.
  • A when describing "A is connected to B,” unless it is explicitly stated that A and B are directly connected, it should be understood that A can be directly connected to B or indirectly connected to B.
  • A when describing "A is above B,” unless it is explicitly stated that A is directly above B (AB is adjacent and A is above B), it should be understood that A can be directly above B or indirectly above B (AB is separated by other elements, and A is above B). And so on.
  • Optical imaging technology has a profound impact on scientific research and industrial applications.
  • Short-wavelength radiation such as X-rays and high-energy electrons, is the primary means of achieving nanometer or atomic resolution imaging.
  • traditional optical microscopes are no longer sufficient to meet the imaging characterization needs of emerging fields such as materials science and biology.
  • Coherent diffraction imaging provides a powerful imaging method for wavelengths where high-quality lenses are difficult to fabricate.
  • Coherent diffraction imaging uses the interaction between a coherent light source and the sample to form a diffraction field, and reconstructs the amplitude and phase information of the sample through iterative calculations. Because it does not require high-performance imaging optics, it can be widely applied to coherent radiation sources such as deep ultraviolet, X-rays, and electron beams.
  • the outgoing wave formed by the coherent wave passing through the sample under test is received by the detector in the form of a diffraction pattern.
  • the diffraction pattern only records amplitude information and loses phase information. Therefore, phase recovery is needed to obtain the structural information of the sample.
  • an iterative update method can be used to gradually approximate the true phase distribution of the sample under test. Due to the non-uniqueness of the solution to the phase recovery problem, certain constraints need to be applied during iteration to update the phase. Typically, alternating application of support constraints and amplitude constraints can be used to gradually converge the phase estimation.
  • This disclosure provides a novel coherent diffraction imaging method that adds adaptive constraints to the iteration to fully utilize the similarity information of varying samples in multi-frame sampling, thereby improving the iterative convergence speed while obtaining better phase retrieval results.
  • FIG. 1 shows a schematic diagram of an imaging apparatus 001 provided according to some embodiments of the present disclosure.
  • the imaging apparatus 001 includes a coherent light source 100, a detector 200, a storage medium, and a processor.
  • a coherent light source 100 emits a coherent wave 110 that illuminates a target object 002.
  • the target object 002 is the sample to be tested.
  • the size of the sample to be tested is matched to the wavelength of the coherent wave 110 to ensure that a sufficiently strong diffraction signal is generated for resolution, while also meeting the required imaging resolution.
  • the target object 002 can be a sample that changes over time.
  • the target object 002 can be a growing crystal, a moving cell, a microorganism, a virus, etc.
  • the imaging device 001 can perform coherent diffraction imaging on the target object 002 that changes over time to obtain structural information of the target object 002 at multiple moments.
  • imaging device 001 can use different types of coherent waves 110.
  • the imaging device 001 provided in this disclosure can be applied to a variety of applications.
  • the specific type of coherent light source 100 used to generate the required coherent waves 110 depends on factors such as the properties of the target 002, the required resolution, and the research objective.
  • the coherent wave 110 can be electromagnetic radiation, such as the high-frequency components of visible light, near-infrared, ultraviolet, vacuum ultraviolet, and even X-rays. Electromagnetic radiation as the coherent wave 110 can be widely used in materials science, biomedicine, semiconductor devices, nanostructure research, and other fields. For example, it can be used for surface morphology analysis of micro/nano structures, crystal structure determination, and protein molecular structure analysis.
  • the coherent wave 110 can be a photon.
  • a photon is the quantum unit of electromagnetic radiation. Photon coherent diffraction imaging can be applied in emerging fields such as quantum information processing and quantum imaging.
  • the coherent wave 110 can be X-rays.
  • X-rays have the characteristics of short wavelength and high penetrating power, allowing them to penetrate deep into the interior of a target object.
  • X-ray coherence can be guaranteed by advanced light sources such as synchrotron radiation sources and free-electron lasers.
  • X-rays as coherent waves 110, can be applied to material structure analysis (such as crystal structures, amorphous materials, and complex macromolecules), nanoscale imaging (such as nanoparticles, biomacromolecules, and intracellular structures), materials science (defect detection, stress analysis), and cultural heritage protection (non-destructive testing).
  • the coherent wave 110 can be electrons.
  • the wave nature of electrons requires high energies to be significant, forming electron waves.
  • Electron diffraction mainly occurs in electron microscopy, where a high-speed electron beam emitted by an electron gun interacts with the sample to produce coherent diffraction patterns.
  • Electron waves, as coherent waves 110 can be applied to electron microscopy (including transmission electron microscopy TEM and scanning electron microscopy SEM), allowing direct observation of the sample's microstructure, chemical composition, crystallographic information, etc.
  • the coherent wave 110 can be neutrons.
  • Neutrons because they have no electric charge and have a specific interaction with the atomic nucleus, can penetrate many materials that are opaque to electromagnetic radiation and are selectively sensitive to certain elements and isotopes.
  • Neutron beams generated by neutron sources (such as reactors or spallation neutron sources) can achieve coherent diffraction.
  • the coherent wave 110 can be used to study the magnetic structure of materials, hydrogen/deuterium positions, stress distribution, and compositional distribution of complex alloys, and has important applications in condensed matter physics, earth sciences, energy materials, and biomedicine.
  • the coherent wave 110 can be a proton.
  • the proton exhibits significant wave properties at extremely high energies.
  • Proton coherent diffraction can play a role in the study of matter structure under extreme conditions, nuclear physics, and cosmic ray detection.
  • the coherent wave 110 is constrained so that the probe 120 illuminates the target object 002 and undergoes diffraction.
  • the plane perpendicular to the axis of the probe 120 and containing the target object 002 can be designated as target plane A.
  • Target plane A is the plane in which the probe 120 passes through the target object 002 and undergoes diffraction. After penetrating the target object 002, the probe 120 forms an outgoing wave, which travels a certain distance before reaching the detector 200.
  • the probe 120 can be a coherent wave constrained to a certain size or shape (e.g., focused into a beam or narrow pulse).
  • the imaging device 001 can obtain a higher energy density, which helps to obtain a sufficiently strong diffraction signal for effective imaging.
  • the imaging device 001 can ensure that the wavefront incident on the target object 002 has high spatial coherence, thereby producing a clear, resolvable, high-resolution diffraction pattern during diffraction.
  • the imaging device 001 can provide higher spatial resolution by using a smaller probe 120.
  • detector 200 receives the outgoing waves generated by diffraction on target object 002 and generates N frames of diffraction patterns.
  • the plane on which detector 200 receives the outgoing waves can be designated as detection plane B.
  • probe 120 passes through target object 002, it carries the microstructural information of target object 002; that is, after diffraction, the outgoing waves carry the information of target object 002 to detection plane B.
  • Detector 200 can record the intensity information of the outgoing waves at different positions on detection plane B as they propagate to detection plane B, forming diffraction patterns. Different positions in the diffraction pattern correspond to different frequencies.
  • the distance between detector 200 and target object 002 satisfies the far-field diffraction condition.
  • the intensity of each point in the diffraction pattern is determined by the Fourier transform relationship; that is, the complex amplitude distribution (first distribution) of the outgoing wave on target plane A, after Fourier transform, forms a corresponding complex amplitude distribution (second distribution) on the far-field detection plane B.
  • the imaging device 001 is capable of acquiring diffraction patterns of the target object 002 at a specific sampling frequency.
  • detector 200 can acquire multiple frames of diffraction patterns at multiple sampling times. These multiple sampling times can be equally spaced or unequally spaced.
  • Target object 002 exhibits spatial local consistency at corresponding multiple diffraction times (i.e., the times when diffraction occurs on target plane A corresponding to the sampling times of the diffraction pattern on detector plane B).
  • target object 002 is a continuously growing crystal. At adjacent diffraction times, some regions of target object 002 change, while other regions remain relatively stationary. Or, for example, target object 002 is an object moving within target plane A. At adjacent diffraction times, target object 002 may undergo overall displacement, but the internal region containing target object 002 remains relatively stationary. Therefore, from the perspective of target plane A, target object 002 exhibits spatial similarity at multiple diffraction times. This is reflected in detector plane B, where inter-frame similarity also exists between multiple diffraction patterns.
  • the imaging device 001 can adapt to application scenarios in different fields.
  • the selection of the detector 200 needs to consider the type of coherent wave 110 used, as well as imaging requirements (such as spatial resolution, energy resolution, dynamic range, temporal resolution, etc.), experimental conditions (such as sample type, experimental environment, data acquisition speed, etc.) and economic factors.
  • detector 200 may be a CCD (Charge-Coupled Device) detector. CCD detectors offer high quantum efficiency and low readout noise, making them suitable for capturing weak diffraction signals.
  • detector 200 may be a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor. CMOS image sensors offer low cost and high readout speed.
  • detector 200 may be a photodiode array detector.
  • detector 200 may be a photon counting detector.
  • detector 200 may be a silicon photomultiplier (SiPM), a single photon avalanche diode (SPAD) detector.
  • detector 200 may be an X-ray detector.
  • detector 200 may be a scintillator-coupled detector, a silicon-based direct electron detector (DEDS), or a hybrid pixel detector (HEPDs).
  • detector 200 can be an electronic detector.
  • detector 200 can be a scintillation detector, a direct electron counting detector, or a fluorescent screen-CCD system detector.
  • detector 200 can be a neutron detector.
  • a scintillator detector where neutrons interact with the scintillator material to generate visible light, which is then recorded by a photoelectric conversion device.
  • a 3He tube detector is another example.
  • a solid-state neutron detector which can directly convert charged particles generated by neutron capture reactions into electrical signals.
  • the imaging device 001 includes at least one storage medium and at least one processor.
  • the at least one processor is communicatively connected to the at least one storage medium.
  • the at least one storage medium stores at least one set of instructions for implementing the coherent diffraction imaging method provided in this disclosure.
  • the at least one processor can be communicatively connected to the detector 200 to apply the acquired N frames of diffraction patterns to the implementation of the coherent diffraction imaging method provided in this disclosure.
  • the at least one processor reads the at least one set of instructions and implements the coherent diffraction imaging method provided in this disclosure.
  • the coherent diffraction method provided in this disclosure will be described in detail below.
  • the imaging device 001 may integrate the at least one storage medium and the at least one processor into a computing device 300.
  • the imaging device 001 may include one or more computing devices 300.
  • Figure 2 illustrates a schematic hardware structure of a computing device 300 according to some embodiments of the present disclosure.
  • the computing device 300 includes at least one storage medium 330 and at least one processor 320.
  • the computing device 300 may further include an internal communication bus 310.
  • the computing device 300 may further include a communication port 350.
  • the computing device 300 may further include I/O components 360.
  • the internal communication bus 310 can connect different system components, including storage medium 330 and processor 320.
  • I/O component 360 supports input/output between computing device 300 and other components.
  • Storage medium 330 may include a data storage device.
  • the data storage device may be a non-transitory storage medium or a temporary storage medium.
  • the data storage device may include one or more of a disk 332, a read-only storage medium (ROM) 334, or a random access storage medium (RAM) 336.
  • Storage medium 330 also includes at least one instruction set stored in the data storage device.
  • the instruction set is computer program code, which may include programs, routines, objects, components, data structures, procedures, modules, etc., that execute the related diffraction imaging methods provided in this disclosure.
  • Communication port 350 is used for data communication between computing device 300 and external sources.
  • computing device 300 can connect to network 330 through communication port 350.
  • At least one processor 320 is communicatively connected to at least one storage medium 330 via internal communication bus 310.
  • At least one processor 320 is used to execute at least one instruction set.
  • the correlation diffraction imaging system 001 is running, at least one processor 320 reads at least one instruction set and executes the correlation diffraction imaging method provided in this disclosure according to the instructions of at least one instruction set.
  • Processor 320 can perform all the steps included in the relevant diffraction imaging method.
  • Processor 320 can be in the form of one or more processors.
  • Processor 320 can issue execution instructions.
  • Processor 320 may include one or more hardware processors, such as microcontrollers, microprocessors, reduced instruction set computers (RISC), application-specific integrated circuits (ASICs), application-specific instruction set processors (ASIPs), central processing units (CPUs), graphics processing units (GPUs), physical processing units (PPUs), microcontroller units, digital signal processors (DSPs), field-programmable gate arrays (FPGAs), advanced RISC machines (ARMs), programmable logic devices (PLDs), any circuit or processor capable of performing one or more functions, or any combination thereof.
  • RISC reduced instruction set computers
  • ASICs application-specific integrated circuits
  • ASIPs application-specific instruction set processors
  • CPUs central processing units
  • GPUs graphics processing units
  • PPUs physical processing units
  • DSPs digital signal
  • processor 320 For illustrative purposes only, only one processor 320 is described in the computing device 300 of this disclosure. However, it should be noted that the computing device 300 of this disclosure may also include multiple processors. Therefore, the operation and/or method steps disclosed in this disclosure may be executed by one processor as described in this disclosure, or they may be executed jointly by multiple processors. For example, if the processor 320 of the computing device 300 of this disclosure executes steps A and B, it should be understood that steps A and B may also be executed jointly or separately by two different processors 320 (e.g., the first processor executes step A, the second processor executes step B, or the first and second processors jointly execute steps A and B).
  • the imaging device 001 can emit coherent waves 110 through a coherent light source 100 and constrain the coherent waves 110 to form a precise probe 120.
  • the probe 120 illuminates the target object 002 and, after diffraction, forms an outgoing wave.
  • the outgoing waves at multiple moments propagate to the far-field detection plane B and are received by the detector 200, forming multi-frame diffraction patterns.
  • the computing device 300 recovers the complex amplitude distribution of the target object 002 based on the received multi-frame diffraction patterns.
  • the imaging device 001 can be applied to coherent wave diffraction imaging in multiple bands, can adapt to various types of detectors 200, has a simple structure, wide adaptability, and broad application scenarios.
  • the coherent diffraction imaging method (hereinafter referred to as the imaging method) provided in this disclosure will now be described in detail.
  • Figure 3 shows a flowchart of an imaging method P100 provided according to some embodiments of the present disclosure.
  • the processor can read an instruction set stored in its local storage medium and/or database, and then execute the imaging method P100 according to the instructions in the instruction set.
  • the imaging method P100 provided in this disclosure includes:
  • the coherent wave 110 generated by the coherent light source 100 diffracts to form an outgoing wave when it passes through the target object 002.
  • the outgoing wave forms N frames of the first distribution ( ⁇ N ) at the target object 002.
  • N is a positive integer
  • the symbol ⁇ N represents the set of N frames.
  • the N frames of the first distribution ( ⁇ N ) is the actual spatial distribution of the outgoing wave at the target plane A.
  • ⁇ (r;n) represents the distribution of the nth frame in the N frames of the first distribution at the target plane A, where r is the coordinate vector on the target plane A.
  • ⁇ (r;n) is abbreviated as ⁇ in the following text.
  • the N frames of the first distribution ( ⁇ N ) is the complex amplitude distribution of the outgoing wave, including amplitude and phase information.
  • the N-frame second distribution ( ⁇ N ) represents the complex amplitude distribution of the outgoing wave as it propagates to the detection plane B.
  • ⁇ (k;n) represents the distribution of the nth frame in the N-frame second distribution at the detection plane B, where k is the coordinate vector on the detection plane B.
  • ⁇ (k;n) is abbreviated as ⁇ in the following text.
  • the N-frame first distribution ( ⁇ N ) and its corresponding N-frame second distribution ( ⁇ N ) satisfy a Fourier transform relationship.
  • the N-frame diffraction pattern ( ⁇ I ⁇ N ) represents the intensity information of the N-frame second distribution ( ⁇ N ) obtained by the detector 200.
  • the detector 200 can record the intensity information of the N-frame second distribution ( ⁇ N ) at different positions on the detection plane B, acquired sequentially over time, and form the N-frame diffraction pattern.
  • the intensity information is positively correlated with the amplitude of the N-frame second distribution ( ⁇ N ).
  • Different positions on the diffraction pattern ( ⁇ I ⁇ N ) correspond to different frequencies.
  • the target complex amplitude distribution (O) of the target object is obtained by performing target iteration until the target complex amplitude distribution (O) satisfies the preset convergence condition.
  • the target complex amplitude distribution (O) is an estimate of the complex amplitude distribution of target object 002 on the target plane A, including amplitude and phase information.
  • the target complex amplitude distribution (O) can be separated from the N-frame first distribution ( ⁇ N ) of the outgoing wave.
  • Imaging method P100 obtains the target complex amplitude distribution (O) by performing multiple target iterations to continuously update the estimate of the N-frame first distribution ( ⁇ N ).
  • Target object 002 exhibits spatial local consistency across N diffraction occurrence times. From target plane A, target object 002 displays spatial similarity across these N diffraction occurrence times, meaning there is a similar distribution ( ⁇ s ) among the N frames of the first distribution ( ⁇ N ). This is reflected in detector plane B, where inter-frame similarity also exists among the N frames of diffraction patterns ( ⁇ I ⁇ N ) . Imaging method P100 applies this similarity as an inter-frame constraint between the N frames of the first distribution ( ⁇ N ) during the estimation process.
  • the inter-frame constraint includes updating the similarity distribution ( ⁇ s ) among the N-frame first distribution ( ⁇ N ) in each target iteration to obtain a first estimate ( ⁇ ′ ⁇ N ) of the N-frame first distribution.
  • the first estimate ( ⁇ ′ ⁇ N) is an estimate of the N-frame first distribution ( ⁇ N ) obtained based on the similarity distribution ( ⁇ s ) in each target iteration.
  • the imaging method P100 can estimate the N -frame first distribution ( ⁇ N ) in one or more ways, and using the similarity distribution ( ⁇ s ) to estimate and update it is one of them.
  • the target complex amplitude distribution (O) is an estimate of the amplitude and phase of the target object 002.
  • the structure of the target object 002 can be presented either as an image reconstructed from the target complex amplitude distribution (O), or by visualizing the amplitude and phase information separately.
  • the imaging method P100 estimates the similarity distribution ( ⁇ s ) among the first distribution ( ⁇ N ) of N frames during target iteration, and then updates the first estimate ( ⁇ ′ ⁇ N ) of the first distribution of N frames based on the similarity distribution ( ⁇ s ). Since the similarity distribution ( ⁇ s ) is adjusted in each iteration as the estimate of the first distribution of N frames is continuously updated, the imaging method P100 can adaptively utilize the similarity information between frames during iteration.
  • Using the similarity distribution ( ⁇ s ) as a constraint for iteratively updating the first estimate ( ⁇ ′ ⁇ N ) of the first distribution of N frames in target iteration can improve the speed at which the estimate ( ⁇ ′ ⁇ N ) of the first distribution of N frames gradually approaches the true distribution ( ⁇ N ), thereby improving the iterative convergence speed of the target complex amplitude distribution (O).
  • step S130 the imaging method P100 updates the similarity distribution ( ⁇ s ) in each round of target iteration to obtain a first estimate ( ⁇ ′ ⁇ N ) of the first distribution of the N frames.
  • the similarity distribution ( ⁇ s ) is continuously updated as the estimate of the first distribution ( ⁇ ′ ⁇ N ) of the N frames changes.
  • Figure 4 illustrates a flowchart of applying inter-frame constraints according to some embodiments of the present disclosure.
  • updating the similarity distribution ( ⁇ s ,m ) among the N-frame first distribution ( ⁇ N ) to obtain a first estimate ( ⁇ ⁇ ′m ⁇ N ) of the N-frame first distribution may include:
  • S135-1 Determine the similarity (S m ) of the initial estimate ( ⁇ 0m ⁇ N ) of the first distribution of the N frames in the current iteration.
  • the initial estimate ( ⁇ ⁇ 0m ⁇ N ) of the first distribution of N frames is the estimate of the first distribution ( ⁇ N ) of N frames at the start of the current iteration.
  • the initial estimate ( ⁇ ⁇ 0m ⁇ N ) of the first distribution of N frames needs to be initialized.
  • imaging device 001 can perform random initialization or zero-phase initialization on the initial estimate ( ⁇ ⁇ 0m ⁇ N ) of the first distribution of N frames.
  • determining the similarity ( Sm ) of the initial estimates ( ⁇ ⁇ 0m ⁇ N ) of the N-frame first distribution in the current iteration may include: calculating a dispersion index of the initial estimates ( ⁇ ⁇ 0m ⁇ N ) of the N-frame first distribution in the current iteration, and obtaining the similarity ( Sm ) based on the dispersion index.
  • the dispersion index can be used to reflect the degree of dispersion among the N-frame first distributions ( ⁇ N ), or the degree of deviation from the mean.
  • the dispersion index can be the standard deviation or the variance. It can be considered that the larger the dispersion index, the lower the similarity ( Sm ) among the initial estimates ( ⁇ ⁇ 0m ⁇ N ) of the N-frame first distributions.
  • S135-3 Based on the similarity (S m ), determine the similarity distribution ( ⁇ s,m ) and the difference distribution ( ⁇ e ,m ⁇ N ) between the initial estimates ( ⁇ 0m ⁇ N ) of the first distribution of the N frames in the current iteration.
  • the similarity distribution ( ⁇ ⁇ sub>s,m ⁇ /sub> ) is an estimate of the similar components in the first distribution ( ⁇ sub> N ⁇ /sub> ) of N frames.
  • each frame can be composed of a similarity distribution ( ⁇ ⁇ sub>s ⁇ /sub> ) and a difference distribution ( ⁇ ⁇ sub> e ⁇ /sub> ).
  • the estimates of the similarity distribution ( ⁇ ⁇ sub>s,m ⁇ /sub> ) and the difference distribution ( ⁇ sub>e ⁇ /sub>,m ⁇ ⁇ sub>N ⁇ /sub> ) of N frames can be updated based on the similarity (S ⁇ sub>m ⁇ /sub>).
  • the first estimate ( ⁇ ′ ⁇ N ) of the first distribution of N frames is the estimate of the first distribution ( ⁇ N ) of N frames obtained based on the similarity distribution ( ⁇ s ) in each target iteration.
  • the imaging method P100 estimates the first distribution ( ⁇ N ) of N frames in one or more ways, and the method of using the similarity distribution ( ⁇ s ) to estimate and update it is one of them.
  • the first estimate ( ⁇ ⁇ ′m ⁇ N ) of the first distribution of N frames is the result after applying inter-frame constraints to the initial estimate ( ⁇ 0m ) of the first distribution of N frames.
  • Each frame ( ⁇ ′m) in the first estimate ( ⁇ ⁇ ′m ⁇ N ) of the first distribution of N frames can be composed of a similar distribution ( ⁇ s ,m ) and a difference distribution ( ⁇ e ,m ).
  • the first estimate ( ⁇ ⁇ ′m ⁇ N ) of the first distribution of N frames can be expressed as the sum or weighted sum of the similar distribution ( ⁇ s,m ) and the difference distribution ( ⁇ ⁇ e,m ⁇ N ) of N frames.
  • a similar distribution ( ⁇ sub> s ⁇ /sub> ) can also be obtained based on the distribution of the N-frame first distribution in the transform domain.
  • the transform domain can be the SVD (Singular Value Decomposition) transform domain.
  • the N-frame first distribution ( ⁇ sub> N ⁇ /sub>)
  • the dimension of the N-frame first distribution ( ⁇ sub> N ⁇ /sub> ) is x ⁇ y ⁇ N.
  • the N-frame first distribution ( ⁇ sub> N ⁇ /sub> ) is rearranged into a two-dimensional matrix ⁇ of size N ⁇ xy.
  • the first row of ⁇ corresponds to the first frame in the N-frame first distribution ( ⁇ sub> N ⁇ /sub> ), and so on for the other rows.
  • An SVD transformation is performed on the two-dimensional matrix ⁇ to decompose it into three matrices: U, S, and V.
  • U is the left singular vector matrix
  • S is the singular value matrix
  • V is the right singular vector matrix.
  • S is a diagonal matrix.
  • V(1,:) represents the principal components of the normalized matrix.
  • the imaging method P100 provided in this disclosure can characterize the similarity distribution ( ⁇ s) in the N-frame first distribution ( ⁇ N ) based on the similarity (S) between the N-frame first distribution ( ⁇ N ) or between the corresponding N-frame distributions of the N-frame first distribution in the transform domain, and then obtain the first estimate ( ⁇ ⁇ ′ ⁇ N ) of the N-frame first distribution based on the similarity distribution ( ⁇ s ).
  • the similarity ( Sm ), the similarity distribution ( ⁇ s,m ), the N-frame difference distribution ( ⁇ ⁇ e,m ⁇ N ), and the first estimate ( ⁇ ⁇ ′m ⁇ N ) of the N-frame first distribution are all updated.
  • the imaging method P100 makes full use of the similarity information between the N-frame first distributions ( ⁇ N ), and the estimate of the N-frame first distribution ( ⁇ N ) can converge at a relatively fast speed.
  • each frame can be obtained by a weighted sum of the similarity distribution ( ⁇ s,m ) and the difference distribution ( ⁇ e ,m ).
  • the coefficient ⁇ of the difference distribution ( ⁇ e,m ) is used to maintain the constant energy in the current iteration.
  • the weighting coefficient ⁇ can be...
  • the summation symbol applies to the first distribution of N frames.
  • ⁇ sub>m ⁇ /sub> is the average value of the initial estimate ( ⁇ ⁇ sub>0m ⁇ /sub> ⁇ ⁇ sub>N ⁇ /sub> ) of the first distribution of the N frames.
  • ⁇ sub> m ⁇ /sub> can be the standard deviation distribution of the initial estimate ( ⁇ ⁇ sub>0m ⁇ /sub> ⁇ sub> N ⁇ /sub> ) of the first distribution of the N frames.
  • C ⁇ sub> m ⁇ /sub> is a transformation function used to transform ⁇ sub> m ⁇ /sub> into the similarity S ⁇ sub>m ⁇ /sub> .
  • ⁇ , ⁇ ⁇ sub>m ⁇ /sub> , and ⁇ sub> m ⁇ /sub> are used to adjust the shape of the transformation function.
  • the transformation function C ⁇ sub>m ⁇ /sub> has an inverse S-shape, and its value is negatively correlated with ⁇ sub> m ⁇ /sub>.
  • ⁇ is a preset parameter value used to ensure that the transformation function C ⁇ sub>m ⁇ /sub> reaches a preset ⁇ when ⁇ sub> m ⁇ /sub> approaches zero.
  • s ⁇ sub>1 ⁇ /sub> is the average of the lowest 10% of ⁇ ⁇ sub> m ⁇ /sub> in the m-th iteration
  • s ⁇ sub>2 ⁇ /sub> is the average of the highest 10% of ⁇ sub>m ⁇ /sub>.
  • Imaging method P100 uses the average of the highest and lowest 10% instead of the highest and lowest values, achieving a more stable convergence effect and improving the robustness of the iteration process.
  • ⁇ sub> m ⁇ /sub> is positively correlated with ⁇ ⁇ sub>m ⁇ /sub>.
  • ⁇ m can be set as a multiple of ⁇ m , for example, ⁇ m can be set as 50 times the standard deviation of ⁇ m .
  • the imaging method P100 estimates the first distribution ( ⁇ N ) of N frames using one or more methods, and using a similar distribution ( ⁇ s ) for estimation and updating is one such method.
  • the first estimate ( ⁇ ⁇ ′m ⁇ N ) of the first distribution of N frames is the result of applying inter-frame constraints to the initial estimate ( ⁇ 0m ) of the first distribution of N frames.
  • the imaging method P100 can also apply other constraints to the first distribution of N frames to obtain the initial estimate ( ⁇ ⁇ 0m ⁇ N ) and the second estimate ( ⁇ ′′ m ⁇ N ) of the first distribution of N frames.
  • Figure 5 illustrates a flowchart of a target iteration provided according to some embodiments of the present disclosure.
  • the target iteration may include:
  • the second estimate ( ⁇ ′′ m-1 ⁇ N ) of the first distribution of N frames is the estimation result of the first distribution ( ⁇ N ) of N frames obtained after the previous iteration.
  • the imaging method P100 applies a support constraint to the second estimate ( ⁇ ′′ m-1 ⁇ N ) of the N-frame first distribution to limit the estimate of the N-frame first distribution ( ⁇ N ) within the constraint range.
  • Obtaining the initial estimate ( ⁇ 0m ⁇ N ) of the N-frame first distribution in the current iteration based on the second estimate ( ⁇ ′′ m-1 ⁇ N ) of the N-frame first distribution from the previous iteration includes: when the second estimate ( ⁇ ′′ m-1 ⁇ N ) of the N-frame first distribution is located within the support region, using the second estimate ( ⁇ ′′ m-1 ⁇ N ) of the N-frame first distribution as the initial estimate ( ⁇ 0m ⁇ N ); or when the second estimate ( ⁇ ′′ m-1 ⁇ N ) of the N-frame first distribution is located outside the support region, the value of the second estimate ( ⁇ ′′ m-1 ⁇ N ) of the N-frame first distribution in that region is zero.
  • S135 Apply the inter-frame constraint to the initial estimate ( ⁇ ⁇ 0m ⁇ N ) of the first distribution of the N frames to obtain the first estimate ( ⁇ ⁇ ′m ⁇ N ) of the first distribution of the N frames in the current iteration.
  • the method for applying the inter-frame constraint to the initial estimate ( ⁇ 0m ⁇ N ) of the first distribution of the N frames is as described in S135-1 to S135-5 above, and will not be repeated here.
  • the imaging method P100 propagates the first estimate ( ⁇ ⁇ ′m ⁇ N ) of the N-frame first distribution in the current iteration and the second estimate ( ⁇ ′′ m-1 ⁇ N ) of the N-frame first distribution in the previous iteration to the detector plane B, and applies a modulus constraint to ensure energy conservation during propagation.
  • the imaging method P100 can obtain the estimate ( ⁇ ′m ⁇ N ) of the N-frame second distribution in the current iteration based on the N-frame diffraction pattern ( ⁇ I ⁇ N ).
  • the estimate ( ⁇ ⁇ ′m ⁇ N ) of the N-frame second distribution in the current iteration is an estimate of the complex amplitude distribution of the outgoing wave at the detector corresponding to the N-frame diffraction pattern ( ⁇ I ⁇ N ). Since the N-frame first distribution ( ⁇ N ) at the target plane A and the N-frame second distribution ( ⁇ N ) at the detector plane B satisfy a Fourier transform relationship, the initial estimate ⁇ 0m of the N-frame second distribution in the current iteration at the detector plane B can be obtained based on the Fourier transform. In some embodiments, in, This represents the Fourier transform. This method effectively avoids iterative stagnation by increasing the increment ⁇ .
  • ⁇ ⁇ ′m (r;n) - ⁇ ′′ m-1 (r;n).
  • the coefficients of ⁇ ′m and ⁇ ′′ m-1 can be adjusted according to actual needs.
  • a modulus constraint is applied to the initial estimate of the N-frame second distribution to obtain the N-frame second distribution estimate ( ⁇ ⁇ ′m ⁇ N ).
  • the modulus constraint can be applied to each frame in the initial estimate of the N-frame second distribution in the following ways: Where I corresponds to each frame in the N-frame diffraction pattern ( ⁇ I ⁇ N ).
  • the imaging method P100 obtains a second estimate ( ⁇ ′′ m ⁇ N ) of the N-frame first distribution by backpropagating the second distribution estimate ( ⁇ ′ m ⁇ N) on the detection plane B to the target plane A.
  • the imaging method P100 can determine the second estimate ( ⁇ ′′ m ⁇ N ) of the N-frame first distribution corresponding to the backpropagation of the second distribution estimate ( ⁇ ′ m ⁇ N ) of the current iteration to the target object based on the first estimate ( ⁇ ′ m ⁇ N ) of the first distribution of the N-frames in the current iteration and the second estimate ( ⁇ ′′ m -1 ⁇ N ) of the first distribution of the N-frames in the previous iteration.
  • the nth frame in the second estimate ( ⁇ ′′ m ⁇ N ) of the first distribution of the N -frames in the current iteration satisfies: in This represents the inverse Fourier transform.
  • the ⁇ value is linearly related to the first estimate ( ⁇ ⁇ ′m ⁇ N ) of the first distribution of the N frames and the second estimate ( ⁇ ′′ m-1 ⁇ N ) of the first distribution of the N frames from the previous iteration.
  • ⁇ ⁇ ′m (r;n) - ⁇ ′′ m-1 (r;n).
  • the coefficients of ⁇ ′m and ⁇ ′′ m-1 can be adjusted according to actual needs. This method, by increasing the increment ⁇ , can effectively prevent the iteration from stalling.
  • the imaging method P100 updates the estimated value of the N-frame first distribution ( ⁇ N ) by propagating the N-frame first distribution ( ⁇ N ) back and forth between the target plane A and the detector plane B. Specifically, in each iteration (taking the m-th iteration as an example), the inter-frame constraint is applied to the initial estimate ( ⁇ ⁇ 0m ⁇ N ) of the N-frame first distribution to obtain the first estimate ( ⁇ ⁇ ′m ⁇ N ) of the N-frame first distribution in the current iteration. This fully utilizes the similarity information between the initial estimates ( ⁇ ⁇ 0m ⁇ N ) of the N-frame first distribution, which is beneficial for improving the convergence speed.
  • FIG. 6 illustrates a flowchart of the target iteration provided according to some embodiments of this disclosure.
  • K can be a preset number of iteration rounds.
  • the target complex amplitude distribution (O) is initialized. The initial estimate of the target complex amplitude distribution (O) satisfies:
  • O K (r; n) is the initial estimate of the nth frame of the target complex amplitude distribution (O)
  • P K (r) is the initial estimate of the probe complex amplitude distribution
  • ⁇ is a preset constant used to prevent the denominator from having a zero value.
  • the target complex amplitude distribution (O) is updated based on the second estimate ( ⁇ ⁇ m ⁇ N ) of the first distribution in the N frames of the current iteration.
  • the imaging method P100 can obtain initial estimates of the target complex amplitude distribution (O) and the probe complex amplitude distribution (P) according to the above initialization method. After obtaining the initial estimate of the probe complex amplitude distribution (P), support constraints and modulus constraints can be applied to it for updating.
  • O ⁇ sub>m-1 ⁇ /sub>(r;n) is the estimate of the target complex amplitude distribution after the previous iteration update.
  • P ⁇ sub>m-1 ⁇ /sub> (r) is the estimate of the probe complex amplitude distribution after the previous iteration update.
  • P ⁇ sub> m ⁇ /sub>(r) is the estimate of the probe complex amplitude distribution after the current iteration update. * denotes a conjugate complex number.
  • ⁇ sub>1 ⁇ /sub> , ⁇ ⁇ sub>2 ⁇ /sub> , ⁇ ⁇ sub>3 ⁇ /sub> , and ⁇ ⁇ sub>4 ⁇ /sub> are preset constants in the range of 0 to 1.
  • the iteration stops when the estimate of the target complex amplitude distribution (O) satisfies the preset convergence condition.
  • Figures 7A-7H show comparative experimental results provided according to some embodiments of the present disclosure.
  • Figures 7A-7H show numerical simulation imaging results of the growth process of polystyrene sulfonate (3,4-ethylenedioxythiophene, PEDOT:PSS) crystal samples.
  • the convergence speed of the imaging method P100 provided by the present disclosure is significantly faster than the conventional CDI method, reaching convergence within a few hundred iterations.
  • the imaging method P100 estimates the similarity distribution ( ⁇ s ) among the first distribution ( ⁇ N ) of N frames during target iteration, and then updates the first estimate ( ⁇ ′ ⁇ N ) of the first distribution of N frames based on the similarity distribution ( ⁇ s ). Since the similarity distribution ( ⁇ s ) is adjusted in each iteration as the estimate of the first distribution of N frames is continuously updated, the imaging method P100 can adaptively utilize the similarity information between frames during iteration.
  • Using the similarity distribution ( ⁇ s ) as a constraint condition for iteratively updating the first estimate ( ⁇ ′ ⁇ N ) of the first distribution of N frames in target iteration can improve the speed at which the estimate ( ⁇ ′ ⁇ N ) of the first distribution of N frames gradually approaches the true distribution ( ⁇ N ), thereby improving the iterative convergence speed of the target complex amplitude distribution (O).

Landscapes

  • Physics & Mathematics (AREA)
  • Chemical & Material Sciences (AREA)
  • Biochemistry (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Analytical Chemistry (AREA)
  • General Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Immunology (AREA)
  • Pathology (AREA)
  • Crystallography & Structural Chemistry (AREA)
  • Electromagnetism (AREA)
  • Analysing Materials By The Use Of Radiation (AREA)

Abstract

本公开提供一种相干衍射成像方法，包括：获得目标物的N帧衍射图样，其中，相干波经过所述目标物时发生衍射形成出射波，在N个发生时刻，所述出射波在所述目标物处分别形成N帧第一分布，所述N为正整数；基于所述N帧第一分布的帧间约束以及所述N帧衍射图样，通过执行目标迭代得到所述目标物的目标复振幅分布，直至所述目标复振幅分布满足预设的收敛条件，其中，所述帧间约束包括在每次目标迭代中通过对所述N帧第一分布之间的相似分布进行更新，从而得到所述N帧第一分布的第一估计；以及输出所述目标复振幅分布。本公开提供的成像方法具有较好的鲁棒性和收敛性。

Description

一种相干衍射成像方法及装置 技术领域
本公开涉及光学衍射成像领域，尤其涉及一种相干衍射成像方法及装置。
 背景技术
相干衍射成像(Coherent Diffractive Imaging，CDI)是一种无透镜成像技术，利用光波或X射线等电磁波的相干性来重构物体的结构信息。该技术突破了传统光学显微镜的衍射极限，能够在原子尺度上解析物质的结构，对材料科学、生物医学、物理学等领域具有重大意义。
相干衍射成像的基本原理基于傅里叶变换和波动光学理论。在实际操作中，衍射信息通常是以衍射图样的形式被探测器接收。衍射图样只能记录到振幅信息而丢失了相位信息，因此需要通过算法进行相位恢复，进而得到样品的结构信息。
本公开提供一种相干衍射成像方法，在提高迭代收敛速度的同时获得更好的样品成像结果。
背景技术部分的内容仅仅是申请人个人所知晓的信息，并不代表上述信息在本公开申请日之前已经进入公共领域，也不代表其可以成为本公开的现有技术。
 发明内容
本公开提供一种相干衍射成像方法及装置，能够提高相位恢复的迭代收敛速度。
第一方面，本公开提供一种相干衍射成像方法，包括：获得目标物的N帧衍射图样({I}N)，其中，相干波经过所述目标物时发生衍射形成出射波，在N个发生时刻，所述出射波在所述目标物处分别形成N帧第一分布({ψ}N)，所述出射波沿出射方向传播至探测器时所述N帧第一分布({ψ}N)分别变成N帧第二分布({Ψ}N)，所述N帧衍射图样({I}N)为所述探测器得到的所述N帧第二分布({Ψ}N)的强度信息，所述N为正整数；基于所述N帧第一分布({ψ}N)的帧间约束以及所述N帧衍射图样({I}N)，通过执行目标迭代得到所述目标物的目标复振幅分布(O)，直至所述目标复振幅分布(O)满足预设的收敛条件，其中，所述帧间约束包括在每次目标迭代中通过对所述N帧第一分布({ψ}N)之间或所述N帧第一分布在变换域中对应的N帧分布之间的相似分布(ψs)进行更新，从而得到所述N帧第一分布的第一估计({ψ′}N)；以及输出所述目标复振幅分布(O)。
在一些实现方式中，对于第m轮迭代，m为正整数，所述通过对所述N帧第一分布({ψ}N)之间的相似分布(ψs,m)进行更新以便得到所述N帧第一分布的第一估计({ψ′m}N)包括：确定当前迭代的所述N帧第一分布的初始估计({ψ0m}N)的相似性(Sm)；基于所述相似性(Sm)，确定所述当前迭代中所述N帧第一分布的初始估计({ψ0m}N)之间的所述相似分布(ψs,m)和所述差异分布({ψe,m}N)；基于所述相似分布(ψs,m)和所述差异分布({ψe,m}N)，确定所述当前迭代的所述N帧第一分布的第一估计({ψ′m}N)。
在一些实现方式中，所述当前迭代的所述N帧第一分布的第一估计({ψ′m}N)满足：ψ′m(r；n)＝ψe,m(r；n)+ρψs,m(r)ψe,m(r；n)＝[1-Sm(r)]ψ0m(r；n)
其中，r为所述目标物所在平面的坐标矢量，n为小于等于N的正整数，ρ用于保持所述当前迭代中能量的恒定，是所述N帧第一分布的初始估计({ψ0m}N)的平均值，σm(r)为所述N帧第一分布的初始估计({ψ0m}N)的标准差分布，Cm为转换函数，用于将所述σm(r)转换成所述相似性Sm(r)，α、βm以及κm用于调整所述转换函数的形状。
在一些实现方式中，对于第m轮迭代，m为正整数，所述目标迭代包括：得到上一轮迭代的所述N帧第一分布的第二估计({ψ″m-1}N)；基于所述上一轮迭代的所述N帧第一分布的第二估计({ψ″m-1}N)获得当前迭代的N帧第一分布的初始估计({ψ0m}N)；对所述N帧第一分布的初始估计({ψ0m}N)施加所述帧间约束，得到当前迭代的所述N帧第一分布的第一估计({ψ′m}N)；基于所述当前迭代的所述N帧第一分布的第一估计({ψ′m}N)、所述上一轮迭代的所述N帧第一分布的第二估计({ψ″m-1}N)、以及所述N帧衍射图样({I}N)，得到当前迭代的所述N帧第一分布的第二估计({ψ″m}N)。
在一些实现方式中，所述基于所述上一轮迭代的所述N帧第一分布的第二估计({ψ″m-1}N)获得当前迭代的所述N帧第一分布的初始估计({ψ0m}N)包括：当所述N帧第一分布的第二估计({ψ″m-1}N)位于支撑区域内时，以所述N帧第一分布的第二估计({ψ″m-1}N)作为N帧第一分布的初始估计({ψ0m}N)；或者当所述N帧第一分布的第二估计({ψ″m-1}N)位于支撑区域外时，所述N帧第一分布的第二估计({ψ″m-1}N)在该区域的值为零。
在一些实现方式中，所述得到当前迭代的所述N帧第一分布的第二估计({ψ″m}N)包括：基于所述N帧衍射图样({I}N)，得到所述当前迭代的N帧第二分布估计({Ψ′m}N)，所述当前迭代的所述N帧第二分布估计({Ψ′m}N)为所述N帧衍射图样({I}N)所对应的所述出射波在所述探测器处的复振幅分布的估计值；以及基于当前迭代的所述N帧第一分布的第一估计({ψ′m}N)和上一轮迭代的所述N帧第一分布的第二估计({ψ″m-1}N)，确定所述当前迭代的N帧第二分布估计({Ψ′m}N)反向传播到目标物处所对应的当前迭代的所述N帧第一分布的第二估计({ψ″m}N)。
在一些实现方式中，所述当前迭代的所述N帧第一分布的第二估计({ψ″m}N)满足：其中，为逆傅里叶变换函数，所述k为所述探测器所在平面的坐标矢量，所述n为小于等于N的正整数，所述δ与所述N帧第一分布的第一估计({ψ′m}N)以及所述上一轮迭代的所述N帧第一分布的第二估计({ψ″m-1}N)成线性相关。
在一些实现方式中，所述第m轮所述目标迭代还包括：当m＝K时，所述K为整数且K≥0，基于所述当前迭代的N帧第一分布的第二估计({ψ″m}N)，得到所述目标复振幅分布(O)的初始估计，其中，所述目标复振幅分布(O)的初始估计满足：
其中，r为所述目标物所在平面的坐标矢量，n为小于等于N的正整数，OK(r；n)为目标复振幅分布(O)的初始估计，PK(r)为探针复振幅分布的初始估计，∈为预设的常数。
在一些实现方式中，所述第m轮所述目标迭代还包括：当m≥K时，所述K为整数且K≥0，基于所述当前迭代的N帧第一分布的第二估计({ψ″m-1}N)，更新所述目标复振幅分布(O)。
在一些实现方式中，当前迭代更新后所述目标复振幅分布(O)的估计满足：
其中，r为所述目标物所在平面的坐标矢量，n为小于等于N的正整数，Om-1(r；n)为上一轮迭代更新后所述目标复振幅分布的估计，Pm-1(r)为上一轮迭代更新后探针复振幅分布的估计，Pm(r)为当前轮迭代更新后探针复振幅分布的估计，*表示共轭复数，|.|max表示非负约束，α1、α2、α3和α4为预设的范围在0到1之间的常数。
第二方面，本公开提供一种成像装置，包括：至少一个存储介质，存储有至少一组指令集用于实现相干衍射成像；以及至少一个处理器，同所述至少一个存储介质通信连接，其中当所述成像装置运行时，所述至少一个处理器读取所述至少一个指令集并实施第一方面所述的方法。
在一些实现方式中，所述装置还包括：相干光源，被配置为发射一束相干波，所述装置运行时，所述相干波被约束成探针照射到目标物上并发生衍射；探测器，被配置为接收所述衍射产生的出射波并生成N帧衍射图样。
在一些实现方式中，所述相干波可以是电磁辐射、光子、X射线、电子、中子和质子中的至少一种。
由以上技术方案可知，本公开提供的成像方法P100在进行目标迭代的过程中对N帧第一分布({ψ}N)之间的相似分布(ψs)进行估计，进而基于相似分布(ψs)对N帧第一分布的第一估计({ψ′}N)进行更新。由于相似分布(ψs)在每轮迭代中随着N帧第一分布的估计的不断更新而进行调整，成像方法P100可以自适应地在迭代中利用帧间的相似信息。在目标迭代中使用相似分布(ψs)作为对N帧第一分布的第一估计({ψ′}N)进行迭代更新的约束条件，可以提高N帧第一分布的估计({ψ′}N)逐步逼近真实分布({ψ}N)的速度，进而可以提高目标复振幅分布(O)的迭代收敛速度。
本公开提供的相干衍射成像方法及装置的其他功能将在以下说明中部分列出。根据描述，以下数字和示例介绍的内容将对那些本领域的普通技术人员显而易见。本公开提供的相干衍射成像方法及装置的创造性方面可以通过实践或使用下面详细示例中所提供的方法、装置和组合得到充分解释。
 附图说明
为了更清楚地说明本公开实施例中的技术方案，下面将对实施例描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本公开的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。
图1示出了根据本公开的一些实施例提供的成像装置的示意图；
图2示出了根据本公开的一些实施例提供的一种计算设备的硬件结构示意图；
图3示出了根据本公开的一些实施例提供的相干衍射成像方法的流程图；
图4示出了根据本公开的一些实施例提供的施加帧间约束的流程图；
图5示出了根据本公开的一些实施例提供的目标迭代的流程图；
图6示出了根据本公开的一些实施例提供的目标迭代的流程图；以及
[根据细则91更正 09.07.2025]
图7A-图7H示出了根据本公开的一些实施例提供的实验对比结果图。
 具体实施方式
以下描述提供了本公开的特定应用场景和要求，目的是使本领域技术人员能够制造和使用本公开中的内容。对于本领域技术人员来说，对所公开的实施例的各种局部修改是显而易见的，并且在不脱离本公开的精神和范围的情况下，可以将这里定义的一般原理应用于其他实施例和应用。因此，本公开不限于所示的实施例，而是与权利要求一致的最宽范围。
这里使用的术语仅用于描述特定示例实施例的目的，而不是限制性的。比如，除非上下文另有明确说明，这里所使用的，单数形式“一”，“一个”和“该”也可以包括复数形式。当在本公开中使用时，术语“包括”、“包含”和/或“含有”意思是指所关联的整数，步骤、操作、元素和/或组件存在，但不排除一个或多个其他特征、整数、步骤、操作、元素、组件和/或组的存在或在该系统/方法中可以添加其他特征、整数、步骤、操作、元素、组件和/或组。
考虑到以下描述，本公开的这些特征和其他特征、以及结构的相关元件的操作和功能、以及部件的组合和制造的经济性可以得到明显提高。参考附图，所有这些形成本公开的一部分。然而，应该清楚地理解，附图仅用于说明和描述的目的，并不旨在限制本公开的范围。还应理解，附图未按比例绘制。
本公开中使用的流程图示出了根据本公开中的一些实施例的系统实现的操作。应该清楚地理解，流程图的操作可以不按顺序实现。相反，操作可以以反转顺序或同时实现。此外，可以向流程图添加一个或多个其他操作。可以从流程图中移除一个或多个操作。
在本公开中，“X包括A、B或C中的至少一个”所表达的意思是X至少包括A，或X至少包括B，或X至少包括C。也就是说，X可以只包括A、B、C的任意一个，或者同时包括A、B、C的任意组合以及其他可能的内容/元素。所述A、B、C的任意组合可以是A、B、C、AB、AC、BC、或ABC。
在本公开中，除非明确说明，否则结构之间产生的关联关系可以是直接的关联关系也可以是间接的关联关系。比如，当描述“A与B连接”时，除非明确说明了A与B直接连接，否则应当理解成A可以与B直接连接，也可以与B间接地连接；再比如，当描述“A在B之上”时，除非明确说明了A直接在B的上方(AB相邻且A在B的上方)，否则应当理解成A可以直接在B的上方，A也可以间接地在B之上(AB之间隔着其他元素，且A在B的上方)。以此类推。
光学成像技术在科学研究和工业应用上都有着深远的影响。短波长的辐射，如X射线和高能电子，是实现纳米或原子分辨率成像的主要手段。然而，由于成像质量高、数值孔径大的光学透镜的制造瓶颈以及合适的制作材料的缺乏，传统的光学显微镜已经难以满足新兴的材料科学以及生物学领域的成像表征需求。
相干衍射成像技术为那些难以制造高质量透镜的波段提供了一种强有力的成像手段。相干衍射成像技术通过相干光源与待测样品相互作用形成衍射场，并通过迭代计算的方式重构出样品的幅值与相位信息。由于其不需要高性能的成像光学元件，因而可以被广泛应用到深极紫外、X射线以及电子束等相干辐射源上。
相干波透过待测样品形成的出射波以衍射图样的形式被探测器接收。衍射图样只能记录到振幅信息而丢失了相位信息。因此，需要通过相位恢复得到样品的结构信息。在进行相位恢复的过程中，可以使用迭代更新的方法逐步逼近待测样品真实的相位分布。由于相位恢复问题的解的非唯一性，在迭代中需要通过施加一定的约束以进行相位更新。通常可以采用交替应用支撑约束以及幅值约束的方式使相位估计逐步收敛。
相关技术中使用增加样品静止区域的先验知识或者增加样品边界信息的方式提高相位恢复的收敛速度。本公开提供一种新的相干衍射成像方法，在迭代中添加自适应的约束条件，以充分利用变化样品在多帧采样中的相似性信息，提高迭代收敛速度的同时获得更好的相位恢复结果。
图1示出了根据本公开的一些实施例提供的成像装置001的示意图。成像装置001包括相干光源100、探测器200、存储介质以及处理器。
相干光源100用于发射一束照射目标物002的相干波110。所述目标物002是待测样品。在相干衍射成像中，待测样品的尺寸与相干波110的波长相匹配，以确保能够产生足够强、可用于解析的衍射信号，同时满足所需的成像分辨率。所述目标物002可以是随时间变化的样品。比如，目标物002可以是生长的晶体、移动的细胞、微生物、病毒等。成像装置001可以对随时间变化的目标物002进行相干衍射成像，以获得目标物002在多个时刻的结构信息。
为了适应不同尺度、不同性质的目标物002或者不同的应用需求，成像装置001可以使用不同类型的相干波110。本公开提供的成像装置001可以适用于多种应用场合，具体使用何种相干光源100以生成所需的相干波110取决于目标物002的性质、所需分辨率以及研究目标等因素。
在一些实施例中，所述相干波110可以是电磁辐射。比如，可见光、近红外、紫外、真空紫外乃至X射线等高频部分。电磁辐射作为相干波110可以广泛应用于材料科学、生物医学、半导体器件、纳米结构研究等领域。例如，微纳结构表面形貌分析、解析晶体结构、蛋白质分子结构等。在一些实施例中，所述相干波110可以是光子。光子是电磁辐射的量子单元。光子相干衍射成像可以应用在量子信息处理和量子成像等新兴领域。例如，量子光学、量子信息科学、量子计算中的量子态制备与测量，以及探索超越经典极限的成像技术。在一些实施例中，所述相干波110可以是X射线。X射线具有短波长、高穿透力的特点，能深入目标物内部。X射线相干性可以通过同步辐射光源、自由电子激光器等高级光源得到保证。X射线作为相干波110可以应用于物质结构分析(如晶体结构、非晶态材料、复杂大分子)、纳米尺度成像(如纳米颗粒、生物大分子、细胞内部结构)、材料科学(缺陷检测、应力分析)、文化遗产保护(无损检测)等。在一些实施例中，所述相干波110可以是电子。电子的波动性表现需要在较高能量下才能显著，形成电子波。电子衍射主要发生在电子显微镜中，通过电子枪发射的高速电子束与样品相互作用产生相干衍射图样。电子波作为相干波110可以应用于电子显微术(包括透射电子显微镜TEM和扫描电子显微镜SEM)，可以直接观察样品的微观结构、化学成分、晶体学信息等。在一些实施例中，所述相干波110可以是中子。中子因其没有电荷且与原子核有特定相互作用，可以穿透许多对电磁辐射不透明的材料，且对某些元素和同位素有选择性敏感性。中子源(如反应堆或散裂中子源)产生的中子束可以实现相干衍射。中字束作为相干波110可以用于研究材料的磁结构、氢/氘位置、应力分布、复杂合金成分分布等，在凝聚态物理、地球科学、能源材料、生物医学等领域有重要应用。在一些实施例中，所述相干波110可以是质子。质子作为带正电的重粒子，其波动性在极高能量下显著。质子相干衍射可以在极端条件下的物质结构研究、核物理学、宇宙射线探测等领域发挥作用。
需要说明的是，上述有关不同类型的相干波110可以应用的场景为对本公开提供的成像装置001可以应用的部分场景的示例性说明。其它可以使用本公开提供的成像装置001的应用场景都在本公开保护的范围内。
成像装置001运行时，所述相干波110被约束成探针120照射到目标物002上并发生衍射。为了便于描述，可以约定目标物002所在的与探针120轴向垂直的平面为目标平面A。目标平面A为探针120透过目标物002并发生衍射现象的平面。探针120穿透目标物002之后形成出射波，经过一段距离的传播之后到达探测器200。
探针120可以是被约束为一定的尺寸或形状(如聚焦成束状或窄脉冲)的相干波。一方面，成像装置001将相干波110约束为探针120可以获得较高的能量密度，有助于获得足够强的衍射信号，以进行有效成像。另一方面，成像装置001将相干波110约束为探针120可以使得入射到目标物002上的波前具有高度的空间相干性，从而在衍射过程中产生清晰、可解析的高分辨率衍射图样。成像装置001可以通过使用更小尺寸的探针120以提供更高的空间分辨率。
继续参照图1，探测器200用于接收目标物002上发生衍射产生的出射波并生成N帧衍射图样。为了便于描述，可以约定探测器200用于接收出射波的平面为探测平面B。探针120透过目标物002之后便会携带目标物002的微观结构信息，即衍射发生之后，出射波会携带着目标物002的信息到达探测平面B。探测器200可以记录出射波传播到探测平面B时在探测平面B上不同位置的强度信息，并形成衍射图样。衍射图样不同位置对应于不同的频率。探测器200与目标物002之间的距离满足远场衍射条件。衍射图样中各点的强度由傅里叶变换关系决定，也就是说，出射波在目标平面A上的复振幅分布(第一分布)经傅里叶变换后，在远场的探测平面B形成了对应的复振幅分布(第二分布)。成像装置001能够以特定的采样频率对目标物002进行衍射图样的采集。
对于变化的目标物002，探测器200可以获取多个采样时刻的多帧衍射图样。采集衍射图样的多个采样时刻可以是等间隔的，也可以是非等间隔的。目标物002在对应的多个衍射时刻(即与探测平面B衍射图样的采样时刻对应的目标平面A上衍射的发生时刻)具有空间上的局部一致性。具体来说，比如，目标物002是不断生长的晶体。在相邻的衍射时刻，目标物002有部分区域发生了变化，而有部分区域维持了相对静止。再比如，目标物002是在目标平面A内移动的物体。在相邻的衍射时刻，目标物002可以发生整体的位移，但目标物002所在的部分区域内部保持相对静止。因此，从目标平面A来看，目标物002在多个衍射时刻表现出空间上的相似性。反映在探测平面B，多帧衍射图样之间也存在帧间相似性。
如前所述，本公开提供的成像装置001可以适应不同领域的应用场景。探测器200的选择需要考虑所用相干波110的类型，还要考虑成像需求(如空间分辨率、能量分辨率、动态范围、时间分辨率等)、实验条件(如样品类型、实验环境、数据采集速度等)以及经济因素等。
在一些实施例中，探测器200可以是CCD(Charge-Coupled Device，电荷耦合器件)探测器。CCD探测器能够提供较高的量子效率和较低的读出噪声，适合于捕获微弱的衍射信号。在一些实施例中，探测器200可以是CMOS(Complementary Metal-Oxide-Semiconductor，互补金属氧化物半导体)图像传感器。CMOS图像传感器具有较低的成本和较高的读出速度。在一些实施例中，探测器200可以是光电二极管阵列探测器。在一些实施例中，探测器200可以是光子计数探测器。比如，探测器200可以是硅光电倍增管(Silicon photomultiplier，SiPM)、单光子雪崩二极管(Single Photon Avalanche Diode，SPAD)探测器。在一些实施例中，探测器200可以是X射线探测器。比如，探测器200可以是闪烁体耦合探测器、硅基直接电子探测器(Silicon based direct electron detectors，DEDS)或者混合像素探测器(Hybrid Pixel Detectors，HEPDs)。在一些实施例中，探测器200可以是电子探测器。比如，探测器200可以是闪烁探测器、直接电子计数探测器或者荧光屏-CCD系统探测器。在一些实施例中，探测器200可以是中子探测器。比如，闪烁体探测器，中子与闪烁体材料相互作用产生可见光，再由光电转换器件记录。比如，3He管探测器。再比如，固态中子探测器，可以直接将中子俘获反应产生的带电粒子转换为电信号。
需要说明的是，上述有关不同类型的探测器200可以应用的场景为对本公开提供的成像装置001可以应用的部分场景的示例性说明。其它可以使用本公开提供的成像装置001的应用场景都在本公开保护的范围内。
继续参照图1，成像装置001包括至少一个存储介质以及至少一个处理器。所述至少一个处理器同所述至少一个存储介质通信连接。所述至少一个存储介质存储有至少一组指令集用于实现本公开提供的相干衍射成像方法。所述至少一个处理器可以与探测器200通信连接，以将采集到的N帧衍射图样应用于本公开提供的相干衍射成像方法的实施中。当所述成像装置001运行时，所述至少一个处理器读取所述至少一个指令集并实施本公开提供的相干衍射成像方法。本公开提供的相干衍射方法将在后文进行详细描述。在一些实施例中，所述成像装置001可以将所述至少一个存储介质以及所述至少一个处理器集成于计算设备300中。成像装置001中可以包括一个或者多个计算设备300。
图2示出了根据本公开的一些实施例提供的一种计算设备300的硬件结构示意图。如图2所示，计算设备300包括至少一个存储介质330和至少一个处理器320。在一些实施例中，计算设备300还可以包括内部通信总线310。在一些实施例中，计算设备300还可以包括通信端口350。在一些实施例中，计算设备300还可以包括I/O组件360。
内部通信总线310可以连接不同的系统组件，包括存储介质330和处理器320。I/O组件360支持计算设备300和其他组件之间的输入/输出。
存储介质330可以包括数据存储装置。数据存储装置可以是非暂时性存储介质，也可以是暂时性存储介质。比如，数据存储装置可以包括磁盘332、只读存储介质(ROM)334或随机存取存储介质(RAM)336中的一种或多种。存储介质330还包括存储在数据存储装置中的至少一个指令集。指令集是计算机程序代码，计算机程序代码可以包括执行本公开提供相关衍射成像方法的程序、例程、对象、组件、数据结构、过程、模块等等。
通信端口350用于计算设备300同外界的数据通讯。比如，计算设备300可以通过通信端口350连接网络330。至少一个处理器320同至少一个存储介质330通过内部通信总线310通信连接。至少一个处理器320用以执行上述至少一个指令集。相关衍射成像系统001运行时，至少一个处理器320读取至少一个指令集，并且根据至少一个指令集的指示执行本公开提供的相关衍射成像方法。
处理器320可以执行相关衍射成像方法包含的所有步骤。处理器320可以是一个或多个处理器的形式。处理器320可以发出执行指令。处理器320可以包括一个或多个硬件处理器，例如微控制器，微处理器，精简指令集计算机(RISC)，专用集成电路(ASIC)，特定于应用的指令集处理器(ASIP)，中央处理单元(CPU)，图形处理单元(GPU)，物理处理单元(PPU)，微控制器单元，数字信号处理器(DSP)，现场可编程门阵列(FPGA)，高级RISC机器(ARM)，可编程逻辑器件(PLD)，能够执行一个或多个功能的任何电路或处理器等，或其任何组合。
仅仅为了说明问题，在本公开中计算设备300中仅描述了一个处理器320。然而，应当注意，本公开中计算设备300还可以包括多个处理器，因此，本公开中披露的操作和/或方法步骤可以如本公开所述的由一个处理器执行，也可以由多个处理器联合执行。例如，如果在本公开中计算设备300的处理器320执行步骤A和步骤B，则应该理解，步骤A和步骤B也可以由两个不同处理器320联合或分开执行(例如，第一处理器执行步骤A，第二处理器执行步骤B，或者第一和第二处理器共同执行步骤A和B)。
综上所述，本公开提供的成像装置001可以通过相干光源100发射相干波110，并对相干波110进行约束以形成精准的探针120。探针120照射目标物002并发生衍射后形成出射波。多个时刻的出射波经过传播后到达远场的探测平面B，并被探测器200接收形成多帧衍射图样。计算设备300基于接收到的多帧衍射图样恢复目标物002的复振幅分布。成像装置001可以应用于多种波段的相干波衍射成像，可以适应多种类型的探测器200，结构简单、适应性广，具有广阔的应用场景。
接下来对本公开提供的相干衍射成像方法(以下简称成像方法)进行详细描述。
图3示出了根据本公开的一些实施例提供的成像方法P100的流程图。如前所述，成像装置001运行时，处理器可以读取存储在其本地存储介质和/或数据库中的指令集，然后根据指令集的规定，执行成像方法P100。
本公开提供的成像方法P100包括：
S110：获得目标物002的N帧衍射图样({I}N)。
如前所述，相干光源100产生的相干波110经过所述目标物002时发生衍射形成出射波。在N个发生时刻，即N个衍射发生的时刻，所述出射波在所述目标物002处分别形成N帧第一分布({ψ}N)。其中，所述N为正整数，符号{·}N表示N帧的集合。所述N帧第一分布({ψ}N)为出射波在目标平面A处真实的空间分布。ψ(r；n)表示所述N帧第一分布中的第n帧在目标平面A处的分布，其中r为目标平面A上的坐标矢量。ψ(r；n)在下文中简写为ψ。所述N帧第一分布({ψ}N)为出射波的复振幅分布，包括振幅和相位信息。所述出射波沿出射方向传播至探测器200时所述N帧第一分布({ψ}N)分别变成对应的N帧第二分布({Ψ}N)。N帧第二分布({Ψ}N)为出射波传播到探测平面B时的复振幅分布。Ψ(k；n)表示所述N帧第二分布中的第n帧在探测平面B处的分布，其中k为探测平面B上的坐标矢量。Ψ(k；n)在下文中简写为Ψ。N帧第一分布({ψ}N)与其对应的N帧第二分布({Ψ}N)满足傅里叶变换的关系。
所述N帧衍射图样({I}N)为所述探测器200得到的所述N帧第二分布({Ψ}N)的强度信息。探测器200可以记录按照时间顺序采集到的所述N帧第二分布({Ψ}N)在探测平面B上不同位置的强度信息，并形成N帧衍射图样。所述强度信息与N帧第二分布({Ψ}N)的振幅成正相关。衍射图样({I}N)上不同位置对应于不同的频率。
S130：基于所述N帧第一分布({ψ}N)的帧间约束以及所述N帧衍射图样({I}N)，通过执行目标迭代得到所述目标物的目标复振幅分布(O)，直至所述目标复振幅分布(O)满足预设的收敛条件。
目标复振幅分布(O)为目标物002在目标平面A上的复振幅分布的估计，包括振幅和相位信息。目标复振幅分布(O)可以从出射波的N帧第一分布({ψ}N)中被分离出来。成像方法P100通过执行多次目标迭代以不断更新对于N帧第一分布({ψ}N)的估计，进而得到目标复振幅分布(O)。
目标物002在N个衍射发生时刻具有空间上的局部一致性。从目标平面A来看，目标物002在N个衍射发生时刻表现出空间上的相似性，即N帧第一分布({ψ}N)之间存在着相似分布(ψs)。反映在探测平面B，N帧衍射图样({I}N)之间也存在帧间相似性。成像方法P100将相似性作为N帧第一分布({ψ}N)之间的帧间约束应用于对N帧第一分布({ψ}N)的估计过程中。
其中，所述帧间约束包括在每次目标迭代中通过对所述N帧第一分布({ψ}N)之间的相似分布(ψs)进行更新，从而得到所述N帧第一分布的第一估计({ψ′}N)。所述N帧第一分布的第一估计({ψ′}N)为在每次目标迭代中基于相似分布(ψs)得到的对于N帧第一分布({ψ}N)的估计。成像方法P100对于N帧第一分布({ψ}N)的估计方式包括一种或者多种，使用相似分布(ψs)对其进行估计更新的方式属于其中一种。
S150：输出所述目标复振幅分布(O)。
目标复振幅分布(O)为目标物002的振幅和相位估计。对于目标物002结构的呈现形式可以是根据目标复振幅分布(O)恢复得到图像，也可以是将振幅信息和相位信息分别进行可视化呈现。
本公开提供的成像方法P100在进行目标迭代的过程中对N帧第一分布({ψ}N)之间的相似分布(ψs)进行估计，进而基于相似分布(ψs)对N帧第一分布的第一估计({ψ′}N)进行更新。由于相似分布(ψs)在每轮迭代中随着N帧第一分布的估计的不断更新而进行调整，成像方法P100可以自适应地在迭代中利用帧间的相似信息。在目标迭代中使用相似分布(ψs)作为对N帧第一分布的第一估计({ψ′}N)进行迭代更新的约束条件，可以提高N帧第一分布的估计({ψ′}N)逐步逼近真实分布({ψ}N)的速度，进而可以提高目标复振幅分布(O)的迭代收敛速度。
前述步骤S130中，成像方法P100在每轮目标迭代中通过对相似分布(ψs)进行更新，进而得到所述N帧第一分布的第一估计({ψ′}N)。在每一轮目标迭代中相似分布(ψs)随着对于N帧第一分布({ψ′}N)的估计的不断变化也在不断更新。
图4示出了根据本公开的一些实施例提供的施加帧间约束的流程图。对于第m轮迭代，m为正整数，所述通过对所述N帧第一分布({ψ}N)之间的相似分布(ψs,m)进行更新以便得到所述N帧第一分布的第一估计({ψ′m}N)可以包括：
S135-1：确定当前迭代的所述N帧第一分布的初始估计({ψ0m}N)的相似性(Sm)。
N帧第一分布的初始估计({ψ0m}N)为在当前轮次的迭代开始时N帧第一分布({ψ}N)的估计。对于第一轮迭代来说，需要对N帧第一分布的初始估计({ψ0m}N)进行初始化操作。比如，成像装置001可以对N帧第一分布的初始估计({ψ0m}N)进行随机初始化或者零相位初始化。
在一些实施例中，所述确定当前迭代的所述N帧第一分布的初始估计({ψ0m}N)的相似性(Sm)可以包括：计算当前迭代的所述N帧第一分布的初始估计({ψ0m}N)的离散指标，以及基于所述离散指标得到所述相似性(Sm)。所述离散指标可以用于反映N帧第一分布({ψ}N)之间的离散程度，或者说偏离均值的程度。比如，所述离散指标可以是标准差或者是方差。可以认为，所述离散指标越大，N帧第一分布的初始估计({ψ0m}N)之间的相似性(Sm)越低。
S135-3：基于所述相似性(Sm)，确定所述当前迭代中所述N帧第一分布的初始估计({ψ0m}N)之间的所述相似分布(ψs,m)和所述差异分布({ψe,m}N)。
相似分布(ψs,m)为对N帧第一分布({ψ}N)中相似部分的估计。对于N帧第一分布({ψ}N)来说，每一帧都可以由相似分布(ψs)与差异分布(ψe)组成。在当前迭代中，可以基于相似性(Sm)对相似分布(ψs,m)和N帧差异分布({ψe,m}N)进行估值更新。
S135-5：基于所述相似分布(ψs,m)和所述差异分布({ψe,m}N)，确定所述当前迭代的所述N帧第一分布的第一估计({ψ′m}N)。
如前所述，N帧第一分布的第一估计({ψ′}N)为在每次目标迭代中基于相似分布(ψs)得到的对于N帧第一分布({ψ}N)的估计。成像方法P100对于N帧第一分布({ψ}N)的估计方式包括一种或者多种，使用相似分布(ψs)对其进行估计更新的方式属于其中一种。换句话说，在第m轮迭代中，N帧第一分布的第一估计({ψ′m}N)为对N帧第一分布的初始估计(ψ0m)施加帧间约束后的结果。
N帧第一分布的第一估计({ψ′m}N)中的每一帧(ψ′m)可以由相似分布(ψs,m)与差异分布(ψe,m)组成。比如，N帧第一分布的第一估计({ψ′m}N)可以表示为相似分布(ψs,m)与N帧差异分布({ψe,m}N)之和或者加权和。
在一些实施例中，还可以基于N帧第一分布在变换域上的分布得到相似分布(ψs)。比如，所述变换域可以是SVD(Singular Value Decomposition，奇异值分解)变换域。对于N帧第一分布({ψ}N)，将每帧ψ的大小定义为x·y，则N帧第一分布({ψ}N)的维度为x·y·N。将N帧第一分布({ψ}N)重新排布成大小为N·xy的二维矩阵χ。其中，χ的第一行对应N帧第一分布({ψ}N)中的第一帧，其它行以此类推。对二维矩阵χ做SVD变换，分解得到U，S，V三个矩阵。其中，U为左奇异向量矩阵，S是奇异值矩阵，V是右奇异向量矩阵。S是对角矩阵，对于连续变化的出射波，也即连接变化的N帧第一分布，S矩阵中的第一项会有较大的值。V的第一行V(1,:)为归一化的主要成分，对应的第一个奇异值S(1,1)值是最大的，代表V(1,:)在二维矩阵χ所有行中占的能量成分最大。因此取其大小为1·xy，把其重新排布成x·y的矩阵，即可得到相似分布(ψs)。
本公开提供的成像方法P100能够根据N帧第一分布({ψ}N)之间或者N帧第一分布在变换域上的对应N帧分布之间的相似性(S)对N帧第一分布({ψ}N)中的相似分布(ψs)进行表征，进而基于相似分布(ψs)得到N帧第一分布的第一估计({ψ′}N)。在每一轮迭代中(以第m轮为例)，所述相似性(Sm)、相似分布(ψs,m)、N帧差异分布({ψe,m}N)以及N帧第一分布的第一估计({ψ′m}N)都得以更新。成像方法P100充分利用了N帧第一分布({ψ}N)之间的相似性信息，对于N帧第一分布({ψ}N)的估计可以以较快速度收敛。
在一些实施例中，第m轮迭代中所述N帧第一分布的第一估计({ψ′m}N)可以满足如下关系：ψ′m(r；n)＝ψe,m(r；n)+ρψs,m(r)ψe,m(r；n)＝[1-Sm(r)]ψ0m(r；n)Sm(r)＝Cm(σm(r))
其中，N帧第一分布的第一估计({ψ′m}N)中的每一帧可以由相似分布(ψs,m)与差异分布(ψe,m)的加权求和得到。差异分布(ψe,m)的系数ρ用于保持所述当前迭代中能量的恒定。在一些实施例中，加权系数ρ可以为其中求和符号作用于N帧第一分布。是所述N帧第一分布的初始估计({ψ0m}N)的平均值。σm可以是所述N帧第一分布的初始估计({ψ0m}N)的标准差分布。
Cm为转换函数，用于将所述σm转换成所述相似性Sm。α、βm以及κm用于调整所述转换函数的形状。转换函数Cm具有反S形，函数值的大小与σm成负相关。α为预设的参数值，用于使得转换函数Cm在σm接近零时达到预设的α。随着迭代的进行，βm和κm的值会不断更新。为了在迭代过程中具有更好的鲁棒性，在一些实施例中，βm的计算方式可以为βm＝0.9·s1+0.2·s2。其中，s1是第m轮迭代中σm中的最低的10％的平均值，s2是σm中最高10％的平均值。成像方法P100使用最高10％和最低10％的平均值，而不是最高值和最低值，可以达到更稳定的收敛效果，使得迭代过程具有更好的鲁棒性。所述κm与σm成正相关。在一些实施例中，κm可以设置为σm的倍数，比如，可以将κm可以设置为σm标准差的50倍。
如前所述，本公开提供的成像方法P100对于N帧第一分布({ψ}N)的估计方式包括一种或者多种，使用相似分布(ψs)对其进行估计更新的方式属于其中一种。换句话说，在第m轮迭代中，N帧第一分布的第一估计({ψ′m}N)为对N帧第一分布的初始估计(ψ0m)施加帧间约束后的结果。除了施加帧间约束这种估计方式外，成像方法P100还可以对N帧第一分布施加其它约束方式以得到N帧第一分布的初始估计({ψ0m}N)以及N帧第一分布的第二估计({ψ″m}N)。
图5示出了根据本公开的一些实施例提供的目标迭代的流程图。对于第m轮迭代，所述目标迭代可以包括：
S131：得到上一轮迭代的所述N帧第一分布的第二估计({ψ″m-1}N)。
所述N帧第一分布的第二估计({ψ″m-1}N)为经过上一轮迭代之后得到的对于N帧第一分布({ψ}N)的估计结果。
S133：基于所述上一轮迭代的所述N帧第一分布的第二估计({ψ″m-1}N)获得当前迭代的N帧第一分布的初始估计({ψ0m}N)。
在一些实施例中，成像方法P100对所述N帧第一分布的第二估计({ψ″m-1}N)施加支撑约束以将所述N帧第一分布({ψ}N)的估计限定在约束范围之内。所述基于所述上一轮迭代的所述N帧第一分布的第二估计({ψ″m-1}N)获得当前迭代的所述N帧第一分布的初始估计({ψ0m}N)包括：当所述N帧第一分布的第二估计({ψ″m-1}N)位于支撑区域内时，以所述N帧第一分布的第二估计({ψ″m-1}N)作为N帧第一分布的初始估计({ψ0m}N)；或者当所述N帧第一分布的第二估计({ψ″m-1}N)位于支撑区域外时，所述N帧第一分布的第二估计({ψ″m-1}N)在该区域的值为零。
S135：对所述N帧第一分布的初始估计({ψ0m}N)施加所述帧间约束，得到当前迭代的所述N帧第一分布的第一估计({ψ′m}N)。
对所述N帧第一分布的初始估计({ψ0m}N)施加所述帧间约束的方法如前文S135-1至S135-5所述，此处不再赘述。
S137：基于所述当前迭代的所述N帧第一分布的第一估计({ψ′m}N)、所述上一轮迭代的所述N帧第一分布的第二估计({ψ″m-1}N)、以及所述N帧衍射图样({I}N)，得到当前迭代的所述N帧第一分布的第二估计({ψ″m}N)。
在一些实施例中，成像方法P100通过将当前迭代的所述N帧第一分布的第一估计({ψ′m}N)以及前轮迭代的所述N帧第一分布的第二估计({ψ″m-1}N)传播到探测平面B，并施加模量约束的方式以保证传播过程的能量守恒。具体来说，成像方法P100可以基于所述N帧衍射图样({I}N)，得到所述当前迭代的N帧第二分布估计({Ψ′m}N)。所述当前迭代的所述N帧第二分布估计({Ψ′m}N)为所述N帧衍射图样({I}N)所对应的所述出射波在所述探测器处的复振幅分布的估计值。由于目标平面A处的N帧第一分布({ψ}N)与探测平面B处的N帧第二分布({Ψ}N)满足傅里叶变换关系。因此，可以依据傅里叶变换得到当前迭代中N帧第二分布在探测平面B的初始估计Ψ0m。在一些实施例中，其中，表示傅里叶变换。该方法通过增加δ这一增量，可以有效避免迭代陷入停滞。在一些实施例中，δ＝ψ′m(r；n)-ψ″m-1(r；n)。当然，ψ′m以及ψ″m-1的系数可以根据实际需要进行调整。对N帧第二分布的初始估计施加模量约束以得到N帧第二分布估计({Ψ′m}N)。对N帧第二分布的初始估计中的每一帧施加模量约束的方式可以为：其中，I对应N帧衍射图样({I}N)中的每一帧。
在一些实施例中，成像方法P100通过将探测平面B上的所述N帧第二分布估计({Ψ′m}N)反向传播到目标平面A，以得到所述N帧第一分布的第二估计({ψ″m}N)。具体来说，成像方法P100可以基于当前迭代的所述N帧第一分布的第一估计({ψ′m}N)和上一轮迭代的所述N帧第一分布的第二估计({ψ″m-1}N)，确定所述当前迭代的N帧第二分布估计({Ψ′m}N)反向传播到目标物处所对应的当前迭代的所述N帧第一分布的第二估计({ψ″m}N)。在一些实施例中，所述当前迭代的所述N帧第一分布的第二估计({ψ″m}N)中的第n帧满足：其中表示逆傅里叶变换。所述δ与所述N帧第一分布的第一估计({ψ′m}N)以及所述上一轮迭代的所述N帧第一分布的第二估计({ψ″m-1}N)成线性相关。在一些实施例中，δ＝ψ′m(r；n)-ψ″m-1(r；n)。当然，ψ′m以及ψ″m-1的系数可以根据实际需要进行调整。该方法通过增加δ这一增量，可以有效避免迭代陷入停滞。
综上所述，本公开提供的成像方法P100通过将N帧第一分布({ψ}N)在目标平面A与探测平面B之间来回传播以进行N帧第一分布({ψ}N)的估计值的更新。其中，在每轮迭代中(以第m轮为例)，对所述N帧第一分布的初始估计({ψ0m}N)施加所述帧间约束，得到当前迭代的所述N帧第一分布的第一估计({ψ′m}N)，可以充分利用N帧第一分布的初始估计({ψ0m}N)之间的相似性信息，有利于提高收敛速度。
前文详细描述了如何得到N帧第一分布({ψ}N)的估计。接下来，对如何得到目标复振幅分布(O)进行详细描述。图6示出了根据本公开的一些实施例提供的目标迭代的流程图。
在一些实施例中，所述第m轮所述目标迭代还包括：当m＝K时，所述K为整数且K≥0，基于所述当前迭代的N帧第一分布的第二估计({ψ″m}N)，得到所述目标复振幅分布(O)的初始估计。其中，所述K可以是预设的迭代轮次数目，当迭代完成K轮之后，进行目标复振幅分布(O)的初始化。其中，所述目标复振幅分布(O)的初始估计满足：
其中，OK(r；n)为目标复振幅分布(O)的第n帧的初始估计，PK(r)为探针复振幅分布的初始估计，∈为预设的常数，用于防止分母出现零值。
当m≥K时，基于所述当前迭代的N帧第一分布的第二估计({ψm}N)，更新所述目标复振幅分布(O)。其中，当m＝K时，成像方法P100可以按照上述初始化方法得到目标复振幅分布(O)以及探针复振幅分布(P)的初始估计。得到探针复振幅分布(P)的初始估计后，可以对其施加支撑约束与模量约束，以进行更新。
当m＞K时，当前迭代更新后所述目标复振幅分布(O)的第n帧的估计满足：
其中，Om-1(r；n)为上一轮迭代更新后所述目标复振幅分布的估计。Pm-1(r)为上一轮迭代更新后探针复振幅分布的估计。Pm(r)为当前轮迭代更新后探针复振幅分布的估计。*表示共轭复数。|.|max表示非负约束。α1、α2、α3和α4为预设的范围在0到1之间的常数。
当所述目标复振幅分布(O)的估计满足预设的收敛条件时，停止迭代。
[根据细则91更正 09.07.2025]
图7A-图7H示出了根据本公开的一些实施例提供的实验对比结果图。图7A-图7H中示出的是聚苯乙烯磺酸盐(3,4-亚乙二氧基噻吩，PEDOT：PSS)晶体样品的生长过程的数值模拟成像结果。由图7A可以看出，本公开提供的成像方法P100的收敛速度相较于传统的CDI方法明显更快，在几百次迭代内就达到了收敛。图7B-图7D分别显示了t1＝0s、t2＝2.5s和t3＝5s时所选帧使用成像方法P100重构的振幅和相位。传统CDI在t3＝5s时的结果见图7F。与传统CDI方法相比，成像方法P100的重构结果质量明显较优。传统CDI的重构结果显示出明显的伪影，妨碍了晶体内部信息的清晰度。图7G和图7H分别显示了相似分布(ψs)和相似性S。从图7B-图7D可以看出，样本变化较大的区域相似性S较低，如图7H中黑色区域所示。同样，在相应位置获得的相似性部分也很小。模拟实验表明，成像方法P100可以有效地提取N帧第一分布之间的相似部分，并通过帧间约束加速算法收敛。
综上所述，本公开提供的成像方法P100在进行目标迭代的过程中对N帧第一分布({ψ}N)之间的相似分布(ψs)进行估计，进而基于相似分布(ψs)对N帧第一分布的第一估计({ψ′}N)进行更新。由于相似分布(ψs)在每轮迭代中随着N帧第一分布的估计的不断更新而进行调整，成像方法P100可以自适应地在迭代中利用帧间的相似信息。在目标迭代中使用相似分布(ψs)作为对N帧第一分布的第一估计({ψ′}N)进行迭代更新的约束条件，可以提高N帧第一分布的估计({ψ′}N)逐步逼近真实分布({ψ}N)的速度，进而可以提高目标复振幅分布(O)的迭代收敛速度。
上述对本公开特定实施例进行了描述。其他实施例在所附权利要求书的范围内。在一些情况下，在权利要求书中记载的动作或步骤可以按照不同于实施例中的顺序来执行并且仍然可以实现期望的结果。另外，在附图中描绘的过程不一定要求示出特定顺序或者连续顺序才能实现期望的结果。在某些实施方式中，多任务处理和并行处理也是可以的或者是可能有利的。
综上所述，在阅读本详细公开内容之后，本领域技术人员可以明白，前述详细公开内容可以仅以示例的方式呈现，并且可以不是限制性的。尽管这里没有明确说明，本领域技术人员可以理解本公开需求囊括对实施例的各种合理改变，改进和修改。这些改变，改进和修改旨在由本公开提出，并且在本公开的示例性实施例的精神和范围内。
此外，本公开中的某些术语已被用于描述本公开的实施例。例如，“一个实施例”，“实施例”和/或“一些实施例”意味着结合该实施例描述的特定特征，结构或特性可以包括在本公开的至少一个实施例中。因此，可以强调并且应当理解，在本公开的各个部分中对“实施例”或“一个实施例”或“替代实施例”的两个或更多个引用不一定都指代相同的实施例。此外，特定特征，结构或特性可以在本公开的一个或多个实施例中适当地组合。
应当理解，在本公开的实施例的前述描述中，为了帮助理解一个特征，出于简化本公开的目的，本公开将各种特征组合在单个实施例、附图或其描述中。然而，这并不是说这些特征的组合是必须的，本领域技术人员在阅读本公开的时候完全有可能将其中一部分设备标注出来作为单独的实施例来理解。也就是说，本公开中的实施例也可以理解为多个次级实施例的整合。而每个次级实施例的内容在于少于单个前述公开实施例的所有特征的时候也是成立的。
本文引用的每个专利，专利申请，专利申请的出版物和其他材料，例如文章，书籍，说明书，出版物，文件，物品等，除了与其相关的任何历史起诉文件、可能与本文件不一致或相冲突的任何相同的、或者任何可能对权利要求的最宽范围具有限制性影响的任何相同的历史起诉文件，均可以通过引用结合于此，并用于现在或以后与本文件相关联的所有目的。此外，如果在与任何所包含的材料相关联的术语的描述、定义和/或使用与本文档相关的术语、描述、定义和/或之间存在任何不一致或冲突时，使用本文件中的术语为准。
最后，应理解，本文公开的申请的实施方案是对本公开的实施方案的原理的说明。其他修改后的实施例也在本公开的范围内。因此，本公开披露的实施例仅仅作为示例而非限制。本领域技术人员可以根据本公开中的实施例采取替代配置来实现本公开中的申请。因此，本公开的实施例不限于申请中被精确地描述过的实施例。

Claims (11)

  1. 一种相干衍射成像方法，其特征在于，包括：
    获得目标物的N帧衍射图样({I}N)，其中，相干波经过所述目标物时发生衍射形成出射波，在N个发生时刻，所述出射波在所述目标物处分别形成N帧第一分布({ψ}N)，所述出射波沿出射方向传播至探测器时所述N帧第一分布({ψ}N)分别变成N帧第二分布({Ψ}N)，所述N帧衍射图样({I}N)为所述探测器得到的所述N帧第二分布({Ψ}N)的强度信息，所述N为正整数；
    基于所述N帧第一分布({ψ}N)的帧间约束以及所述N帧衍射图样({I}N)，通过执行目标迭代得到所述目标物的目标复振幅分布(O)，直至所述目标复振幅分布(O)满足预设的收敛条件，其中，所述帧间约束包括在每次目标迭代中通过对所述N帧第一分布({ψ}N)之间或所述N帧第一分布在变换域中对应的N帧分布之间的相似分布(ψs)进行更新，从而得到所述N帧第一分布的第一估计({ψ′}N)；以及
    输出所述目标复振幅分布(O)，其中，
    对于第m轮迭代，m为正整数：所述通过对所述N帧第一分布({ψ}N)之间的相似分布(ψs,m)进行更新以便得到所述N帧第一分布的第一估计({ψ′m}N)包括：确定当前迭代的所述N帧第一分布的初始估计({ψ0m}N)的相似性(Sm)；基于所述相似性(Sm)，确定所述当前迭代中所述N帧第一分布的初始估计({ψ0m}N)之间的所述相似分布(ψs,m)和差异分布({ψe,m}N)；以及基于所述相似分布(ψs,m)和所述差异分布({ψe,m}N)，确定所述当前迭代的所述N帧第一分布的第一估计({ψ′m}N)；
    对于第m轮迭代，所述目标迭代包括：得到上一轮迭代的所述N帧第一分布的第二估计({ψ″m-1}N)；基于所述上一轮迭代的所述N帧第一分布的第二估计({ψ″m-1}N)获得当前迭代的N帧第一分布的初始估计({ψ0m}N)；对所述N帧第一分布的初始估计({ψ0m}N)施加所述帧间约束，得到当前迭代的所述N帧第一分布的第一估计({ψ′m}N)；以及基于所述当前迭代的所述N帧第一分布的第一估计({ψ′m}N)、所述上一轮迭代的所述N帧第一分布的第二估计({ψ″m-1}N)、以及所述N帧衍射图样({I}N)，得到当前迭代的所述N帧第一分布的第二估计({ψ″m}N)。
  2. 如权利要求1所述的方法，其特征在于，所述当前迭代的所述N帧第一分布的第一估计({ψ′m}N)满足：
    ψ′m(r；n)＝ψe,m(r；n)+ρψs,m(r)

    ψe,m(r；n)＝[1-Sm(r)]ψ0m(r；n)
    Sm(r)＝Cm(σm(r))
    其中，r为所述目标物所在平面的坐标矢量，n为小于等于N的正整数，ρ用于保持所述当前迭代中能量的恒定，是所述N帧第一分布的初始估计({ψ0m}N)的平均值，σm(r)为所述N帧第一分布的初始估计({ψ0m}N)的标准差分布，Cm为转换函数，用于将所述σm(r)转换成所述相似性Sm(r)，α、βm以及κm用于调整所述转换函数的形状。
  3. 如权利要求1所述的方法，其特征在于，所述基于所述上一轮迭代的所述N帧第一分布的第二估计({ψ″m-1}N)获得当前迭代的所述N帧第一分布的初始估计({ψ0m}N)包括：
    当所述N帧第一分布的第二估计({ψ″m-1}N)位于支撑区域内时，以所述N帧第一分布的第二估计({ψ″m-1}N)作为N帧第一分布的初始估计({ψ0m}N)；或者
    当所述N帧第一分布的第二估计({ψ″m-1}N)位于支撑区域外时，所述N帧第一分布的第二估计({ψ″m-1}N)在该区域的值为零。
  4. 如权利要求1所述的方法，其特征在于，所述得到当前迭代的所述N帧第一分布的第二估计({ψ″m}N)包括：
    基于所述N帧衍射图样({I}N)，得到所述当前迭代的N帧第二分布估计({Ψ′m}N)，所述当前迭代的所述N帧第二分布估计({Ψ′m}N)为所述N帧衍射图样({I}N)所对应的所述出射波在所述探测器处的复振幅分布的估计值；以及
    基于当前迭代的所述N帧第一分布的第一估计({ψ′m}N)和上一轮迭代的所述N帧第一分布的第二估计({ψ″m-1}N)，确定所述当前迭代的N帧第二分布估计({Ψ′m}N)反向传播到目标物处所对应的当前迭代的所述N帧第一分布的第二估计({ψ″m}N)。
  5. 如权利要求1所述的方法，其特征在于，所述当前迭代的所述N帧第一分布的第二估计({ψ″m}N)满足：
    其中，为逆傅里叶变换函数，所述k为所述探测器所在平面的坐标矢量，所述n为小于等于N的正整数，所述δ与所述N帧第一分布的第一估计({ψ′m}N)以及所述上一轮迭代的所述N帧第一分布的第二估计({ψ″m-1}N)成线性相关。
  6. 如权利要求1所述的方法，其特征在于，所述第m轮所述目标迭代还包括：
    当m＝K时，所述K为整数且K≥0，基于所述当前迭代的N帧第一分布的第二估计({ψ″m}N)，得到所述目标复振幅分布(O)的初始估计，其中，所述目标复振幅分布(O)的初始估计满足：
    其中，r为所述目标物所在平面的坐标矢量，n为小于等于N的正整数，OK(r；n)为目标复振幅分布(O)的初始估计，PK(r)为探针复振幅分布的初始估计，∈为预设的常数，是所述N帧第一分布的初始估计({ψ0m}N)的平均值。
  7. 如权利要求1所述的方法，其特征在于，所述第m轮所述目标迭代还包括：
    当m≥K时，所述K为整数且K≥0，基于所述当前迭代的N帧第一分布的第二估计({ψ″m-1}N)，更新所述目标复振幅分布(O)。
  8. 如权利要求7所述的方法，其特征在于，当前迭代更新后所述目标复振幅分布(O)的估计满足：
    其中，r为所述目标物所在平面的坐标矢量，n为小于等于N的正整数，Om-1(r；n)为上一轮迭代更新后所述目标复振幅分布的估计，Pm-1(r)为上一轮迭代更新后探针复振幅分布的估计，Pm(r)为当前轮迭代更新后探针复振幅分布的估计，*表示共轭复数，|.|max表示非负约束，α1、α2、α3和α4为预设的范围在0到1之间的常数，
  9. 一种成像装置，包括：
    至少一个存储介质，存储有至少一组指令集用于实现相干衍射成像；以及
    至少一个处理器，同所述至少一个存储介质通信连接，
    其中当所述成像装置运行时，所述至少一个处理器读取所述至少一个指令集并实施权利要求1-8中任一项所述的方法。
  10. 如权利要求9所述的装置，还包括：
    相干光源，被配置为发射一束相干波，所述装置运行时，所述相干波被约束成探针照射到目标物上并发生衍射；
    探测器，被配置为接收所述衍射产生的出射波并生成N帧衍射图样。
  11. 如权利要求10所述的装置，其特征在于，
    所述相干波包括电磁辐射、光子、X射线、电子、中子或质子中的至少一种。
PCT/CN2025/098904 2024-06-03 2025-06-03 一种相干衍射成像方法及装置 Pending WO2025252091A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202410708006.8A CN118294414B (zh) 2024-06-03 2024-06-03 一种相干衍射成像方法及装置
CN202410708006.8 2024-06-03

Publications (1)

Publication Number Publication Date
WO2025252091A1 true WO2025252091A1 (zh) 2025-12-11

Family

ID=91680668

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2025/098904 Pending WO2025252091A1 (zh) 2024-06-03 2025-06-03 一种相干衍射成像方法及装置

Country Status (2)

Country Link
CN (2) CN119223922A (zh)
WO (1) WO2025252091A1 (zh)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN119223922A (zh) * 2024-06-03 2024-12-31 南方科技大学 一种相干衍射成像方法及装置

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105548080A (zh) * 2016-01-15 2016-05-04 北京工业大学 一种连续太赫兹波空间扫描相干衍射成像系统及方法
US20160202163A1 (en) * 2013-08-16 2016-07-14 The General Hospital Corporation Portable diffraction-based imaging and diagnostic systems and methods
CN112697751A (zh) * 2020-12-07 2021-04-23 杭州电子科技大学 多角度照明无透镜成像方法、系统及装置
CN115128042A (zh) * 2022-06-30 2022-09-30 华中科技大学 一种离焦扫描的轴向叠层衍射成像方法和系统
CN115436404A (zh) * 2022-08-04 2022-12-06 上海精测半导体技术有限公司 一种计算相位成像方法及系统
CN117203516A (zh) * 2021-02-04 2023-12-08 南方科技大学 用于表征物体的方法和设备
CN118294414A (zh) * 2024-06-03 2024-07-05 南方科技大学 一种相干衍射成像方法及装置

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11221205B2 (en) * 2019-05-28 2022-01-11 University Of Central Florida Research Foundation, Inc. Iterative optical diffraction tomography (iODT) method and applications
CN115452743B (zh) * 2022-09-13 2025-11-11 南京理工大学 基于部分相干发光二极管照明的无透镜单帧相位恢复方法
CN116858809A (zh) * 2023-07-12 2023-10-10 上海交通大学 基于相位掩模增强的散射介质传输矩阵测量方法和装置
CN117274096B (zh) * 2023-09-28 2026-01-20 深圳技术大学 一种生成二值化空域约束的x射线相干衍射图像恢复方法

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20160202163A1 (en) * 2013-08-16 2016-07-14 The General Hospital Corporation Portable diffraction-based imaging and diagnostic systems and methods
CN105548080A (zh) * 2016-01-15 2016-05-04 北京工业大学 一种连续太赫兹波空间扫描相干衍射成像系统及方法
CN112697751A (zh) * 2020-12-07 2021-04-23 杭州电子科技大学 多角度照明无透镜成像方法、系统及装置
CN117203516A (zh) * 2021-02-04 2023-12-08 南方科技大学 用于表征物体的方法和设备
CN115128042A (zh) * 2022-06-30 2022-09-30 华中科技大学 一种离焦扫描的轴向叠层衍射成像方法和系统
CN115436404A (zh) * 2022-08-04 2022-12-06 上海精测半导体技术有限公司 一种计算相位成像方法及系统
CN118294414A (zh) * 2024-06-03 2024-07-05 南方科技大学 一种相干衍射成像方法及装置

Also Published As

Publication number Publication date
CN118294414A (zh) 2024-07-05
CN118294414B (zh) 2024-10-18
CN119223922A (zh) 2024-12-31

Similar Documents

Publication Publication Date Title
Godard et al. Noise models for low counting rate coherent diffraction imaging
Yan Ptychographic phase retrieval by proximal algorithms
Chang et al. Advanced denoising for X-ray ptychography
WO2025252091A1 (zh) 一种相干衍射成像方法及装置
US12154758B2 (en) Observation method by means of scanning transmission electron microscope, scanning transmission electron microscope system, and computer readable medium
Leidl et al. Influence of loss function and electron dose on ptychography of 2D materials using the Wirtinger flow
Girard et al. Sparse representations and convex optimization as tools for LOFAR radio interferometric imaging
CN113870184B (zh) 基于x射线吸收图像生成相位衬度图像和暗场图像的方法、装置、计算设备和存储介质
Pascazio et al. Statistical regularization in linearized microwave imaging through MRF-based MAP estimation: Hyperparameter estimation and image computation
Moths et al. Bayesian algorithms for recovering structure from single-particle diffraction snapshots of unknown orientation: a comparison
Wang et al. Fast compressive beamforming with a modified fast iterative shrinkage-thresholding algorithm
Geng et al. NUDIM: A non-uniform fast Fourier transform based dual-space constraint iterative reconstruction method in biological electron tomography
Manekar et al. Low-light phase retrieval with implicit generative priors
Balakin et al. Improvement of the optical image reconstruction based on multiplexed quantum ghost images
Zhao et al. Global ghost imaging
Yokoyama et al. Phase retrieval algorithm based on total variation regularization for ferromagnetic domain patterns
EP4075126B1 (en) Phase retrieval in electron microscopy
Xian et al. Randomized Kaczmarz method for single-particle x-ray image phase retrieval
Bortel et al. Fast inside-source X-ray fluorescent holography
Bangun et al. FlowTIE: Flow-based Transport of Intensity Equation for Phase Gradient Estimation from 4D-STEM Data
Tong et al. Learning-based surface deformation recovery for large radio telescope antennas
Deng et al. Deep learning-based compressed sampling reconstruction algorithm for digitizing intensive neutron ToF signals: C. Deng et al.
Massone et al. The use of electron maps to constrain some physical properties of solar flares
Chowdhury et al. Fast hyperspectral reconstruction for neutron computed tomography using subspace extraction
Manekar et al. LoDIP: Low light phase retrieval with deep image prior

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 25819203

Country of ref document: EP

Kind code of ref document: A1