Image super-resolution reconstruction method and device based on quantum efficiency

Abstract

The application discloses a kind of quantum efficiency-based image super-resolution reconstruction method and equipment, method includes: the corresponding calibration image of each detection period detected by detector is obtained, the corresponding received photon number of each sub-region of each pixel in each detection period is determined;Based on the corresponding calibration image of each detection period and the corresponding received photon number of each sub-region of each pixel in each detection period, the quantum efficiency value of each sub-region of each pixel is determined, and based on the quantum efficiency value of each sub-region of each pixel, quantum efficiency distribution matrix is determined;Obtain the observation image detected by detector when each sub-pixel displacement, obtain multiple observation images;Based on quantum efficiency distribution matrix, each frame of observation image is reconstructed, and the corresponding photon distribution diagram of each frame of observation image is obtained, based on the corresponding photon distribution diagram of each frame of observation image, multiple observation images are fused, and super-resolution image is determined.

Description

Technical Field

[0001] This invention relates to the field of computer tomography, and in particular to a quantum-efficiency-based image super-resolution reconstruction method and computing device. Background Technology

[0002] X-ray imaging technology has become the most important non-destructive testing technology in fields such as medical diagnosis, industrial flaw detection, and materials analysis. Due to its penetrating power into the interior of objects, it plays a crucial role in scenarios such as micron-level defect detection in aerospace components, early screening of medical diseases, and non-destructive analysis of the internal patterns of cultural relics. Spatial resolution directly determines the ability to resolve minute structures in X-ray images and is a key factor in evaluating the performance of an imaging system.

[0003] In X-ray imaging systems, spatial resolution is primarily constrained by two physical factors: the focal size of the X-ray source and the unit size of the detector. These two factors together determine the theoretical upper limit of the imaging system's resolution. The X-ray focal point has a certain size and is not an ideal point source, introducing geometric blurring in X-ray imaging. Furthermore, when the focal size is five times the detector unit size, the focal size becomes the primary factor limiting resolution, while the influence of the detector unit size becomes negligible. Therefore, under typical imaging conditions, the focal size is the most significant factor limiting the improvement of X-ray image resolution.

[0004] To overcome the limitation of focal size on X-ray image resolution, super-resolution reconstruction methods based on convolutional neural networks (CNNs) are currently the mainstream technology. This method improves image resolution at the algorithmic level by training and learning the mapping relationship between low-resolution and high-resolution images. However, this type of method has obvious limitations: First, it is highly dependent on large-scale paired datasets that can accurately reflect the real degradation process of X-ray imaging. However, X-ray imaging degradation is affected by multiple physical factors, making it difficult to obtain truly paired datasets, resulting in high computational difficulty and limited algorithm generalization ability. Second, it achieves pixel-level pure algorithmic mapping without touching the core physical processes of imaging. It cannot optimize underlying physical parameters such as quantum efficiency, and the reconstructed details lack physical realism, making it prone to artifacts and failing to meet the requirements for image realism and accuracy. Summary of the Invention

[0005] To address the existing technical problems, this invention provides a quantum-efficiency-based image super-resolution reconstruction method, computer program product, and computing device, which can effectively eliminate image distortion caused by detector response non-uniformity, noise, and nonlinearity, restore the true incident X-ray photon distribution, and improve the physical fidelity of the image.

[0006] In a first aspect, a quantum-efficiency-based image super-resolution reconstruction method is provided, comprising: controlling a radiation source in a reconstruction system to generate an X-ray standing wave field, and driving the X-ray standing wave field to translate, so that in each detection cycle, the interference fringes of the X-ray standing wave field sequentially illuminate each sub-region of each pixel on the detector; acquiring calibration images corresponding to each detection cycle detected by the detector, and determining the number of received photons in each sub-region of each pixel in each detection cycle; based on the calibration images corresponding to each detection cycle and the number of received photons in each sub-region of each pixel in each detection cycle... The number of photons received is used to determine the quantum efficiency value of each sub-region of each pixel, and a quantum efficiency distribution matrix is ​​determined based on the quantum efficiency value of each sub-region of each pixel. The X-ray standing wave field is controlled to perform multiple sub-pixel displacements centered on a reference position, and the observation image detected by the detector at each sub-pixel displacement is acquired to obtain multiple frames of observation images. Based on the quantum efficiency distribution matrix, each frame of observation images is reconstructed to obtain a photon distribution map corresponding to each frame of observation images. Based on the photon distribution map corresponding to each frame of observation images, the multiple frames of observation images are fused to determine a super-resolution image.

[0007] In a second aspect, a computer program product is provided, comprising a computer program, characterized in that, when executed by a processor, the computer program implements the quantum-efficiency-based image super-resolution reconstruction method of the first aspect of this application.

[0008] Thirdly, a computing device is provided, characterized in that it includes a memory and a processor, the memory storing a computer program, which, when executed by the processor, causes the processor to perform the quantum-efficiency-based image super-resolution reconstruction method of the first aspect of this application.

[0009] This application fully utilizes the high spatial coherence and fine interference fringe structure of X-ray standing wave fields to achieve precise subwavelength-level optical field excitation within detector pixels. By sequentially irradiating each sub-region, the response characteristics of each position within each pixel are fully acquired, thereby achieving precise calibration of pixel-level response inhomogeneities. By constructing a quantum efficiency distribution matrix, the photoelectric conversion differences of the detector at the sub-pixel scale can be accurately characterized, providing a high-precision system response model for subsequent imaging correction and significantly improving the accuracy of incident light field reconstruction. Reconstructing the observed image based on the calibrated quantum efficiency distribution matrix can effectively eliminate image distortion caused by detector response inhomogeneity, noise, and nonlinearity, restoring the true incident X-ray photon distribution and improving the physical fidelity of the image. Attached Figure Description

[0010] Figure 1 This is an application environment diagram of a quantum-efficiency-based image super-resolution reconstruction method in one embodiment. Figure 2 This is a flowchart of an image super-resolution reconstruction method based on quantum efficiency in one embodiment; Figure 3 This is a schematic diagram of a quantum-efficiency-based image super-resolution reconstruction device in one embodiment; Figure 4 This is a schematic diagram of a computing device in one embodiment. Detailed Implementation

[0011] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0012] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the specification of this invention is for the purpose of describing particular embodiments only and is not intended to limit the scope of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0013] In the following description, the expression “some embodiments” refers to a subset of all possible embodiments. However, it should be understood that “some embodiments” can be the same subset or different subsets of all possible embodiments and can be combined with each other without conflict.

[0014] See Figure 1 This diagram illustrates the application environment of a quantum-efficiency-based image super-resolution reconstruction method in one embodiment. The quantum-efficiency-based image super-resolution reconstruction method is applied in a computing device 10, which can acquire observation images obtained through a detection device. A radiation source in the detection device rotates around the object under test, acquiring observation images of the object from different angles. These projection data are then used to reconstruct a cross-sectional image of the scanned area. The acquisition of the observation images involves the attenuation of the radiation source; tissues of different densities absorb radiation to varying degrees, and this difference is used to distinguish different tissue structures. During scanning, a narrow beam of radiation is emitted from a X-ray tube, and the attenuation of the radiation is measured by a detector opposite the tube. The data collected by the detection device are radiation projection and contour data. The computing device 10 can reconstruct the observation images based on the quantum-efficiency-based image super-resolution reconstruction method to obtain a super-resolution image corresponding to the object under test.

[0015] Please see Figure 2 This is a flowchart of a quantum-efficiency-based image super-resolution reconstruction method provided in an embodiment of this application. The quantum-efficiency-based image super-resolution reconstruction method is applied in a computing device and includes the following steps: S11. The X-ray source in the control reconstruction system generates an X-ray standing wave field and drives the X-ray standing wave field to translate so that the interference fringes of the X-ray standing wave field sequentially illuminate each sub-region of each pixel on the detector during each detection cycle.

[0016] In this embodiment, the reconstruction system includes a radiation source, a displacement platform for controlling the movement of the radiation source, a laser interferometer for calibrating the displacement platform, and a detector. The detector surface has multiple pixel regions, and each pixel region is divided into multiple sub-regions. For example, one pixel region is divided into 2... A sub-region of 2. Each sub-region is equivalent to a sub-pixel region.

[0017] Two coherent X-ray beams are emitted from a radiation source. The interference of these two beams generates a stable X-ray standing wave field. By precisely controlling the wavelength, phase, and amplitude of the X-rays, an X-ray standing wave field with a known photon distribution is generated, which can illuminate each sub-region within each pixel area of ​​the detector. By generating a stable X-ray standing wave field, quantum efficiency measurements can be performed in a stable environment, thereby enabling precise measurement of the quantum efficiency distribution matrix within each pixel.

[0018] S12. Obtain the calibration image corresponding to each detection period detected by the detector, and determine the number of received photons for each sub-region of each pixel in each detection period.

[0019] To pre-calibrate the quantum efficiency distribution matrix corresponding to the detector, different calibration images need to be obtained at different detection periods. This can be achieved by performing multiple measurements at a single projection angle, resulting in multiple different calibration images. The different detection periods correspond to different detection time periods.

[0020] An X-ray standing wave field consists of slowly moving bright and dark fringes. When an X-ray standing wave field scans a pixel, different positions of the fringes (bright / dark fringes) will successively illuminate different sub-regions within the pixel. The X-ray standing wave field interference fringe pattern consists of alternating bright and dark fringes that move slowly. The fringes will scan from left to right and from top to bottom, covering every pixel of the detector. Each scan yields a standing wave field calibration map. Each movement assigns a different grayscale value to the pixel. At the same time, based on the extent of each sub-region and the light intensity illuminating each sub-region, the number of photons received in each sub-region can be estimated.

[0021] S13. Based on the calibration image corresponding to each detection period and the number of received photons of each sub-region of each pixel in each detection period, determine the quantum efficiency value of each sub-region of each pixel, and determine the quantum efficiency distribution matrix based on the quantum efficiency value of each sub-region of each pixel.

[0022] In this embodiment, the quantum efficiency distribution matrix includes the quantum efficiency values ​​of each sub-region of each pixel. The quantum efficiency distribution matrix is ​​a detector response matrix obtained through X-ray standing wave field calibration. Each element of the matrix corresponds to the quantum efficiency value of each sub-region within a single pixel of the detector, characterizing the distribution of the ability of each sub-region to convert incident X-ray photons into electrical signals. The quantum efficiency distribution matrix is ​​used to correct the non-uniformity of the detector's response within a pixel, establishing a precise mapping relationship between the number of incident photons and the detector's output grayscale value, providing a basis for response correction for subsequent incident light field reconstruction and sub-pixel-level fusion. The quantum efficiency value of a sub-region characterizes the ability of each sub-region within a pixel of the detector to convert incident X-ray photons into detectable electrical signals; its magnitude reflects the photoelectric conversion efficiency of the sub-region, and is used to correct the non-uniformity of the detector's response within a pixel, establishing a precise mapping relationship between the number of incident photons and the detector's output grayscale value, providing a basis for response correction for incident light field reconstruction and sub-pixel-level fusion.

[0023] The quantum efficiency values ​​of each sub-region of each pixel are used to generate a two-dimensional quantum efficiency distribution according to the order of the pixels. Following the one-dimensional index order corresponding to a preset photon distribution map size, the two-dimensional quantum efficiency distribution is mapped to a one-dimensional array with a length equal to the total number of pixels corresponding to the preset photon distribution map size. The quantum efficiency response matrix is ​​determined using the elements of the one-dimensional array as diagonal elements. The preset photon distribution map size is the size of the photon distribution map, for example, W. H. Where H is the height of the photon distribution map and W is the width of the photon distribution map.

[0024] S14. Control the X-ray standing wave field to perform multiple sub-pixel displacements centered on the reference position, and acquire the observation image detected by the detector at each sub-pixel displacement to obtain multiple frames of observation images.

[0025] In this embodiment, at a projection angle, a set of sub-pixel-level displacement sequences is preset with the reference position as the center; the displacement step size is smaller than the pixel size of the detector (e.g., the step size is 1 / 4 to 1 / 10 of the pixel size), and the displacement direction covers at least two orthogonal directions (such as horizontal and vertical directions) to ensure that the standing wave field fringes can sweep through all sub-regions within each pixel. The phase difference or propagation angle of the X-ray interferometry system is adjusted to drive the X-ray standing wave field to perform sub-pixel-level translation along the preset path, thus obtaining the observation image acquired at each sub-pixel displacement.

[0026] Optionally, before controlling the X-ray standing wave field to perform multiple sub-pixel displacements centered on a reference position, the method further includes: The displacement platform of the control X-ray source is moved to the preset initial calibration position. The laser interferometer in the reconstruction system emits laser and receives the reflected signal. The actual coordinate data of the displacement platform at the preset initial calibration position is collected. The actual coordinate data is compared with the preset initial coordinates of the preset initial calibration position to obtain the displacement deviation value. Based on the displacement deviation value, the displacement platform is corrected in real time through a preset compensation algorithm, and the motion parameters of the displacement platform are dynamically adjusted.

[0027] Before controlling the X-ray standing wave field to perform multiple sub-pixel displacements centered on a reference position, an initial accuracy calibration procedure for the displacement platform is first executed. The displacement platform corresponding to the X-ray source is moved to a preset initial calibration position. A laser interferometer in the reconstruction system emits a measurement laser and receives the returned reflected signal, acquiring the actual coordinate data of the displacement platform at this preset initial calibration position. The acquired actual coordinates are compared with the preset initial coordinates to calculate the displacement deviation value. Based on this displacement deviation value, a preset compensation algorithm is used to correct the motion control quantity of the displacement platform in real time, dynamically adjusting the driving parameters of the displacement platform. This eliminates systematic deviations caused by initial positioning errors, environmental drift, and mechanical backlash, providing a reliable position reference for subsequent high-precision sub-pixel displacement of the X-ray standing wave field and ensuring spatial positioning accuracy during multi-frame observation image acquisition.

[0028] S15. Based on the quantum efficiency distribution matrix, reconstruct each frame of observation image to obtain the photon distribution map corresponding to each frame of observation image. Based on the photon distribution map corresponding to each frame of observation image, fuse multiple frames of observation images to determine the super-resolution image.

[0029] In this embodiment, based on the pre-calibrated quantum efficiency distribution matrix, incident light field reconstruction is performed on each frame of the observed image. The quantum efficiency distribution matrix is ​​introduced, and iterative inversion is performed using the observed images to accurately reconstruct the true incident X-ray photon distribution. Then, CT tomography reconstruction is completed based on the true photon distribution map. Using the grayscale value output by the detector as the known observed value, and the quantum efficiency distribution matrix as the system response model, the inverse problem is solved through iterative optimization. Distortions caused by factors such as uneven response within detector pixels and readout noise are eliminated, reconstructing the true incident X-ray photon distribution map corresponding to each frame of the observed image, resulting in multiple frames of physically consistent photon distribution data. After reconstructing the photon distribution maps of multiple frames, super-resolution fusion processing is performed based on a sub-pixel grid. For each sub-region location, the photon distribution maps of all frames are traversed, and the photon count value and corresponding weight information corresponding to that sub-pixel in each frame are found and fused weighted to determine the super-resolution image. By using the aforementioned sub-pixel-level weighted fusion, the complementary information between multiple frames is fully utilized to ultimately generate a super-resolution X-ray image with higher spatial resolution, clearer details, and more uniform response.

[0030] Using the above method, the high spatial coherence and fine interference fringe structure of X-ray standing wave fields can be fully utilized to achieve precise subwavelength-level optical field excitation within the detector pixels. By sequentially irradiating each sub-region, the response characteristics of each location within each pixel can be fully acquired, thereby achieving precise calibration of pixel-level response inhomogeneities. By constructing a quantum efficiency distribution matrix, the photoelectric conversion differences of the detector at the sub-pixel scale can be accurately characterized, providing a high-precision system response model for subsequent imaging correction and significantly improving the accuracy of incident light field reconstruction.

[0031] By driving the X-ray standing wave field to undergo multiple sub-pixel displacements centered on a reference position, complementary observation data with sub-pixel offset information across multiple frames can be acquired without altering the detector's hardware structure. This overcomes the spatial resolution limitation imposed by the detector's inherent pixel size, providing ample redundancy for super-resolution reconstruction. Simultaneously, high-precision displacement calibration before each displacement ensures that the standing wave field position deviation is controlled within an extremely small range, guaranteeing a reliable spatial correspondence between multiple frames and avoiding reconstruction artifacts caused by displacement errors.

[0032] During the imaging stage, the observed image is reconstructed based on the calibrated quantum efficiency distribution matrix. This effectively eliminates image distortion caused by detector response inhomogeneity, noise, and nonlinearity, restoring the true incident X-ray photon distribution and improving the physical fidelity of the image. Furthermore, sub-pixel-level weighted fusion of multiple photon distribution maps fully integrates complementary detail information at different displacement positions, suppresses single-frame noise, and further improves image spatial resolution and contrast. Ultimately, a super-resolution image with higher resolution, clearer details, and superior imaging quality than the detector's original resolution is obtained. The overall scheme combines high-precision calibration, sub-pixel sampling, and iterative reconstruction, solving the accuracy bottleneck caused by uneven response within detector pixels in traditional X-ray imaging and overcoming the resolution limitations imposed by hardware pixel size. Without increasing detector cost, it significantly improves the overall performance of the X-ray imaging system.

[0033] In some embodiments, determining the number of received photons for each sub-region of each pixel in each detection period includes: Within a detection cycle, for a sub-region, obtain the time range of the detection cycle and the spatial range of the sub-region; Based on the intensity distribution function of the X-ray standing wave field, a triple integral is performed on the spatial range and time range of the sub-region to obtain the cumulative number of photons received by the sub-region within the detection period, and the cumulative number of photons is used as the number of received photons.

[0034] In this embodiment, Indicates the first In the first detection cycle Within the first pixel The number of photons received in each sub-region is obtained by integrating the intensity distribution function of the X-ray standing wave field. The calculation formula is as follows: in: , These represent the starting and ending points of the horizontal coordinate within the spatial range of the region, respectively. , These represent the starting and ending points of the vertical coordinate within the spatial area, respectively. , These represent the start and end points of time within a time range, respectively. When the first... When calibrating an image, the above integration can be used to accurately calculate the first exposure at this moment. Within the first pixel The number of photons received in each sub-region. Indicates the first Each detection cycle is based on position coordinates The light intensity distribution function.

[0035] By using the intensity distribution function of the X-ray standing wave field to calculate the cumulative photon count over the spatial range and detection time period of a sub-region through triple integration, the cumulative photon count can be accurately reflected in terms of the continuous spatial distribution and temporal cumulative effect of the incident X-rays. This avoids the approximate errors caused by simply equating pixels or sub-regions to single points and using average intensity calculations. This method fully considers the intensity fluctuations, edge gradations, and light field changes during exposure of the standing wave field interference fringes within the sub-region, accurately reconstructing the total number of photons actually received in the sub-region from a physical perspective. This provides a high-precision, high-fidelity theoretical input for subsequent quantum efficiency calibration. Compared to traditional discrete sampling or approximate estimation methods, triple integration can more realistically match the physical process of the interaction between X-rays and the detector, significantly improving the calculation accuracy of the number of photons received in the sub-region, thereby improving the calibration accuracy of the quantum efficiency distribution matrix and laying a reliable foundation for subsequent incident light field reconstruction and super-resolution imaging.

[0036] In some embodiments, the quantum efficiency value of each sub-region of each pixel is determined based on the calibration image corresponding to each detection period and the number of received photons in each sub-region of each pixel in each detection period, including: The measured pixel values ​​of each pixel in each detection cycle are obtained from the calibration images corresponding to each detection cycle. Based on the measured pixel value of each pixel in each detection period and the number of received photons of each sub-region of each pixel in each detection period, the response model corresponding to each pixel is fitted to determine the response parameter value corresponding to each sub-region of each pixel in the response model corresponding to each pixel. The quantum efficiency value of each sub-region of each pixel is determined based on the response parameter values ​​corresponding to each sub-region of each pixel.

[0037] In this embodiment, the measured pixel value of a pixel represents the pixel value obtained from the calibration image.

[0038] Optionally, the expression for the response model is: in, This represents the calibration image corresponding to one detection cycle. Line 1 The pixel value of the column pixel. These represent the row and column indices of the sub-region within a pixel, respectively. Indicates the total number of indexes. Indicates the first Within the first pixel The number of photons received in each sub-region Indicates the first Within the first pixel The offset coefficient of each sub-region Indicates the first Within the first pixel Linear response coefficients of each subregion Indicates the first Within the first pixel Nonlinear response coefficients of each subregion.

[0039] Fine, stable, and slowly moving bright and dark fringes are generated using an X-ray standing wave field. These fringes sequentially sweep across different locations within a pixel, allowing each sub-region to receive X-ray photons of varying intensities at different times. During the standing wave field scan, the detector records multiple calibration images at different times. Each calibration image corresponds to a set of known incident photon distribution data, representing the number of photons received by each sub-region of each pixel in each detection cycle. Since the photoelectric response capability of each sub-region differs, a system of equations can be established using multiple sets of observations to deduce the response parameter values ​​for each sub-region. These response parameter values ​​reflect the response sensitivity of each sub-region.

[0040] For example, if the response parameter value is three, each pixel region is divided into... If there are multiple sub-regions, then each sub-region requires three parameter values. Based on the above model, establishing the mapping relationship requires... By establishing a response model for each sub-region within a pixel in the calibrated image, and then performing fitting analysis based on least squares method or nonlinear regression analysis, the response parameter values ​​corresponding to each sub-region can be obtained.

[0041] Optionally, the response parameter values ​​include the values ​​of the offset coefficient, the linear response coefficient, and the nonlinear response coefficient. Based on the response parameter values ​​corresponding to each sub-region of each pixel, the quantum efficiency value of each sub-region of each pixel is determined, including: Using the quantum efficiency calculation formula, based on the values ​​of the offset coefficient, linear response coefficient, and nonlinear response coefficient corresponding to each sub-region of each pixel, the quantum efficiency value of each sub-region of each pixel is determined. The quantum efficiency calculation formula is as follows: in Indicates the first Within the first pixel Quantum efficiency values ​​for each subregion.

[0042] A sub-region-level response model incorporating offset, linear, and quadratic nonlinear terms is employed. The model is solved by fitting the measured pixel values ​​across multiple cycles with precisely calculated received photon counts. This approach accurately characterizes the detector's true photoelectric conversion properties at the sub-pixel scale, offering significant advantages over traditional single quantum efficiency models. This model utilizes offset coefficients... Effectively compensates for fixed background errors such as dark current and circuit bias, avoiding the impact of signal offset on calibration results when there is no light input; through the linear response coefficient Accurate characterization of the fundamental photoelectric conversion capability of the sub-region provides the core physical basis for quantum efficiency; at the same time, a second-order nonlinear coefficient is introduced. It can accurately describe the nonlinear distortions such as response saturation and curve bending that occur in the detector under high photon number incident conditions, making the model more consistent with the actual physical process.

[0043] By performing overall fitting based on measurement data from multiple detection cycles, multiple sets of redundant information can be fully utilized to suppress noise interference and improve the stability and accuracy of solving sub-region-level response parameters. By solving the response parameters independently pixel-by-pixel and sub-region-by-subregion, the non-uniform distribution characteristics of the response within a pixel can be fully reflected. The resulting quantum efficiency value is no longer a single fixed value, but a precise characterization related to incident intensity and spatial position. This approach significantly improves the accuracy and applicability of detector response modeling, providing a high-fidelity system model for subsequent incident light field reconstruction, effectively reducing nonlinear errors and imaging artifacts caused by fixed bias, and significantly improving the realism and reliability of super-resolution reconstruction results.

[0044] In some embodiments, based on the quantum efficiency distribution matrix, each frame of the observed image is reconstructed to obtain a super-resolution photon distribution map corresponding to each frame of the observed image, including: For the reconstruction process of any observed image, the photon distribution map is initialized to obtain an initial image. Based on the initial image, iteration is performed to obtain the current photon distribution map corresponding to the current iteration number. Based on the quantum efficiency distribution matrix and the current photon distribution map, the current predicted image is determined; Calculate the current reconstruction error between the observed image and the current predicted image. If the reconstruction error is greater than the preset error, continue to iterate and update the current photon distribution map until the current reconstruction error is less than the preset error. Then, use the photon distribution map after stopping the iteration as the photon distribution map corresponding to the observed image.

[0045] In this embodiment, the initial image represents the initialized photon distribution map used for the first iteration. The current photon distribution map represents the photon distribution map corresponding to the current iteration number. The current predicted image is equal to the product of the quantum efficiency distribution matrix and the current photon distribution map.

[0046] For the reconstruction of the incident photon distribution in each frame of the observed image, an iterative optimization approach is adopted. First, the photon distribution map is initialized based on prior information to obtain an initial photon distribution image. In each iteration, based on the current photon distribution map obtained in the current iteration, and combined with the determined quantum efficiency distribution matrix, a forward calculation is performed to obtain the current predicted image corresponding to the current photon distribution. Then, the reconstruction error between the actual observed image and the current predicted image is calculated, and it is determined whether this error meets the preset accuracy requirement. If the current reconstruction error is still greater than the preset error, the photon distribution map is iteratively updated according to the error magnitude, continuously approximating the true incident light field distribution. When the reconstruction error after iteration is less than the preset error, the iteration process stops, and the finally converged photon distribution map is taken as the true incident photon distribution result corresponding to that frame of the observed image.

[0047] In some embodiments, fusing multiple observation images based on the photon distribution map corresponding to each frame of observation image to determine the super-resolution image includes: Perform image registration on multiple frames of observation images to determine the corresponding registered observation image for each frame of observation images; The quality of each frame of the registered observation image is evaluated, and the quality weight corresponding to each frame of the registered observation image is determined. Based on the quality weights corresponding to each frame of the registered observation image, the photon distribution values ​​of each sub-region of each pixel in each frame of the registered observation image are fused to determine the fused pixel values ​​of each sub-region of each pixel in the super-resolution image.

[0048] In this embodiment, before performing multi-frame image fusion, precise image registration is performed on each observation image. Based on the sub-pixel displacement information of the X-ray standing wave field, images with slight offsets between different frames are spatially aligned to eliminate relative deviations between frames caused by displacement platform movement, mechanical vibration, or optical path disturbances. This establishes a unified and precise spatial correspondence between frames, resulting in registered observation images with strictly matched spatial positions. Through image registration, it is ensured that the photon distribution data of each frame accurately correspond spatially during subsequent sub-pixel-level fusion, avoiding image blurring, artifacts, and detail distortion caused by inter-frame misalignment. This provides a reliable spatial reference for subsequent weighted fusion, further improving the clarity and structural fidelity of the super-resolution image.

[0049] The quality of each frame of the spatially registered observation image is evaluated. By analyzing indicators such as image signal-to-noise ratio, sharpness, uniformity, and effective information content, the imaging quality and reliability of each registered observation image are comprehensively assessed, and a matching quality weight is assigned to each frame. Frames with higher imaging quality, lower noise, and more stable effective signals are assigned a larger quality weight; conversely, frames with high noise, abnormal signals, or local distortion are assigned a smaller quality weight. This allows for the effective enhancement of high-quality frames and the reasonable suppression of low-quality frames during subsequent fusion. When evaluating the quality of each registered observation image frame, a combination of objective evaluation methods can be used. For example, the ratio of signal strength to noise standard deviation can be calculated based on the image signal-to-noise ratio (SNR), with a higher SNR indicating better image quality. Image sharpness can be assessed by calculating image gradient energy or spatial frequency response, with frames exhibiting clear textures and sharp edges receiving higher weights. The stability of the standing wave field signal can be analyzed based on the image's local uniformity, eliminating abnormal frames caused by field strength fluctuations or noise abrupt changes. Furthermore, similarity measurements between a single frame and the mean images of multiple frames can be used, such as calculating mutual information or dot product similarity; higher similarity indicates stronger data reliability for that frame. A comprehensive score is obtained by combining these indicators, and the normalized score is used as the quality weight for each frame, making the fusion process more reasonable and robust.

[0050] Optionally, based on the fusion weights of each sub-region of each pixel and the quality weights corresponding to each frame of registered observation images, the photon distribution values ​​of each sub-region of each pixel in each frame of registered observation images are fused to determine the fused pixel values ​​of each sub-region of each pixel in the super-resolution image, including: For a sub-region of a pixel, the quality weights corresponding to each frame of the registered observation image and the photon distribution values ​​of the sub-regions of the pixels in each frame of the registered observation image are substituted into the sub-region pixel value fusion formula to determine the fused pixel value of the sub-region of the photon distribution map pixel. The sub-region pixel value fusion formula is as follows: ,in Represents the first pixel in a super-resolution image. The merged pixel values ​​of each sub-region Indicates the first frame, Indicates the total number of frames. Indicates the first Quality weights corresponding to the frame-registered observation images Indicates the first The first photon distribution map corresponding to the frame registration observation image Photon distribution values ​​for each sub-region.

[0051] Based on the quality weights corresponding to the registered observation images of each frame, the photon distribution values ​​of sub-regions at the same spatial location in each frame are weighted, summed, and normalized to obtain the fused pixel values ​​corresponding to that sub-region. By adaptively adjusting the contribution of each frame's data to the fusion result through quality weights, the noise and distortion caused by low-quality frames are suppressed while fully utilizing the complementary information of sub-pixels from multiple frames. Finally, the accurate fused pixel values ​​of each sub-region within each pixel in the super-resolution image are obtained.

[0052] X-ray standing wave fields inherently possess high coherence and fine interference fringes. By controlling sub-pixel displacement, the fringes sequentially sweep across different sub-regions within each pixel of the detector, effectively performing multiple fine samplings within a single physical pixel and accumulating a wealth of previously indistinguishable internal spatial details. Secondly, by precisely modeling and correcting intra-pixel response inhomogeneities, nonlinearities, and offset errors using a quantum efficiency distribution matrix, a true and distortion-free incident photon distribution can be reconstructed from the observed image, ensuring high fidelity and spatial consistency for each frame. Based on this, multiple frames of photon distribution maps with sub-pixel offsets are registered and weighted, integrating complementary details scattered across different frames into a single high-resolution grid. Inter-frame redundancy is used to suppress noise and fill in undersampled areas, ultimately reconstructing a super-resolution image with a spatial resolution significantly higher than the detector's original physical resolution.

[0053] In another aspect, this application provides a computer program product, including a computer program that, when executed by a processor, implements the quantum-efficiency-based image super-resolution reconstruction method of any embodiment of this application.

[0054] Among them, in the computer program product, the optional implementation form of the program module architecture of the computer program that implements each step of the target recognition method can be a quantum efficiency-based image super-resolution reconstruction device.

[0055] Please see Figure 3One embodiment of this application provides an image super-resolution reconstruction device based on quantum efficiency, comprising: a control module 31, which controls a radiation source in the reconstruction system to generate an X-ray standing wave field and drives the X-ray standing wave field to translate such that the interference fringes of the X-ray standing wave field sequentially illuminate each sub-region of each pixel on the detector during each detection cycle; an acquisition module 32, which acquires calibration images corresponding to each detection cycle detected by the detector; and a determination module 33, which determines the number of received photons in each sub-region of each pixel during each detection cycle; the determination module 33 is further configured to determine the number of received photons in each sub-region of each pixel during each detection cycle based on the calibration images corresponding to each detection cycle and the number of received photons in each sub-region of each pixel during each detection cycle. The control module 31 is further configured to control the X-ray standing wave field to perform multiple sub-pixel displacements centered on a reference position, and the acquisition module 32 is further configured to acquire the observation image detected by the detector at each sub-pixel displacement, thereby obtaining multiple frames of observation images. The determination module 33 is further configured to reconstruct each frame of the observation images based on the quantum efficiency distribution matrix, thereby obtaining a photon distribution map corresponding to each frame of the observation images, and to fuse multiple frames of the observation images based on the photon distribution map corresponding to each frame of the observation images to determine a super-resolution image.

[0056] Please see Figure 4 In another aspect of this application, a computing device 10 is also provided, including a memory 3011 and a processor 3012. The memory 3011 stores a computer program, and when the computer program is executed by the processor, the processor 3012 performs the steps of the quantum efficiency-based image super-resolution reconstruction method provided in any of the above embodiments of this application. The computing device 10 can be (e.g., a desktop computer, laptop computer, tablet computer, handheld computer, smart speaker, server, etc.), a mobile phone (e.g., a smartphone, cordless phone, etc.), a wearable device (e.g., a pair of smart glasses or a smartwatch), or a similar device.

[0057] The processor 3012 is the control center, connecting various parts of the computer device via various interfaces and lines. It executes software programs and / or modules stored in the memory 3011, and calls data stored in the memory 3011 to perform various functions and process data. Optionally, the processor 3012 may include one or more processing cores; preferably, the processor 3012 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user page, and applications, and the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into the processor 3012.

[0058] The memory 3011 can be used to store software programs and modules. The processor 3012 executes various functional applications and data processing by running the software programs and modules stored in the memory 3011. The memory 3011 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, application programs required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 3011 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, the memory 3011 may also include a memory controller to provide the processor 3012 with access to the memory 3011.

[0059] In another aspect, this application also provides a storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the quantum-efficiency-based image super-resolution reconstruction method provided in any of the above embodiments of this application.

[0060] Those skilled in the art will understand that all or part of the processes in the methods provided in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0061] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. The scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A quantum-efficiency-based image super-resolution reconstruction method, characterized in that, include: The X-ray source in the control reconstruction system generates an X-ray standing wave field and drives the X-ray standing wave field to translate so that the interference fringes of the X-ray standing wave field sequentially illuminate each sub-region of each pixel on the detector during each detection cycle. Acquire calibration images corresponding to each detection period detected by the detector, and determine the number of received photons in each sub-region of each pixel in each detection period. Based on the calibration image corresponding to each detection period and the number of received photons of each sub-region of each pixel in each detection period, the quantum efficiency value of each sub-region of each pixel is determined, and based on the quantum efficiency value of each sub-region of each pixel, the quantum efficiency distribution matrix is ​​determined. The X-ray standing wave field is controlled to undergo multiple sub-pixel displacements centered on a reference position, and the observation image detected by the detector at each sub-pixel displacement is acquired to obtain multiple frames of observation images. Based on the quantum efficiency distribution matrix, each frame of the observed image is reconstructed to obtain the photon distribution map corresponding to each frame of the observed image. Based on the photon distribution map corresponding to each frame of the observed image, multiple frames of the observed image are fused to determine the super-resolution image.

2. The quantum-efficiency-based image super-resolution reconstruction method as described in claim 1, characterized in that, Determining the number of received photons for each sub-region of each pixel in each detection period includes: Within one detection cycle, for the sub-region, the time range of the detection cycle and the regional spatial range of the sub-region are obtained; Based on the intensity distribution function of the X-ray standing wave field, a triple integral is performed on the spatial range of the sub-region and the time range to obtain the cumulative number of photons received by the sub-region within the detection period, and the cumulative number of photons is used as the number of received photons.

3. The quantum-efficiency-based image super-resolution reconstruction method as described in claim 1, characterized in that, The step of determining the quantum efficiency value of each sub-region of each pixel based on the calibration image corresponding to each detection period and the number of received photons of each sub-region of each pixel in each detection period includes: From the calibration images corresponding to each detection cycle, obtain the measured pixel value of each pixel in each detection cycle; Based on the measured pixel value of each pixel in each detection cycle and the number of received photons of each sub-region of each pixel in each detection cycle, the response model corresponding to each pixel is fitted to determine the response parameter value corresponding to each sub-region of each pixel in the response model corresponding to each pixel. The quantum efficiency value of each sub-region of each pixel is determined based on the response parameter values ​​corresponding to each sub-region of each pixel.

4. The quantum-efficiency-based image super-resolution reconstruction method as described in claim 3, characterized in that, The expression for the response model is: in, This represents the calibration image corresponding to one detection cycle. Line 1 The pixel value of the column pixel. These represent the row and column indices of the sub-region within a pixel, respectively. Indicates the total number of indexes. Indicates the first Within the first pixel The number of photons received in each sub-region Indicates the first Within the first pixel The offset coefficient of each sub-region Indicates the first Within the first pixel Linear response coefficients of each subregion Indicates the first Within the first pixel Nonlinear response coefficients of each subregion.

5. The quantum-efficiency-based image super-resolution reconstruction method as described in claim 4, characterized in that, The response parameter values ​​include the values ​​of the offset coefficient, the linear response coefficient, and the nonlinear response coefficient. Determining the quantum efficiency value of each sub-region of each pixel based on the response parameter values ​​corresponding to each sub-region of each pixel includes: Using the quantum efficiency calculation formula, based on the values ​​of the offset coefficient, the linear response coefficient, and the nonlinear response coefficient corresponding to each sub-region of each pixel, the quantum efficiency value of each sub-region of each pixel is determined, wherein the quantum efficiency calculation formula is: in Indicates the first Within the first pixel Quantum efficiency values ​​for each subregion.

6. The quantum-efficiency-based image super-resolution reconstruction method as described in claim 5, characterized in that, The process of reconstructing each frame of the observed image based on the quantum efficiency distribution matrix to obtain a super-resolution photon distribution map corresponding to each frame of the observed image includes: For the reconstruction process of any of the observed images, the photon distribution map is initialized to obtain an initial image. Based on the initial image, iteration is performed to obtain the current photon distribution map corresponding to the current iteration number. Based on the quantum efficiency distribution matrix and the current photon distribution map, the current predicted image is determined; Calculate the current reconstruction error between the observed image and the current predicted image. If the reconstruction error is greater than a preset error, continue to iteratively update the current photon distribution map until the current reconstruction error is less than the preset error. Then, use the photon distribution map after stopping the iteration as the photon distribution map corresponding to the observed image.

7. The quantum-efficiency-based image super-resolution reconstruction method as described in claim 1, characterized in that, Based on the photon distribution map corresponding to each frame of the observed image, the super-resolution image is determined by fusing multiple frames of the observed image, including: Image registration is performed on multiple frames of the observed images to determine the registered observed image corresponding to each frame of the observed images; The quality of each frame of the registered observation image is evaluated, and the quality weight corresponding to each frame of the registered observation image is determined. Based on the quality weights corresponding to the registered observation images in each frame, the photon distribution values ​​of each sub-region of each pixel in each frame of the registered observation images are fused to determine the fused pixel values ​​of each sub-region of each pixel in the super-resolution image.

8. The quantum-efficiency-based image super-resolution reconstruction method as described in claim 7, characterized in that, The method of fusing the photon distribution values ​​of each sub-region of each pixel in each frame of the registered observation image based on the fusion weights of each sub-region of each pixel and the quality weights corresponding to each frame of the registered observation image, and determining the fused pixel values ​​of each sub-region of each pixel in the super-resolution image, includes: For the sub-region of the pixel, the quality weights corresponding to each frame of the registration observation image and the photon distribution values ​​of the sub-region of the pixel in each frame of the registration observation image are substituted into the sub-region pixel value fusion formula to determine the fused pixel value of the sub-region of the pixel in the photon distribution map, wherein the sub-region pixel value fusion formula is: ,in Represents the first pixel in a super-resolution image. The fused pixel values ​​of the sub-regions Indicates the first frame, Indicates the total number of frames. Indicates the first Quality weights corresponding to the frame-registered observation images Indicates the first The first photon distribution map corresponding to the frame-registered observation image The photon distribution values ​​of each of the sub-regions.

9. The quantum-efficiency-based image super-resolution reconstruction method according to any one of claims 1 to 8, characterized in that, Before controlling the X-ray standing wave field to undergo multiple sub-pixel displacements centered on a reference position, the method further includes: The displacement platform of the X-ray source is controlled to move to a preset initial calibration position. A laser is emitted and a reflected signal is received by the laser interferometer in the reconstruction system. The actual coordinate data of the displacement platform at the preset initial calibration position is collected. The actual coordinate data is compared with the preset initial coordinates of the preset initial calibration position to obtain the displacement deviation value. Based on the displacement deviation value, the displacement platform is corrected in real time using a preset compensation algorithm, and the motion parameters of the displacement platform are dynamically adjusted.

10. A computing device, characterized in that, The system includes a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the quantum-efficiency-based image super-resolution reconstruction method as described in any one of claims 1 to 9.

Images