WO2025129804A1 - 图像恢复方法、装置、计算机设备和存储介质 - Google Patents

图像恢复方法、装置、计算机设备和存储介质 Download PDF

Info

Publication number
WO2025129804A1
WO2025129804A1 PCT/CN2024/075909 CN2024075909W WO2025129804A1 WO 2025129804 A1 WO2025129804 A1 WO 2025129804A1 CN 2024075909 W CN2024075909 W CN 2024075909W WO 2025129804 A1 WO2025129804 A1 WO 2025129804A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
detected
target
speckle image
objects
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/CN2024/075909
Other languages
English (en)
French (fr)
Inventor
何文奇
陈俊成
卢大江
彭翔
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.)
Shenzhen University
Original Assignee
Shenzhen University
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 Shenzhen University filed Critical Shenzhen University
Publication of WO2025129804A1 publication Critical patent/WO2025129804A1/zh
Anticipated expiration legal-status Critical
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/50—Image enhancement or restoration using two or more images, e.g. averaging or subtraction
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G06N3/0455—Auto-encoder networks; Encoder-decoder networks
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]

Definitions

  • the present application relates to the field of image processing technology, and in particular to an image restoration method, apparatus, computer equipment and storage medium.
  • Scattering imaging is an imaging method based on the principle of scattering. It achieves imaging by measuring the scattering behavior of the sample to the incident light. When the sample is irradiated with light, the reflection and scattering of the light by its components and structure affect the propagation of the light. After multiple scattering, these lights constitute the scattered light emitted by the sample.
  • the present application provides an image restoration method.
  • the method comprises:
  • the central autocorrelation data is processed by sidelobe recovery to obtain a target image containing each object to be detected.
  • determining the central autocorrelation data of the object contained in the target speckle image according to the target speckle image includes:
  • the target speckle image respectively obtaining a candidate object intensity map of each of the objects to be detected
  • correlation operations are performed on the candidate object intensity maps of the objects to be detected to obtain the central autocorrelation data.
  • determining the central autocorrelation data of the object contained in the target speckle image according to the target speckle image includes:
  • the sum of the first autocorrelation data and the second autocorrelation data is used as the central autocorrelation data.
  • acquiring a target speckle image containing at least two objects to be detected includes:
  • the initial speckle image is subjected to envelope correction processing to obtain the target speckle image.
  • performing envelope correction processing on the initial speckle image to obtain the target speckle image includes:
  • the initial speckle image is divided by the fitting surface to obtain the target speckle image.
  • the training process of the sidelobe recovery network includes:
  • the reference image is an image obtained by performing sidelobe recovery processing on sample autocorrelation data of the sample object by an initial recovery network
  • the standard image is an image containing sample sidelobes of the sample object
  • the parameters of the initial restoration network are adjusted to obtain the sidelobe restoration network.
  • the present application also provides an image restoration device.
  • the device comprises:
  • An acquisition module used for acquiring a target speckle image containing at least two objects to be detected
  • a determination module configured to determine central autocorrelation data of an object contained in the target speckle image according to the target speckle image
  • the present application further provides a computer-readable storage medium.
  • the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following steps are implemented:
  • the central autocorrelation data is processed by sidelobe recovery to obtain a target image containing each object to be detected.
  • the present application further provides a computer program product.
  • the computer program product includes a computer program, and when the computer program is executed by a processor, the following steps are implemented:
  • the central autocorrelation data is processed by sidelobe recovery to obtain a target image containing each object to be detected.
  • the above-mentioned image restoration method, device, computer equipment and storage medium through the target speckle image, realizes determining the central autocorrelation data of the object contained in the target speckle image, and then, based on the sidelobe recovery network, performs sidelobe recovery processing on the central autocorrelation data to obtain the target image containing each object to be detected; because in the above-mentioned process, the present application does not directly obtain the target image of the object to be detected based on the target speckle image; but by determining the central autocorrelation data of the object contained in the target speckle image, realizes the recovery of the complete amplitude information of each object to be detected, realizes the sidelobe recovery processing of each object to be detected, and ensures that the target image containing each object to be detected can be successfully obtained according to the sidelobe recovery network in the subsequent process; therefore, the present application can realize the recovery of the speckle autocorrelation sidelobe of each object to be detected through the central autocorrelation data, realizes the imaging of multiple different objects to
  • FIG1 is a schematic diagram of a flow chart of an image restoration method provided in an embodiment of the present application.
  • FIG2 is a schematic diagram of a process for obtaining an initial speckle image provided by an embodiment of the present application
  • FIG3 is a schematic diagram of the structure of a sidelobe recovery network provided in an embodiment of the present application.
  • FIG4 is a schematic diagram of a process for determining central autocorrelation data provided in an embodiment of the present application.
  • FIG5 is a schematic diagram of a process for determining a target speckle image according to an embodiment of the present application.
  • FIG6 is a flow chart of a sidelobe recovery network training process provided in an embodiment of the present application.
  • FIG7 is a schematic diagram of a flow chart of another image restoration method provided in an embodiment of the present application.
  • FIG8 is a structural block diagram of a first image restoration device provided in an embodiment of the present application.
  • FIG9 is a structural block diagram of a second image restoration device provided in an embodiment of the present application.
  • FIG10 is a structural block diagram of a third image restoration device provided in an embodiment of the present application.
  • FIG11 is a structural block diagram of a fourth image restoration device provided in an embodiment of the present application.
  • FIG12 is a structural block diagram of a fourth image restoration device provided in an embodiment of the present application.
  • FIG. 13 is a diagram showing the internal structure of a computer device in one embodiment.
  • an image recovery method is provided.
  • This embodiment uses the method applied to a terminal as an example for illustration. It is understandable that the method can also be applied to a server, and can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server.
  • the terminal can be but is not limited to various personal computers, laptops, smart phones, tablet computers, Internet of Things devices and portable wearable devices.
  • the Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc.
  • Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc.
  • the server can be implemented as an independent server or a server cluster consisting of multiple servers. In this embodiment, the method includes the following steps:
  • the target speckle image refers to the image obtained by performing envelope correction processing on the initial speckle image collected by the target camera; therefore, compared with the initial speckle image, the target speckle image reduces the occlusion of the speckle texture by the envelope, so that a more accurate target image can be obtained later based on the target speckle image.
  • the process of obtaining the initial speckle image may specifically include the following: the object is projected by a projector The object is then magnified by the lens and the magnified image is scattered by the scattering medium to form a scattered image. The scattered image is captured by the target camera to obtain the initial speckle image.
  • an aperture and an attenuation sheet may be added during the scattering imaging process, so that the propagation range of the scattered light can be limited by the aperture, the incident angle and the light intensity distribution of the scattered light can be controlled, and the scattering imaging effect can be improved; and the light intensity of different areas in the image of the object can be adjusted by the attenuation sheet, so that the light intensity distribution of each area of the image of the object is more uniform, and the scattering imaging effect can be further improved.
  • the present application in order to reduce the occlusion effect of the envelope on the speckle texture, may use Zernike polynomials to perform envelope correction processing on the initial speckle image to obtain a target speckle image after envelope correction.
  • S102 Determine central autocorrelation data of an object included in the target speckle image according to the target speckle image.
  • the present application determines the corresponding relationship of the OME range of each object to be detected through deep learning, expands the OME range of each object to be detected, and obtains central autocorrelation data, so that the target image containing each object to be detected can be successfully acquired based on the central autocorrelation data.
  • OME Optical Memory Effect
  • the field of view needs to be customized using the proposed method, which may specifically include the following contents: obtaining the candidate object intensity map of each object to be detected according to the target speckle image; performing correlation operation on the candidate object intensity map of each object to be detected to obtain central autocorrelation data.
  • the sidelobe recovery network can be a U-Net++ (densely linked neural network) network, which consists of an encoder and a decoder with dense jump connections, and the feature maps in the encoder and the decoder are fused through jump connections, thereby achieving the complementarity of high-resolution feature maps and low-resolution feature maps.
  • the sidelobe recovery network can be shown in Figure 3, where the sidelobe recovery network includes multiple x (i, j) modules, each of which includes a relu ( linear rectification function) activation function, a convolution layer and a batch normalization layer; and the number of channels of the sidelobe recovery network is completely symmetrically distributed.
  • the central autocorrelation data can be input to the input of the sidelobe recovery network terminal.
  • the output result of the output end of the sidelobe recovery network is obtained, and the output result is the target image containing each object to be detected.
  • the phase recovery algorithm can be used to perform phase recovery on each object to be detected contained in the target image, so as to reconstruct the shape of each object to be detected in the target image and the relative position of each object to be detected.
  • the above-mentioned image restoration method determines the central autocorrelation data of the object contained in the target speckle image through the target speckle image, and then performs sidelobe recovery processing on the central autocorrelation data based on the sidelobe recovery network to obtain the target image containing each object to be detected; because in the above-mentioned process, the present application does not directly obtain the target image of the object to be detected based on the target speckle image; but by determining the central autocorrelation data of the object contained in the target speckle image, the complete amplitude information of each object to be detected is restored, and the sidelobe recovery processing is performed on each object to be detected, which ensures that the target image containing each object to be detected can be successfully obtained according to the sidelobe recovery network in the subsequent process; therefore, the present application can realize the restoration of the speckle autocorrelation sidelobe of each object to be detected through the central autocorrelation data, realizes the imaging of multiple different objects to be detected through incident light scattering in one space, so that the scatter
  • the computer device of this embodiment can determine the central autocorrelation data of the object contained in the target speckle image according to the target speckle image in the manner shown in FIG. 4, which specifically includes the following steps:
  • S401 acquiring candidate object intensity maps of each object to be detected according to the target speckle image.
  • calculation formula (1) can be as follows:
  • the central autocorrelation data when the central autocorrelation data needs to be determined, the following may be specifically included: performing autocorrelation operations on the candidate object intensity maps of each object to be detected to obtain first autocorrelation data; performing cross-correlation operations on the candidate object intensity maps of each object to be detected to obtain second autocorrelation data; and combining the first autocorrelation data and the second autocorrelation data. The sum of is taken as the central autocorrelation data.
  • calculation formula (2) taking the scattering system including two objects to be detected as an example, the above content can be expressed as calculation formula (2), therefore, the central autocorrelation data can be obtained by substituting the two objects to be detected into calculation formula (2).
  • the calculation formula (2) is as follows:
  • O 1 refers to the first object to be detected in the target speckle image
  • O 2 refers to the second object to be detected in the target speckle image
  • r refers to the position of the image center in the target speckle image
  • ⁇ r 1 refers to the position relationship of the first object to be detected in the target speckle image
  • ⁇ r 2 refers to the position relationship of the second object to be detected in the target speckle image
  • the function representation of the target speckle image in order to ensure that when the target speckle image contains two objects to be detected, can be used to perform noise analysis and fitting approximation to obtain the calculation formula (2); wherein the function representation (3) of the target speckle image can be as follows:
  • O 1 (r- ⁇ r 1 ) refers to the first object to be detected
  • O 2 (r- ⁇ r 2 ) refers to the second object to be detected
  • psf 1 (r- ⁇ r 1 ) refers to the point spread function of the field of view where the first object to be detected is located
  • psf 2 (r- ⁇ r 2 ) refers to the point spread function of the second object to be detected
  • O 1 refers to the first object to be detected in the target speckle image
  • O 2 refers to the second object to be detected in the target speckle image
  • r refers to the position of the image center in the target speckle image
  • ⁇ r 1 refers to the position relationship of the first object to be detected in the target speckle image
  • ⁇ r 2 refers to the position relationship of the second object to be detected in the target speckle image.
  • O 1 (r- ⁇ r 1 ) refers to the candidate object intensity map of the first object to be detected
  • O 2 (r- ⁇ r 2 ) refers to the candidate object intensity map of the second object to be detected
  • psf 1 (r- ⁇ r 1 ) refers to the point spread function of the first object to be detected
  • psf 2 (r- ⁇ r 2 ) refers to the point spread function of the second object to be detected
  • O 1 refers to the first object to be detected in the target speckle image
  • O 2 refers to the second object to be detected in the target speckle image
  • r refers to the position of the image center in the target speckle image
  • ⁇ r 1 refers to the position relationship of the first object to be detected in the target speckle image
  • ⁇ r 2 refers to the position relationship of the second object to be detected in the target speckle image.
  • the point spread functions of the objects to be detected within the same OME range are highly correlated, and since the point spread functions of the objects to be detected are randomly distributed speckle patterns, the point spread functions of the objects to be detected within the same OME range can be approximated as pulse functions; while the point spread functions of the objects to be detected within different OME ranges have low correlation, so the point spread functions of the objects to be detected within different OME ranges can be approximated as background noise; the above content can be expressed by calculation formula (5):
  • psf i refers to the point spread function of the i-th object to be detected
  • psf j refers to the point spread function of the j-th object to be detected
  • r refers to the position of the image center in the target speckle image
  • ⁇ ri refers to the position relationship of the i-th object to be detected in the target speckle image
  • ⁇ r j refers to the position relationship of the j-th object to be detected in the target speckle image.
  • calculation formula (4) is simplified by the calculation formula (5), and the cross-correlation operation between the point spread functions of different objects to be detected is used as the background noise; the cross-correlation operation between the point spread functions of the same object to be detected is used as the pulse function.
  • the candidate calculation formula (6) is obtained, and the candidate calculation formula (6) is as follows:
  • O 1 refers to the first object to be detected in the target speckle image
  • O 2 refers to the second object to be detected in the target speckle image
  • r refers to the position of the image center in the target speckle image
  • ⁇ r 1 refers to the position relationship of the first object to be detected in the target speckle image
  • ⁇ r 2 refers to the position relationship of the second object to be detected in the target speckle image
  • C refers to the background noise in the target speckle image.
  • background noise is introduced into the candidate calculation formula (6), there are no autocorrelation side lobes of the objects to be detected in the candidate calculation formula (6). Therefore, the side lobe recovery processing is performed on each object to be detected, and the background noise C in the candidate calculation formula (6) can be used as the cross-correlation information of the annihilated objects in the center. Then, the background noise C corresponding to the object spacing of each object to be detected is learned and analyzed through deep learning to determine the corresponding relationship of the OME range of each object to be detected.
  • the process of determining the corresponding relationship can be regarded as fitting and approximating the point spread function of the two positions, so that psf 2 (x 2 , y 2 ) ⁇ psf 1 (x 1 , y 1 ), and then the calculation formula (2) is obtained. Furthermore, by adopting different object spacings The neural network is trained with samples to achieve customized image restoration operations for multi-scale multi-purpose isolated or continuous fields of view according to the object distance of each object to be detected.
  • the above-mentioned image restoration method by determining the candidate object intensity map of each object to be detected, and determining the central autocorrelation data according to the candidate object intensity map of each object to be detected, realizes the restoration of the complete amplitude information of each object to be detected, realizes the sidelobe recovery processing of each object to be detected, and ensures that the target image containing each object to be detected can be successfully obtained according to the sidelobe recovery network in the subsequent
  • the following components may be sequentially arranged between the projector and the target camera: an aperture, an attenuation plate, a lens, an attenuation plate, and a scattering medium, so that the target camera can capture and shoot the scattered image to obtain an initial speckle image.
  • an object is projected onto an aperture through a projector so that the aperture limits the propagation range of scattered light.
  • the scattered light passes through the aperture and reaches an attenuation plate.
  • the attenuation plate adjusts the light intensity of scattered light in different areas.
  • the scattered light passes through the attenuation plate and reaches a lens to form an image of the object.
  • the image of the object is amplified by the lens and reaches a scattering medium through the attenuation plate.
  • the scattering medium scatters the amplified image of the object to form a scattered image.
  • the scattered image is captured and photographed by a target camera to obtain an initial speckle image.
  • S502 Perform envelope correction processing on the initial speckle image to obtain a target speckle image.
  • the following contents may be specifically included: fitting the initial speckle image with Zernike polynomials to obtain a fitting surface; dividing the initial speckle image by the fitting surface to obtain a target speckle image.
  • the initial speckle image may be preprocessed to improve the accuracy of subsequent envelope correction processing of the initial speckle image.
  • the preprocessing may include but is not limited to: noise removal, image enhancement and other steps.
  • the Zernike polynomials are a set of basis functions describing the speckle shape
  • the Zernike polynomial coefficients can be obtained according to the speckle shape of the initial speckle image; further, the fitting surface can be obtained according to the Zernike polynomial coefficients; and the shape features described by the Zernike polynomials are subtracted from the initial speckle image by dividing the initial speckle image by the fitting surface, thereby obtaining the target speckle image.
  • the target speckle image can also be post-processed to achieve the purpose of improving the clarity and image quality of the target speckle image; wherein the post-processing may include but is not limited to: smoothing, sharpening Processing, etc.
  • the above-mentioned image restoration method reduces the occlusion effect of the envelope on the speckle texture by performing envelope correction processing on the initial speckle image, thereby ensuring that a target image containing each object to be detected can be successfully obtained based on the target speckle image.
  • the training process of the sidelobe recovery network may specifically include the following contents:
  • the reference image is an image obtained by performing sidelobe recovery processing on sample autocorrelation data of the sample object by the initial recovery network
  • the standard image is an image containing the sample sidelobe of the sample object.
  • the similarity comparison can be performed from three aspects: image brightness, image contrast and image structure.
  • SSIM(x, y) refers to the image similarity between the reference image and the standard image
  • ⁇ x and ⁇ y refer to the means of the measurement items of the reference image and the standard image (the measurement items can be image brightness, image contrast and image structure)
  • ⁇ x and ⁇ y refer to the standard deviations of the measurement items of the reference image and the standard image
  • ⁇ xy refers to the covariance of the measurement items of the reference image and the standard image
  • C 1 and C2 refer to constant terms.
  • the calculation formula (8) is as follows:
  • MAE(x, y) refers to the average error between the reference image and the standard image
  • Ii (x, y) refers to the pixel prediction value of the reference image and the standard image
  • I ⁇ i (x, y) refers to the pixel true value of the reference image and the standard image
  • n refers to the reference The sum of pixels between the image and the standard image.
  • the multiple n can be pre-set, and then, according to the average error amplitude between the n-fold reference image and the standard image and the image similarity between the reference image and the standard image, the loss function (9) is determined, and then, according to the loss function (9), the parameters of the initial recovery network are adjusted to obtain the sidelobe recovery network.
  • the loss function (9) is as follows:
  • the process of adjusting the parameters of the sidelobe recovery network it can be divided into at least two training stages, and different training stages use different learning rates. Then, according to the training stages with different learning rates, the sidelobe recovery network is subjected to multi-cycle parameter adjustment to obtain the trained sidelobe recovery network.
  • the parameter adjustment process of the sidelobe recovery network can be divided into three training stages, which are the first training stage, the second training stage and the third training stage, wherein the first training stage corresponds to the first ten rounds of iterative training of the sidelobe recovery network; the second training stage corresponds to the tenth to thirtieth rounds of iterative training of the sidelobe recovery network; the third training stage corresponds to the thirty-first to fiftieth rounds of iterative training of the sidelobe recovery network; and the learning rate of the first training stage can be 0.001; the learning rate of the second training stage can be 0.0003; and the learning rate of the third training stage can be 0.00003.
  • the sample objects can be divided into three groups, specifically: the distance between two objects in the first group of sample objects is 2mm; and the point spread functions of objects that do not belong to the same OME range in the first group are set to have no correlation; the distance between two objects in the second group of sample objects includes: 1mm, 2mm and 3mm; and the point spread functions of objects in the second group with a distance less than 2mm are set to have correlation; the point spread functions of objects with a distance greater than 1mm are set to have no correlation; the distance between two objects in the third group of sample objects is 2mm; and the three sample objects are distributed in a triangular shape, and a single object is within a memory effect range, and the point spread functions of the three objects at their respective locations are set to have no correlation.
  • the above image restoration method realizes parameter adjustment of the initial restoration network by determining the image similarity and the average error amplitude, and obtains the sidelobe restoration network. It ensures that the central autocorrelation data can be side-by-side restored according to the sidelobe restoration network. The lobe is restored to obtain a target image containing each object to be detected.
  • S704 Determine candidate object intensity maps of each object to be detected according to the target speckle image.
  • S707 Use the sum of the first autocorrelation data and the second autocorrelation data as central autocorrelation data.
  • the above-mentioned image restoration method determines the central autocorrelation data of the object contained in the target speckle image through the target speckle image, and then performs sidelobe recovery processing on the central autocorrelation data based on the sidelobe recovery network to obtain the target image containing each object to be detected; because in the above-mentioned process, the present application does not directly obtain the target image of the object to be detected based on the target speckle image; but by determining the central autocorrelation data of the object contained in the target speckle image, the complete amplitude information of each object to be detected is restored, and the sidelobe recovery processing is performed on each object to be detected, which ensures that the target image containing each object to be detected can be successfully obtained according to the sidelobe recovery network in the subsequent process; therefore, the present application can realize the restoration of the speckle autocorrelation sidelobe of each object to be detected through the central autocorrelation data, realizes the imaging of multiple different objects to be detected through incident light scattering in one space, so that the scatter
  • the present application also provides a method for realizing the above-mentioned image restoration method.
  • the implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations of one or more image restoration device embodiments provided below can refer to the limitations of the image restoration method above, and will not be repeated here.
  • an image restoration device comprising: an acquisition module 10, a determination module 20 and a restoration module 30, wherein:
  • the acquisition module 10 is used to acquire a target speckle image containing at least two objects to be detected.
  • the determination module 20 is used to determine the central autocorrelation data of the object contained in the target speckle image according to the target speckle image.
  • the recovery module 30 is used to perform sidelobe recovery processing on the central autocorrelation data based on a sidelobe recovery network to obtain a target image containing each object to be detected.
  • the training process of the sidelobe recovery network includes: obtaining a reference image and a standard image of a sample object; wherein the reference image is an image obtained by performing sidelobe recovery processing on sample autocorrelation data of the sample object by the initial recovery network, and the standard image is an image containing sample sidelobes of the sample object; performing a similarity comparison between the reference image and the standard image to obtain image similarity between the reference image and the standard image; determining an average error amplitude between the reference image and the standard image according to a reference pixel value of the reference image and a standard pixel value of the standard image; and adjusting parameters of the initial recovery network according to the image similarity and the average error amplitude to obtain a sidelobe recovery network.
  • a determination module 20 includes: a first determination unit 21 and a second determination unit 22, wherein:
  • the first determining unit 21 is used to respectively obtain a candidate object intensity map of each to-be-detected object according to the target speckle image.
  • the second determination unit 22 is used to perform a correlation operation on the candidate object intensity map of each object to be detected according to the object distance of each object to be detected, so as to obtain central autocorrelation data.
  • an image restoration device in which the second determination unit 22 includes: a first determination subunit 221, a second determination subunit 222 and a third determination subunit 223, wherein:
  • the first determining subunit 221 is used to perform autocorrelation operations on the candidate object intensity maps of each object to be detected to obtain first autocorrelation data.
  • the second determining subunit 222 is used to perform cross-correlation operations on the candidate object intensity maps of each object to be detected to obtain second autocorrelation data;
  • the third determining subunit 223 is configured to use the sum of the first autocorrelation data and the second autocorrelation data as central autocorrelation data.
  • an image restoration device in which an acquisition module 10 includes: an acquisition unit 11 and a correction unit 12, wherein:
  • the acquisition unit 11 is used to acquire an initial speckle image containing at least two objects to be detected.
  • the correction unit 12 is used to perform envelope correction processing on the initial speckle image to obtain a target speckle image.
  • an image restoration device in which a correction unit 12 includes a fitting subunit 121 and a processing subunit 122 , wherein:
  • the fitting subunit 121 is used to perform fitting processing on the initial speckle image using Zernike polynomials to obtain a fitting surface.
  • the processing subunit 122 is used to perform a division process on the initial speckle image and the fitting surface to obtain a target speckle image.
  • Each module in the above-mentioned image restoration device can be implemented in whole or in part by software, hardware or a combination thereof.
  • Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module.
  • a computer device which may be a terminal, and its internal structure diagram may be shown in FIG13.
  • the computer device includes a processor, a memory, an input/output interface, a communication interface, a display unit, and an input device.
  • the processor, the memory, and the input/output interface are connected via a system bus, and the communication interface, the display unit, and the input device are connected to the system bus via the input/output interface.
  • the processor of the computer device is used to provide computing and control capabilities.
  • the memory of the computer device includes a non-volatile storage medium and an internal memory.
  • the non-volatile storage medium stores an operating system and a computer program.
  • the internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium.
  • the input/output interface of the computer device is used to exchange information between the processor and an external device.
  • the communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner may be implemented through WIFI, a mobile cellular network, NFC (near field communication) or other technologies.
  • WIFI wireless a mobile cellular network
  • NFC near field communication
  • an image restoration method is implemented.
  • the display unit of the computer device is used to form a visually visible picture, which may be a display screen, a projection device, or a virtual reality imaging device.
  • the display screen can be a liquid crystal display screen or an electronic ink display screen
  • the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.
  • FIG. 13 is merely a block diagram of a partial structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied.
  • the specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.
  • a computer device including a memory and a processor, wherein the memory stores a computer program
  • the processor implements the following steps when executing the computer program:
  • the target speckle image determining the central autocorrelation data of the object contained in the target speckle image
  • the central autocorrelation data is processed for sidelobe recovery to obtain a target image containing each object to be detected.
  • the processor when the processor executes the computer program, the processor further implements the following steps:
  • the target speckle image respectively obtain the candidate object intensity map of each object to be detected
  • correlation operations are performed on the candidate object intensity maps of the objects to be detected to obtain central autocorrelation data.
  • the processor when the processor executes the computer program, the processor further implements the following steps:
  • the sum of the first autocorrelation data and the second autocorrelation data is taken as the central autocorrelation data.
  • the processor when the processor executes the computer program, the processor further implements the following steps:
  • the initial speckle image is subjected to envelope correction to obtain the target speckle image.
  • the processor when the processor executes the computer program, the processor further implements the following steps:
  • the Zernike polynomials are used to fit the initial speckle image to obtain a fitting surface
  • the initial speckle image is divided by the fitting surface to obtain the target speckle image.
  • the processor when the processor executes the computer program, the processor further implements the following steps:
  • the reference image is an image obtained by performing sidelobe recovery processing on sample autocorrelation data of the sample object by the initial recovery network
  • the standard image is an image containing sample sidelobes of the sample object
  • the parameters of the initial restoration network are adjusted to obtain the sidelobe restoration network.
  • a computer readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:
  • the target speckle image determining the central autocorrelation data of the object contained in the target speckle image
  • the central autocorrelation data is processed for sidelobe recovery to obtain a target image containing each object to be detected.
  • the target speckle image respectively obtain the candidate object intensity map of each object to be detected
  • correlation operations are performed on the objects to be detected to obtain autocorrelation data.
  • the sum of the first autocorrelation data and the second autocorrelation data is taken as the central autocorrelation data.
  • the initial speckle image is subjected to envelope correction to obtain the target speckle image.
  • the Zernike polynomials are used to fit the initial speckle image to obtain a fitting surface
  • the initial speckle image is divided by the fitting surface to obtain the target speckle image.
  • the reference image is an image obtained by performing sidelobe recovery processing on sample autocorrelation data of the sample object by the initial recovery network
  • the standard image is an image containing sample sidelobes of the sample object
  • the parameters of the initial restoration network are adjusted to obtain the sidelobe restoration network.
  • a computer program product comprising a computer program, which, when executed by a processor, implements the following steps:
  • the target speckle image determining the central autocorrelation data of the object contained in the target speckle image
  • the central autocorrelation data is processed for sidelobe recovery to obtain a target image containing each object to be detected.
  • the target speckle image respectively obtain the candidate object intensity map of each object to be detected
  • correlation operations are performed on the candidate object intensity maps of the objects to be detected to obtain central autocorrelation data.
  • the sum of the first autocorrelation data and the second autocorrelation data is taken as the central autocorrelation data.
  • the initial speckle image is subjected to envelope correction to obtain the target speckle image.
  • the Zernike polynomials are used to fit the initial speckle image to obtain a fitting surface
  • the initial speckle image is divided by the fitting surface to obtain the target speckle image.
  • the reference image is an image obtained by performing sidelobe recovery processing on sample autocorrelation data of the sample object by the initial recovery network
  • the standard image is an image containing sample sidelobes of the sample object
  • the parameters of the initial restoration network are adjusted to obtain the sidelobe restoration network.
  • user information including but not limited to user device information, user personal information, etc.
  • data including but not limited to data used for analysis, stored data, displayed data, etc.
  • any reference to the memory, database or other media used in the embodiments provided in this application can include non-volatile and volatile memory. At least one of volatile memories.
  • Non-volatile memories may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc.
  • Volatile memory may include random access memory (RAM) or external cache memory, etc.
  • RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).
  • SRAM static random access memory
  • DRAM dynamic random access memory
  • the database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database.
  • Non-relational databases may include distributed databases based on blockchains, etc., but are not limited thereto.
  • the processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, etc., but is not limited thereto.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Image Processing (AREA)

Abstract

本申请涉及一种图像恢复方法、装置、计算机设备和存储介质，涉及图像处理技术领域。方法包括：获取包含至少两个待检测物体的目标散斑图像；根据目标散斑图像，确定目标散斑图像中所包含物体的中心自相关数据；基于旁瓣恢复网络，对中心自相关数据进行旁瓣恢复处理，得到包含各待检测物体的目标图像。本申请可实现通过中心自相关数据对各待检测物体的散斑自相关旁瓣进行恢复，实现了定制化视场的散射成像处理，使得散射成像技术能够应用在多个待检测物体的成像应用场景，保证了针对多个待检测物体的散射成像准确性。

Description

图像恢复方法、装置、计算机设备和存储介质 技术领域
本申请涉及图像处理技术领域，特别是涉及一种图像恢复方法、装置、计算机设备和存储介质。
背景技术
散射成像技术是一种基于散射原理的成像方法，通过测量样品对入射光的散射行为来实现成像。当样品受到光辐射时，其成分和结构对光线的反射和散射影响光线的传播。经过多次散射后，这些光线构成了样品发出的散射光。
但是，现有的散射成像技术无法在一个空间内通过入射光散射成像多个不同的目标物体，进而，限制了散射成像技术的应用场景。
发明内容
基于此，有必要针对上述技术问题，提供一种能够在一个空间内通过入射光散射成像多个不同的目标物体的图像恢复方法、装置、计算机设备和存储介质。
第一方面，本申请提供了一种图像恢复方法。该方法包括：
获取包含至少两个待检测物体的目标散斑图像；
根据所述目标散斑图像，确定所述目标散斑图像中所包含物体的中心自相关数据；
基于旁瓣恢复网络，对所述中心自相关数据进行旁瓣恢复处理，得到包含各待检测物体的目标图像。
在其中一个实施例中，所述根据所述目标散斑图像，确定所述目标散斑图像中所包含物体的中心自相关数据，包括：
根据所述目标散斑图像，分别获取各所述待检测物体的候选物体强度图；
根据各所述待检测物体的物体间距，对各所述待检测物体的候选物体强度图进行相关运算，得到所述中心自相关数据。
在其中一个实施例中，所述根据所述目标散斑图像，确定所述目标散斑图像中所包含物体的中心自相关数据，包括：
分别对各所述待检测物体的候选物体强度图进行自相关运算，得到第一自相关数据；
分别对各所述待检测物体的候选物体强度图进行互相关运算，得到第二自相关数据；
将所述第一自相关数据和所述第二自相关数据的和值，作为所述中心自相关数据。
在其中一个实施例中，所述获取包含至少两个待检测物体的目标散斑图像，包括：
获取包含至少两个待检测物体的初始散斑图像；
对所述初始散斑图像进行包络校正处理，得到所述目标散斑图像。
在其中一个实施例中，所述对所述初始散斑图像进行包络校正处理，得到所述目标散斑图像，包括：
采用泽尼克多项式对所述初始散斑图像进行拟合处理，得到拟合曲面；
将所述初始散斑图像与所述拟合曲面进行相除处理，得到所述目标散斑图像。
在其中一个实施例中，所述旁瓣恢复网络的训练过程，包括：
获取样本物体的参考图像和标准图像；其中，所述参考图像为初始恢复网络对所述样本物体的样本自相关数据进行旁瓣恢复处理得到的图像，所述标准图像为包含所述样本物体的样本旁瓣的图像；
对所述参考图像与所述标准图像进行相似度比较，得到所述参考图像与所述标准图像的图像相似度；
根据所述参考图像的参考像素值和所述标准图像的标准像素值，确定所述参考图像与所述标准图像的平均误差幅度；
根据所述图像相似度和所述平均误差幅度，对所述初始恢复网络进行参数调整，得到所述旁瓣恢复网络。
第二方面，本申请还提供了一种图像恢复装置。该装置包括：
获取模块，用于获取包含至少两个待检测物体的目标散斑图像；
确定模块，用于根据所述目标散斑图像，确定所述目标散斑图像中所包含物体的中心自相关数据；
恢复模块，用于基于旁瓣恢复网络，对所述中心自相关数据进行旁瓣恢复处理，得到包含各待检测物体的目标图像。
第三方面，本申请还提供了一种计算机设备。所述计算机设备包括存储器和处理器，所述存储器存储有计算机程序，所述处理器执行所述计算机程序时实现以下步骤：
获取包含至少两个待检测物体的目标散斑图像；
根据所述目标散斑图像，确定所述目标散斑图像中所包含物体的中心自相关数据；
基于旁瓣恢复网络，对所述中心自相关数据进行旁瓣恢复处理，得到包含各待检测物体的目标图像。
第四方面，本申请还提供了一种计算机可读存储介质。所述计算机可读存储介质，其上存储有计算机程序，所述计算机程序被处理器执行时实现以下步骤：
获取包含至少两个待检测物体的目标散斑图像；
根据所述目标散斑图像，确定所述目标散斑图像中所包含物体的中心自相关数据；
基于旁瓣恢复网络，对所述中心自相关数据进行旁瓣恢复处理，得到包含各待检测物体的目标图像。
第五方面，本申请还提供了一种计算机程序产品。所述计算机程序产品，包括计算机程序，该计算机程序被处理器执行时实现以下步骤：
获取包含至少两个待检测物体的目标散斑图像；
根据所述目标散斑图像，确定所述目标散斑图像中所包含物体的中心自相关数据；
基于旁瓣恢复网络，对所述中心自相关数据进行旁瓣恢复处理，得到包含各待检测物体的目标图像。
上述图像恢复方法、装置、计算机设备和存储介质，通过目标散斑图像，实现确定目标散斑图像中所包含物体的中心自相关数据，进而，基于旁瓣恢复网络，对中心自相关数据进行旁瓣恢复处理，得到包含各待检测物体的目标图像；由于上述过程中，本申请并不是直接根据目标散斑图像，得到待检测物体的目标图像；而是通过确定目标散斑图像中所包含物体的中心自相关数据，实现对各待检测物体的完整振幅信息进行恢复，实现对各待检测物体进行旁瓣恢复处理，保证了后续能够顺利根据旁瓣恢复网络得到包含各待检测物体的目标图像；因此，本申请可实现通过中心自相关数据对各待检测物体的散斑自相关旁瓣进行恢复，实现了在一个空间内通过入射光散射成像多个不同的待检测物体，使得散射成像技术能够应用在多个待检测物体的成像应用场景，保证了针对多个待检测物体的散射成像准确性。
附图说明
图1为本申请实施例提供的一种图像恢复方法的流程示意图；
图2为本申请实施例提供的获取初始散斑图像的流程示意图；
图3为本申请实施例提供的旁瓣恢复网络的结构示意图；
图4为本申请实施例提供的确定中心自相关数据的流程示意图；
图5为本申请实施例提供的确定目标散斑图像的流程示意图；
图6为本申请实施例提供的旁瓣恢复网络训练过程的流程示意图；
图7为本申请实施例提供的另一种图像恢复方法的流程示意图；
图8为本申请实施例提供的第一种图像恢复装置的结构框图；
图9为本申请实施例提供的第二种图像恢复装置的结构框图；
图10为本申请实施例提供的第三种图像恢复装置的结构框图；
图11为本申请实施例提供的第四种图像恢复装置的结构框图；
图12为本申请实施例提供的第四种图像恢复装置的结构框图；
图13为一个实施例中计算机设备的内部结构图。
具体实施方式
为了使本申请的目的、技术方案及优点更加清楚明白，以下结合附图及实施例，对本申请进行进一步详细说明。应当理解，此处描述的具体实施例仅仅用以解释本申请，并不用于限定本申请。
应当理解，此处描述的具体实施例仅仅用以解释本申请，并不用于限定本申请。在本申请的描述中，参考术语“一个实施例”、“一些实施例”、“示例”、“具体示例”、或“一些示例”等的描述意指结合该实施例或示例描述的具体特征、结构、材料或者特点包含于本申请的至少一个实施例或示例中。在本说明书中，对上述术语的示意性表述不必须针对的是相同的实施例或示例。而且，描述的具体特征、结构、材料或者特点可以在任一个或多个实施例或示例中以合适的方式结合。此外，在不相互矛盾的情况下，本领域的技术人员可以将本说明书中描述的不同实施例或示例以及不同实施例或示例的特征进行结合和组合。
在一个实施例中，如图1所示，提供了一种图像恢复方法，本实施例以该方法应用于终端进行举例说明，可以理解的是，该方法也可以应用于服务器，还可以应用于包括终端和服务器的系统，并通过终端和服务器的交互实现。其中，终端可以但不限于是各种个人计算机、笔记本电脑、智能手机、平板电脑、物联网设备和便携式可穿戴设备，物联网设备可为智能音箱、智能电视、智能空调、智能车载设备等。便携式可穿戴设备可为智能手表、智能手环、头戴设备等。服务器可以用独立的服务器或者是多个服务器组成的服务器集群来实现。本实施例中，该方法包括以下步骤：
S101，获取包含至少两个待检测物体的目标散斑图像。
需要说明的是，目标散斑图像指的是将目标相机采集到的初始散斑图像进行包络校正处理后的图像；因此，目标散斑图像相比起初始散斑图像，减少了包络对散斑纹理的遮挡，使得后续能够根据目标散斑图像得到更加准确的目标图像。
进一步说明，初始散斑图像的获取过程，具体可包括以下内容：物体通过投影仪投射 到透镜上，形成物体的像；通过透镜将物体的像进行放大处理，并通过散射介质对放大后物体的像进行散射处理，形成散射像；通过目标相机对散射像进行捕捉拍摄，得到初始散斑图像。
在本申请的一种实施例中，为进一步提高初始散斑图像的散射成像效果，如图2所示，可在散射成像过程中加入光阑和衰减片，以通过光阑能够限制散射光线的传播范围，控制散射光线的入射角度和光强分布，提高散射成像的效果；并且，通过衰减片对物体的像中不同区域的光线强度进行调整，使得物体的像各区域的光强分布更加均匀，进一步提高散射成像的效果。
在本申请的另一种实施例中，为降低包络对散斑纹理的遮挡影响，本申请可采用泽尼克多项式对初始散斑图像对初始散斑图像进行包络校正处理，得到包络校正后的目标散斑图像。
S102，根据目标散斑图像，确定目标散斑图像中所包含物体的中心自相关数据。
由于，在对目标散斑图像进行旁瓣恢复处理时，会收到目标散斑图像中包含各待检测物体的OME(Optical Memory Effect，光学记忆效应)范围大小的影响，并且，目标散斑图像中的噪声也会对旁瓣恢复的准确性造成影响，影响恢复效果；因此，本申请通过深度学习，确定各待检测物体的OME范围的对应关系，实现对各待检测物体的OME范围进行扩展，得到中心自相关数据，使得后续能够根据中心自相关数据，顺利获取包含各待检测物体的目标图像。
具体的，当待检测物体所处的视场超过OME范围时，需要用所提方法定制视场时，具体可包括以下内容：根据目标散斑图像，分别获取各待检测物体的候选物体强度图；对各待检测物体的候选物体强度图进行相关运算，得到中心自相关数据。
S103，基于旁瓣恢复网络，对中心自相关数据进行旁瓣恢复处理，得到包含各待检测物体的目标图像。
需要说明的是，旁瓣恢复网络可以为U-Net++(稠密链接神经网络)网络，旁瓣恢复网络由密集跳跃连接的编码器和解码器两部分组成，并通过跳跃连接将编码器和解码器中的特征图进行融合，从而实现了高分辨率特征图和低分辨率特征图的互补。旁瓣恢复网络可如图3所示，其中，旁瓣恢复网络中包含多个x(i,j)模块，各x(i,j)模块中均包含relu(线性整流函数)激活函数，卷积层和批量归一化层；且旁瓣恢复网络的通道数呈完全对称分布。
在本申请的一种实施例中，可通过将中心自相关数据输入至旁瓣恢复网络终端的输入 端，得到旁瓣恢复网络的输出端输出的输出结果，该输出结果即为包含各待检测物体的目标图像。
进一步说明，由于目标图像中包含各待检测物体的完整傅里叶振幅信息，因此，可通过相位恢复算法对目标图像中包含的各待检测物体进行相位恢复，实现重构目标图像中各待检测物体的形状，以及各待检测物体的相对位置。
上述图像恢复方法，通过目标散斑图像，实现确定目标散斑图像中所包含物体的中心自相关数据，进而，基于旁瓣恢复网络，对中心自相关数据进行旁瓣恢复处理，得到包含各待检测物体的目标图像；由于上述过程中，本申请并不是直接根据目标散斑图像，得到待检测物体的目标图像；而是通过确定目标散斑图像中所包含物体的中心自相关数据，实现对各待检测物体的完整振幅信息进行恢复，实现对各待检测物体进行旁瓣恢复处理，保证了后续能够顺利根据旁瓣恢复网络得到包含各待检测物体的目标图像；因此，本申请可实现通过中心自相关数据对各待检测物体的散斑自相关旁瓣进行恢复，实现了在一个空间内通过入射光散射成像多个不同的待检测物体，使得散射成像技术能够应用在多个待检测物体的成像应用场景，保证了针对多个待检测物体的散射成像准确性。
在一种实施例中，由于现有的散射成像技术无法在一个空间内通过入射光散射成像多个不同的目标物体，对散射成像技术的应用场景造成了限制，因此，为解决上述技术问题，本实施例的计算机设备可通过如图4所示的方式，根据目标散斑图像，确定目标散斑图像中所包含物体的中心自相关数据，具体包括如下步骤：
S401，根据目标散斑图像，分别获取各待检测物体的候选物体强度图。
需要说明的是，可用计算公式(1)表征各待检测物体的候选物体强度图，具体的，计算公式(1)可如下所示：
候选物体强度图＝On(r-△rn)    (1)
其中，On指的是目标散斑图像中第n个待检测物体；r指的是目标散斑图像中图像中心的位置；△rn指的是目标散斑图像中第n个待检测物体的位置关系。
S402，根据各待检测物体的物体间距，对各待检测物体进行相关运算，得到中心自相关数据。
需要说明的是，当需要确定中心自相关数据时，具体可包括以下内容：分别对各待检测物体的候选物体强度图进行自相关运算，得到第一自相关数据；分别对各待检测物体的候选物体强度图进行互相关运算，得到第二自相关数据；将第一自相关数据和第二自相关数据 的和值，作为中心自相关数据。
在本申请的一种实施例中，以散射系统中中包含两个待检测物体为例，根据上述内容可表示为计算公式(2)，因此，可通过将两个待检测物体代入计算公式(2)中，得到中心自相关数据。其中，计算公式(2)如下所示：
其中，和指的是各待检测物体进行自相关运算得到的第一自相关数据；和指的是各待检测物体进行互相关运算得到的第二自相关数据。O1指的是目标散斑图像中第1个待检测物体；O2指的是目标散斑图像中第2个待检测物体；r指的是目标散斑图像中图像中心的位置；△r1指的是目标散斑图像中第1个待检测物体的位置关系；△r2指的是目标散斑图像中第2个待检测物体的位置关系；指的是互相关运算。
在本申请的一种实施例中，为保证能够当目标散斑图像中包含两个待检测物体时，可通过对目标散斑图像的函数表示，进行噪声分析和拟合逼近，得到计算公式(2)；其中，目标散斑图像的函数表示(3)可如下所示：
I(r)＝O1(r-Δr1)*psf1(r-Δr1)+O2(r-Δr2)*psf2(r-Δr2)      (3)
其中，O1(r-Δr1)指的是第一个待检测物体；O2(r-Δr2)指的是第二个待检测物体；psf1(r-Δr1)指的是第一个待检测物体所在视场的点扩散函数；psf2(r-Δr2)指的是第二个待检测物体的点扩散函数；O1指的是目标散斑图像中第1个待检测物体；O2指的是目标散斑图像中第2个待检测物体；r指的是目标散斑图像中图像中心的位置；△r1指的是目标散斑图像中第1个待检测物体的位置关系；△r2指的是目标散斑图像中第2个待检测物体的位置关系。
进一步的，对目标散斑图像进行自相关运算，得到计算公式(4)
其中，O1(r-Δr1)指的是第一个待检测物体的候选物体强度图；O2(r-Δr2)指的是第二个待检测物体的候选物体强度图；psf1(r-Δr1)指的是第一个待检测物体的点扩散函数； psf2(r-Δr2)指的是第二个待检测物体的点扩散函数；O1指的是目标散斑图像中第1个待检测物体；O2指的是目标散斑图像中第2个待检测物体；r指的是目标散斑图像中图像中心的位置；△r1指的是目标散斑图像中第1个待检测物体的位置关系；△r2指的是目标散斑图像中第2个待检测物体的位置关系。
进一步的，根据光学记忆效应可知，同一OME范围内待检测物体的点扩散函数具有高度相关性，并且，由于待检测物体的点扩散函数是随机分布的散斑图，因此，可将同一OME范围内待检测物体的点扩散函数近似作为脉冲函数；而不同OME范围内待检测物体的点扩散函数的相关性较低，因此，将不同OME范围内待检测物体的点扩散函数近似作为背景噪声；上述内容可用计算公式(5)表示：
其中，psfi(r-Δri)指的是第i个待检测物体的点扩散函数；psfj(r-Δrj)指的是第j个待检测物体的点扩散函数；r指的是目标散斑图像中图像中心的位置；△ri指的是目标散斑图像中第i个待检测物体的位置关系；△rj指的是目标散斑图像中第j个待检测物体的位置关系。
进而，通过计算公式(5)对计算公式(4)进行简化处理，将不同待检测物体的点扩散函数之间的互相关运算作为背景噪声；将同一待检测物体的点扩散函数之间的互相关运算作为脉冲函数。得到候选计算公式(6)，候选计算公式(6)如下所示：
其中，和指的是各待检测物体的候选物体强度图进行自相关运算得到的第一自相关数据；O1指的是目标散斑图像中第1个待检测物体；O2指的是目标散斑图像中第2个待检测物体；r指的是目标散斑图像中图像中心的位置；△r1指的是目标散斑图像中第1个待检测物体的位置关系；△r2指的是目标散斑图像中第2个待检测物体的位置关系；指的是互相关运算；C指的是目标散斑图像中的背景噪声。
进一步的，由于候选计算公式(6)中引入了背景噪声，此时候选计算公式(6)中不存在各待检测对象的自相关旁瓣，因此，对各待检测物体进行旁瓣恢复处理，可将候选计算公式(6)中背景噪声C作为中心各待检测物体被湮灭的互相关项信息，进而，通过深度学习对各待检测物体的物体间距对应的背景噪声C进行学习分析，确定各待检测物体的OME范围的对应关系，确定对应关系的过程可看作对两个位置的点扩散函数进行拟合逼近，使得psf2(x2，y2)≈psf1(x1，y1)，进而得到计算公式(2)。进一步的，可通过采用不同物体间距 的样本对神经网络进行训练，以实现根据各待检测物体的物体间距，实现针对多尺度多数目的孤立或连续的视场进行定制化图像恢复操作。
上述图像恢复方法，通过确定各待检测物体的候选物体强度图，并根据各待检测物体的候选物体强度图，确定中心自相关数据，实现对各待检测物体的完整振幅信息进行恢复，实现对各待检测物体进行旁瓣恢复处理，保证了后续能够顺利根据旁瓣恢复网络得到包含各待检测物体的目标图像
在一种实施例中，如图5所示，当需要获取包含至少两个待检测物体的目标散斑图像时，具体可包括以下内容：
S501，获取包含至少两个待检测物体的初始散斑图像。
在本申请的一种实施例中，可通过在投影仪与目标相机之间依次设置：光阑、衰减片、透镜、衰减片、散射介质，实现通过目标相机对散射像进行捕捉拍摄，得到初始散斑图像。
具体的，物体通过投影仪投射到光阑上，以使光阑限制散射光线的传播范围，散射光线透过光阑达到衰减片，衰减片对不同区域的散射光线的光线强度进行调整，散射光线通过衰减片达到透镜，形成物体的像；通过透镜将物体的像进行放大处理，并在此通过衰减片达到散射介质，散射介质对放大后物体的像进行散射处理，形成散射像；通过目标相机对散射像进行捕捉拍摄，得到初始散斑图像。
S502，对初始散斑图像进行包络校正处理，得到目标散斑图像。
需要说明的是，当需要对初始散斑图像进行包络校正处理时，具体可包括以下内容：采用泽尼克多项式对初始散斑图像进行拟合处理，得到拟合曲面；将初始散斑图像与拟合曲面进行相除处理，得到目标散斑图像。
进一步说明，在对初始散斑图像进行包络校正处理之前，可通过对初始散斑图像进行预处理，已达到提高后续对初始散斑图像进行包络校正处理准确信的效果。其中，预处理可以包括但不限于：噪声去除、图像增强等步骤。
在本申请的一种实施例中，由于泽尼克多项式是一组描述散斑形状的基函数，因此，可根据初始散斑图像的散斑形状，得到泽尼克多项式系数；进而，根据泽尼克多项式系数，得到拟合曲面；通过将初始散斑图像与拟合曲面进行相除处理，得到从初始散斑图像中减去由泽尼克多项式描述的形状特征的目的，从而得到目标散斑图像。
进一步说明，在得到目标散斑图像后，还可对目标散斑图像进行后处理，已达到提高目标散斑图像清晰度和图像质量的目的；其中，后处理可以包括但不限于：平滑处理、锐化 处理等。
上述图像恢复方法，通过对初始散斑图像进行包络校正处理，降低了降低包络对散斑纹理的遮挡影响，保证了后续能够顺利根据目标散斑图像，得到包含各待检测物体的目标图像。
在一种实施例中，如图6所示，旁瓣恢复网络的训练过程具体可包括以下内容：
S601，获取样本物体的参考图像和标准图像。
其中，参考图像为初始恢复网络对样本物体的样本自相关数据进行旁瓣恢复处理得到的图像，标准图像为包含样本物体的样本旁瓣的图像。
S602，对参考图像与标准图像进行相似度比较，得到参考图像与标准图像的图像相似度。
需要说明的是，在对参考图像与标准图像进行相似度比较时，可从图像亮度、图像对比度和图像结构，三个部分对参考图像与标准图像进行相似度比较。
进一步说明，可通过计算公式(7)对参考图像与标准图像进行相似度比较；计算公式(7)如下所示：
其中，SSIM(x，y)指的是参考图像与标准图像的图像相似度；μx和μy指的是参考图像与标准图像的测量项(测量项可以为图像亮度、图像对比度和图像结构)均值；σx和σy指的是参考图像与标准图像的测量项的标准差；σxy指的是参考图像与标准图像的测量项的协方差；C1和C2指的是常数项。
S603，根据参考图像的参考像素值和标准图像的标准像素值，确定参考图像与标准图像的平均误差幅度。
需要说明的是，可将参考图像的参考像素值和标准图像的标准像素值代入计算公式(8)，进而，通过计算公式(8)确定参考图像与标准图像的平均误差幅度，计算公式(8)如下所示：
其中，MAE(x，y)指的是参考图像与标准图像的平均误差幅度；Ii(x，y)指的是参考图像与标准图像的像素预测值；I`i(x，y)指的是参考图像与标准图像的像素真实值；n指的是参考 图像与标准图像的像素总和。
S604，根据图像相似度和平均误差幅度，对初始恢复网络进行参数调整，得到旁瓣恢复网络。
需要说明的是，由于参考图像与标准图像的平均误差幅度的取值较小，为更好的对初始恢复网络进行参数调整，可预先设定倍数n，进而，根据n倍的参考图像与标准图像的平均误差幅度和参考图像与标准图像的图像相似度，确定损失函数(9)，进而，根据损失函数(9)对初始恢复网络进行参数调整，得到旁瓣恢复网络。
具体的，损失函数(9)如下所示：
LOSS＝MAE*n+(1-SSIM)    (9)
其中，n指的是预先设定的倍数(例如，n可以取10)；MAE指的是参考图像与标准图像的平均误差幅度；SSIM指的是参考图像与标准图像的图像相似度。
进一步说明，在对旁瓣恢复网络进行参数调整的过程中，可划分为至少两个训练阶段，不同训练阶段采用不同的学习速率，进而，根据不同学习速率的训练阶段，对旁瓣恢复网络进行多循环参数调整，得到训练后的旁瓣恢复网络。
举例说明，例如可将旁瓣恢复网络进行参数调整过程划分为三个训练阶段，三个训练阶段分别为第一训练阶段、第二训练阶段和第三训练阶段，其中，第一训练阶段对应旁瓣恢复网络的前十轮迭代训练；第二训练阶段对应旁瓣恢复网络的第十轮迭代训练至第三十轮迭代训练；第三训练阶段对应旁瓣恢复网络的第三十一轮迭代训练至第五十轮迭代训练；并且，第一训练阶段的学习速率可以为0.001；第二训练阶段的学习速率可以为0.0003；第三训练阶段的学习速率可以为0.00003。
举例说明，当需要根据样本物品对旁瓣恢复网络进行训练时，可将样本物体可划分为三组，具体的：第一组样本物体中两个物体之间的间距为2mm；并且，设定第一组中不属于同一个OME范围的物体的点扩散函数不存在相关性；第二组样本物体中两个物体之间的间距包含：1mm、2mm和3mm；并且，设定第二组间距小于2mm物体的点扩散函数存在相关性；间距大于1mm物体的点扩散函数不存在相关性；第三组样本物体中两个物体之间的间距为2mm；并且，三个样本物体之间程序三角形分布，并且，单个物体处于一个记忆效应范围内，三个物体各自所在位置的点扩散函数不存在相关性。
上述图像恢复方法，通过确定图像相似度和平均误差幅度，实现了对初始恢复网络进行参数调整，得到旁瓣恢复网络。保证了能够根据旁瓣恢复网络，对中心自相关数据进行旁 瓣恢复处理，得到包含各待检测物体的目标图像。
在一种实施例中，如图7所示，当需要确定包含各待检测物体的目标图像时，具体可包括以下内容：
S701，获取包含至少两个待检测物体的初始散斑图像。
S702，采用泽尼克多项式对初始散斑图像进行拟合处理，得到拟合曲面。
S703，将初始散斑图像与拟合曲面进行相除处理，得到目标散斑图像。
S704，根据目标散斑图像，分别确定各待检测物体的候选物体强度图。
S705，分别对各待检测物体的候选物体强度图进行自相关运算，得到第一自相关数据。
S706，分别对各待检测物体的候选物体强度图进行互相关运算，得到第二自相关数据。
S707，将第一自相关数据和第二自相关数据的和值，作为中心自相关数据。
S708，基于旁瓣恢复网络，对中心自相关数据进行旁瓣恢复处理，得到包含各待检测物体的目标图像。
上述图像恢复方法，通过目标散斑图像，实现确定目标散斑图像中所包含物体的中心自相关数据，进而，基于旁瓣恢复网络，对中心自相关数据进行旁瓣恢复处理，得到包含各待检测物体的目标图像；由于上述过程中，本申请并不是直接根据目标散斑图像，得到待检测物体的目标图像；而是通过确定目标散斑图像中所包含物体的中心自相关数据，实现对各待检测物体的完整振幅信息进行恢复，实现对各待检测物体进行旁瓣恢复处理，保证了后续能够顺利根据旁瓣恢复网络得到包含各待检测物体的目标图像；因此，本申请可实现通过中心自相关数据对各待检测物体的散斑自相关旁瓣进行恢复，实现了在一个空间内通过入射光散射成像多个不同的待检测物体，使得散射成像技术能够应用在多个待检测物体的成像应用场景，保证了针对多个待检测物体的散射成像准确性。
应该理解的是，虽然如上的各实施例所涉及的流程图中的各个步骤按照箭头的指示依次显示，但是这些步骤并不是必然按照箭头指示的顺序依次执行。除非本文中有明确的说明，这些步骤的执行并没有严格的顺序限制，这些步骤可以以其它的顺序执行。而且，如上的各实施例所涉及的流程图中的至少一部分步骤可以包括多个步骤或者多个阶段，这些步骤或者阶段并不必然是在同一时刻执行完成，而是可以在不同的时刻执行，这些步骤或者阶段的执行顺序也不必然是依次进行，而是可以与其它步骤或者其它步骤中的步骤或者阶段的至少一部分轮流或者交替地执行。
基于同样的发明构思，本申请实施例还提供了一种用于实现上述所涉及的图像恢复方 法的图像恢复装置。该装置所提供的解决问题的实现方案与上述方法中所记载的实现方案相似，故下面所提供的一个或多个图像恢复装置实施例中的具体限定可以参见上文中对于图像恢复方法的限定，在此不再赘述。
在一个实施例中，如图8所示，提供了一种图像恢复装置，包括：获取模块10、确定模块20和恢复模块30，其中：
获取模块10，用于获取包含至少两个待检测物体的目标散斑图像。
确定模块20，用于根据目标散斑图像，确定目标散斑图像中所包含物体的中心自相关数据。
恢复模块30，用于基于旁瓣恢复网络，对中心自相关数据进行旁瓣恢复处理，得到包含各待检测物体的目标图像。
其中，旁瓣恢复网络的训练过程包括：获取样本物体的参考图像和标准图像；其中，参考图像为初始恢复网络对样本物体的样本自相关数据进行旁瓣恢复处理得到的图像，标准图像为包含样本物体的样本旁瓣的图像；对参考图像与标准图像进行相似度比较，得到参考图像与标准图像的图像相似度；根据参考图像的参考像素值和标准图像的标准像素值，确定参考图像与标准图像的平均误差幅度；根据图像相似度和平均误差幅度，对初始恢复网络进行参数调整，得到旁瓣恢复网络。
在一个实施例中，如图9所示，提供了一种图像恢复装置，该图像恢复装置中确定模块20包括：第一确定单元21和第二确定单元22，其中：
第一确定单元21，用于根据目标散斑图像，分别获取各待检测物体的候选物体强度图。
第二确定单元22，用于根据各待检测物体的物体间距，对各待检测物体的候选物体强度图进行相关运算，得到中心自相关数据。
在一个实施例中，如图10所示，提供了一种图像恢复装置，该图像恢复装置中第二确定单元22包括：第一确定子单元221、第二确定子单元222和第三确定子单元223，其中：
第一确定子单元221，用于分别对各待检测物体的候选物体强度图进行自相关运算，得到第一自相关数据。
第二确定子单元222，用于分别对各待检测物体的候选物体强度图进行互相关运算，得到第二自相关数据；
第三确定子单元223，用于将第一自相关数据和第二自相关数据的和值，作为中心自相关数据。
在一个实施例中，如图11所示，提供了一种图像恢复装置，该图像恢复装置中获取模块10包括：获取单元11和校正单元12，其中：
获取单元11，用于获取包含至少两个待检测物体的初始散斑图像。
校正单元12，用于对初始散斑图像进行包络校正处理，得到目标散斑图像。
在一个实施例中，如图12所示，提供了一种图像恢复装置，该图像恢复装置中校正单元12包括：拟合子单元121和处理子单元122，其中：
拟合子单元121，用于采用泽尼克多项式对初始散斑图像进行拟合处理，得到拟合曲面。
处理子单元122，用于将初始散斑图像与拟合曲面进行相除处理，得到目标散斑图像。
上述图像恢复装置中的各个模块可全部或部分通过软件、硬件及其组合来实现。上述各模块可以硬件形式内嵌于或独立于计算机设备中的处理器中，也可以以软件形式存储于计算机设备中的存储器中，以便于处理器调用执行以上各个模块对应的操作。
在一个实施例中，提供了一种计算机设备，该计算机设备可以是终端，其内部结构图可以如图13所示。该计算机设备包括处理器、存储器、输入/输出接口、通信接口、显示单元和输入装置。其中，处理器、存储器和输入/输出接口通过系统总线连接，通信接口、显示单元和输入装置通过输入/输出接口连接到系统总线。其中，该计算机设备的处理器用于提供计算和控制能力。该计算机设备的存储器包括非易失性存储介质和内存储器。该非易失性存储介质存储有操作系统和计算机程序。该内存储器为非易失性存储介质中的操作系统和计算机程序的运行提供环境。该计算机设备的输入/输出接口用于处理器与外部设备之间交换信息。该计算机设备的通信接口用于与外部的终端进行有线或无线方式的通信，无线方式可通过WIFI、移动蜂窝网络、NFC(近场通信)或其他技术实现。该计算机程序被处理器执行时以实现一种图像恢复方法。该计算机设备的显示单元用于形成视觉可见的画面，可以是显示屏、投影装置或虚拟现实成像装置。显示屏可以是液晶显示屏或者电子墨水显示屏，该计算机设备的输入装置可以是显示屏上覆盖的触摸层，也可以是计算机设备外壳上设置的按键、轨迹球或触控板，还可以是外接的键盘、触控板或鼠标等。
本领域技术人员可以理解，图13中示出的结构，仅仅是与本申请方案相关的部分结构的框图，并不构成对本申请方案所应用于其上的计算机设备的限定，具体的计算机设备可以包括比图中所示更多或更少的部件，或者组合某些部件，或者具有不同的部件布置。
在一个实施例中，提供了一种计算机设备，包括存储器和处理器，存储器中存储有计 算机程序，该处理器执行计算机程序时实现以下步骤：
获取包含至少两个待检测物体的目标散斑图像；
根据目标散斑图像，确定目标散斑图像中所包含物体的中心自相关数据；
基于旁瓣恢复网络，对中心自相关数据进行旁瓣恢复处理，得到包含各待检测物体的目标图像。
在一个实施例中，处理器执行计算机程序时还实现以下步骤：
根据目标散斑图像，分别获取各待检测物体的候选物体强度图；
根据各待检测物体的物体间距，对各待检测物体的候选物体强度图进行相关运算，得到中心自相关数据。
在一个实施例中，处理器执行计算机程序时还实现以下步骤：
分别对各待检测物体的候选物体强度图进行自相关运算，得到第一自相关数据；
分别对各待检测物体的候选物体强度图进行互相关运算，得到第二自相关数据；
将第一自相关数据和第二自相关数据的和值，作为中心自相关数据。
在一个实施例中，处理器执行计算机程序时还实现以下步骤：
获取包含至少两个待检测物体的初始散斑图像；
对初始散斑图像进行包络校正处理，得到目标散斑图像。
在一个实施例中，处理器执行计算机程序时还实现以下步骤：
采用泽尼克多项式对初始散斑图像进行拟合处理，得到拟合曲面；
将初始散斑图像与拟合曲面进行相除处理，得到目标散斑图像。
在一个实施例中，处理器执行计算机程序时还实现以下步骤：
获取样本物体的参考图像和标准图像；其中，参考图像为初始恢复网络对样本物体的样本自相关数据进行旁瓣恢复处理得到的图像，标准图像为包含样本物体的样本旁瓣的图像；
对参考图像与标准图像进行相似度比较，得到参考图像与标准图像的图像相似度；
根据参考图像的参考像素值和标准图像的标准像素值，确定参考图像与标准图像的平均误差幅度；
根据图像相似度和平均误差幅度，对初始恢复网络进行参数调整，得到旁瓣恢复网络。
在一个实施例中，提供了一种计算机可读存储介质，其上存储有计算机程序，计算机程序被处理器执行时实现以下步骤：
获取包含至少两个待检测物体的目标散斑图像；
根据目标散斑图像，确定目标散斑图像中所包含物体的中心自相关数据；
基于旁瓣恢复网络，对中心自相关数据进行旁瓣恢复处理，得到包含各待检测物体的目标图像。
在一个实施例中，计算机程序被处理器执行时还实现以下步骤：
根据目标散斑图像，分别获取各待检测物体的候选物体强度图；
根据各待检测物体的物体间距，对各待检测物体进行相关运算，得到自相关数据。
在一个实施例中，计算机程序被处理器执行时还实现以下步骤：
分别对各待检测物体的候选物体强度图进行自相关运算，得到第一自相关数据；
分别对各待检测物体的候选物体强度图进行互相关运算，得到第二自相关数据；
将第一自相关数据和第二自相关数据的和值，作为中心自相关数据。
在一个实施例中，计算机程序被处理器执行时还实现以下步骤：
获取包含至少两个待检测物体的初始散斑图像；
对初始散斑图像进行包络校正处理，得到目标散斑图像。
在一个实施例中，计算机程序被处理器执行时还实现以下步骤：
采用泽尼克多项式对初始散斑图像进行拟合处理，得到拟合曲面；
将初始散斑图像与拟合曲面进行相除处理，得到目标散斑图像。
在一个实施例中，计算机程序被处理器执行时还实现以下步骤：
获取样本物体的参考图像和标准图像；其中，参考图像为初始恢复网络对样本物体的样本自相关数据进行旁瓣恢复处理得到的图像，标准图像为包含样本物体的样本旁瓣的图像；
对参考图像与标准图像进行相似度比较，得到参考图像与标准图像的图像相似度；
根据参考图像的参考像素值和标准图像的标准像素值，确定参考图像与标准图像的平均误差幅度；
根据图像相似度和平均误差幅度，对初始恢复网络进行参数调整，得到旁瓣恢复网络。
在一个实施例中，提供了一种计算机程序产品，包括计算机程序，该计算机程序被处理器执行时实现以下步骤：
获取包含至少两个待检测物体的目标散斑图像；
根据目标散斑图像，确定目标散斑图像中所包含物体的中心自相关数据；
基于旁瓣恢复网络，对中心自相关数据进行旁瓣恢复处理，得到包含各待检测物体的目标图像。
在一个实施例中，计算机程序被处理器执行时还实现以下步骤：
根据目标散斑图像，分别获取各待检测物体的候选物体强度图；
根据各待检测物体的物体间距，对各待检测物体的候选物体强度图进行相关运算，得到中心自相关数据。
在一个实施例中，计算机程序被处理器执行时还实现以下步骤：
分别对各待检测物体的候选物体强度图进行自相关运算，得到第一自相关数据；
分别对各待检测物体的候选物体强度图进行互相关运算，得到第二自相关数据；
将第一自相关数据和第二自相关数据的和值，作为中心自相关数据。
在一个实施例中，计算机程序被处理器执行时还实现以下步骤：
获取包含至少两个待检测物体的初始散斑图像；
对初始散斑图像进行包络校正处理，得到目标散斑图像。
在一个实施例中，计算机程序被处理器执行时还实现以下步骤：
采用泽尼克多项式对初始散斑图像进行拟合处理，得到拟合曲面；
将初始散斑图像与拟合曲面进行相除处理，得到目标散斑图像。
在一个实施例中，计算机程序被处理器执行时还实现以下步骤：
获取样本物体的参考图像和标准图像；其中，参考图像为初始恢复网络对样本物体的样本自相关数据进行旁瓣恢复处理得到的图像，标准图像为包含样本物体的样本旁瓣的图像；
对参考图像与标准图像进行相似度比较，得到参考图像与标准图像的图像相似度；
根据参考图像的参考像素值和标准图像的标准像素值，确定参考图像与标准图像的平均误差幅度；
根据图像相似度和平均误差幅度，对初始恢复网络进行参数调整，得到旁瓣恢复网络。
需要说明的是，本申请所涉及的用户信息(包括但不限于用户设备信息、用户个人信息等)和数据(包括但不限于用于分析的数据、存储的数据、展示的数据等)，均为经用户授权或者经过各方充分授权的信息和数据，且相关数据的收集、使用和处理需要遵守相关国家和地区的相关法律法规和标准。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程，是可以通过计算机程序来指令相关的硬件来完成，的计算机程序可存储于一非易失性计算机可读取存储介质中，该计算机程序在执行时，可包括如上述各方法的实施例的流程。其中，本申请所提供的各实施例中所使用的对存储器、数据库或其它介质的任何引用，均可包括非易失性和易 失性存储器中的至少一种。非易失性存储器可包括只读存储器(Read-Only Memory，ROM)、磁带、软盘、闪存、光存储器、高密度嵌入式非易失性存储器、阻变存储器(ReRAM)、磁变存储器(Magnetoresistive Random Access Memory，MRAM)、铁电存储器(Ferroelectric Random Access Memory，FRAM)、相变存储器(Phase Change Memory，PCM)、石墨烯存储器等。易失性存储器可包括随机存取存储器(Random Access Memory，RAM)或外部高速缓冲存储器等。作为说明而非局限，RAM可以是多种形式，比如静态随机存取存储器(Static Random Access Memory，SRAM)或动态随机存取存储器(Dynamic Random Access Memory，DRAM)等。本申请所提供的各实施例中所涉及的数据库可包括关系型数据库和非关系型数据库中至少一种。非关系型数据库可包括基于区块链的分布式数据库等，不限于此。本申请所提供的各实施例中所涉及的处理器可为通用处理器、中央处理器、图形处理器、数字信号处理器、可编程逻辑器、基于量子计算的数据处理逻辑器等，不限于此。
以上实施例的各技术特征可以进行任意的组合，为使描述简洁，未对上述实施例中的各个技术特征所有可能的组合都进行描述，然而，只要这些技术特征的组合不存在矛盾，都应当认为是本说明书记载的范围。
以上实施例仅表达了本申请的几种实施方式，其描述较为具体和详细，但并不能因此而理解为对本申请专利范围的限制。应当指出的是，对于本领域的普通技术人员来说，在不脱离本申请构思的前提下，还可以做出若干变形和改进，这些都属于本申请的保护范围。因此，本申请的保护范围应以所附权利要求为准。

Claims (10)

  1. 一种图像恢复方法，其特征在于，所述方法包括：
    获取包含至少两个待检测物体的目标散斑图像；
    根据所述目标散斑图像，确定所述目标散斑图像中所包含物体的中心自相关数据；
    基于旁瓣恢复网络，对所述中心自相关数据进行旁瓣恢复处理，得到包含各待检测物体的目标图像。
  2. 根据权利要求1所述的方法，其特征在于，所述根据所述目标散斑图像，确定所述目标散斑图像中所包含物体的中心自相关数据，包括：
    根据所述目标散斑图像，分别获取各所述待检测物体的候选物体强度图；
    根据各所述待检测物体的物体间距，对各所述待检测物体的候选物体强度图进行相关运算，得到所述中心自相关数据。
  3. 根据权利要求2所述的方法，其特征在于，所述根据所述目标散斑图像，确定所述目标散斑图像中所包含物体的中心自相关数据，包括：
    分别对各所述待检测物体的候选物体强度图进行自相关运算，得到第一自相关数据；
    分别对各所述待检测物体的候选物体强度图进行互相关运算，得到第二自相关数据；
    将所述第一自相关数据和所述第二自相关数据的和值，作为所述中心自相关数据。
  4. 根据权利要求1-3任一项所述的方法，其特征在于，所述获取包含至少两个待检测物体的目标散斑图像，包括：
    获取包含至少两个待检测物体的初始散斑图像；
    对所述初始散斑图像进行包络校正处理，得到所述目标散斑图像。
  5. 根据权利要求4所述的方法，其特征在于，所述对所述初始散斑图像进行包络校正处理，得到所述目标散斑图像，包括：
    采用泽尼克多项式对所述初始散斑图像进行拟合处理，得到拟合曲面；
    将所述初始散斑图像与所述拟合曲面进行相除处理，得到所述目标散斑图像。
  6. 根据权利要求1-3任一项所述的方法，其特征在于，所述旁瓣恢复网络的训练过程，包括：
    获取样本物体的参考图像和标准图像；其中，所述参考图像为初始恢复网络对所述样本物体的样本自相关数据进行旁瓣恢复处理得到的图像，所述标准图像为包含所述样本物体的样本旁瓣的图像；
    对所述参考图像与所述标准图像进行相似度比较，得到所述参考图像与所述标准图像的图像 相似度；
    根据所述参考图像的参考像素值和所述标准图像的标准像素值，确定所述参考图像与所述标准图像的平均误差幅度；
    根据所述图像相似度和所述平均误差幅度，对所述初始恢复网络进行参数调整，得到所述旁瓣恢复网络。
  7. 一种图像恢复装置，其特征在于，所述装置包括：
    获取模块，用于获取包含至少两个待检测物体的目标散斑图像；
    确定模块，用于根据所述目标散斑图像，确定所述目标散斑图像中所包含物体的中心自相关数据；
    恢复模块，用于基于旁瓣恢复网络，对所述中心自相关数据进行旁瓣恢复处理，得到包含各待检测物体的目标图像。
  8. 一种计算机设备，包括存储器和处理器，所述存储器存储有计算机程序，其特征在于，所述处理器执行所述计算机程序时实现权利要求1至6中任一项所述的方法的步骤。
  9. 一种计算机可读存储介质，其上存储有计算机程序，其特征在于，所述计算机程序被处理器执行时实现权利要求1至6中任一项所述的方法的步骤。
  10. 一种计算机程序产品，包括计算机程序，其特征在于，该计算机程序被处理器执行时实现权利要求1至6中任一项所述的方法的步骤。
PCT/CN2024/075909 2023-12-21 2024-02-05 图像恢复方法、装置、计算机设备和存储介质 Pending WO2025129804A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202311765123.X 2023-12-21
CN202311765123.XA CN117689564A (zh) 2023-12-21 2023-12-21 图像恢复方法、装置、计算机设备和存储介质

Publications (1)

Publication Number Publication Date
WO2025129804A1 true WO2025129804A1 (zh) 2025-06-26

Family

ID=90135099

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2024/075909 Pending WO2025129804A1 (zh) 2023-12-21 2024-02-05 图像恢复方法、装置、计算机设备和存储介质

Country Status (2)

Country Link
CN (1) CN117689564A (zh)
WO (1) WO2025129804A1 (zh)

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR101858137B1 (ko) * 2017-03-08 2018-05-16 대진대학교 산학협력단 컴퓨터 시뮬레이션을 이용한 초음파 영상의 부엽 억제 필터 평가 방법
CN111179368A (zh) * 2019-12-26 2020-05-19 西安电子科技大学 大视场多目标散斑成像方法、装置、电子设备及存储介质
CN111795949A (zh) * 2020-06-12 2020-10-20 北京理工大学 抗散射成像方法与装置
CN112950482A (zh) * 2019-12-10 2021-06-11 深圳大学 物体信息恢复方法、装置、计算机设备和存储介质
CN113902819A (zh) * 2020-06-22 2022-01-07 深圳大学 透过散射介质成像的方法、装置、计算机设备和存储介质
CN114549486A (zh) * 2022-02-25 2022-05-27 杭州电子科技大学 一种无先验多目标的实时散射成像方法、装置及存储介质
CN115082364A (zh) * 2021-03-10 2022-09-20 深圳大学 基于散斑图叠加和平均的超记忆效应范围成像方法和装置
CN116506714A (zh) * 2023-06-26 2023-07-28 之江实验室 成像方法、装置、系统、存储介质以及电子设备

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR101858137B1 (ko) * 2017-03-08 2018-05-16 대진대학교 산학협력단 컴퓨터 시뮬레이션을 이용한 초음파 영상의 부엽 억제 필터 평가 방법
CN112950482A (zh) * 2019-12-10 2021-06-11 深圳大学 物体信息恢复方法、装置、计算机设备和存储介质
CN111179368A (zh) * 2019-12-26 2020-05-19 西安电子科技大学 大视场多目标散斑成像方法、装置、电子设备及存储介质
CN111795949A (zh) * 2020-06-12 2020-10-20 北京理工大学 抗散射成像方法与装置
CN113902819A (zh) * 2020-06-22 2022-01-07 深圳大学 透过散射介质成像的方法、装置、计算机设备和存储介质
CN115082364A (zh) * 2021-03-10 2022-09-20 深圳大学 基于散斑图叠加和平均的超记忆效应范围成像方法和装置
CN114549486A (zh) * 2022-02-25 2022-05-27 杭州电子科技大学 一种无先验多目标的实时散射成像方法、装置及存储介质
CN116506714A (zh) * 2023-06-26 2023-07-28 之江实验室 成像方法、装置、系统、存储介质以及电子设备

Also Published As

Publication number Publication date
CN117689564A (zh) 2024-03-12

Similar Documents

Publication Publication Date Title
CN110660066B (zh) 网络的训练方法、图像处理方法、网络、终端设备及介质
CN110148157B (zh) 画面目标跟踪方法、装置、存储介质及电子设备
CN110084155B (zh) 一种密集人数计数的方法、装置、设备以及存储介质
WO2021139178A1 (zh) 图像合成方法及相关设备
CN115100286B (zh) 无人机采集视点确定方法、装置、计算机设备和存储介质
CN116304677B (zh) 模型的通道剪枝方法、装置、计算机设备和存储介质
CN116228753A (zh) 肿瘤预后评估方法、装置、计算机设备和存储介质
CN117058022A (zh) 深度图像去噪方法、装置、计算机设备及存储介质
CN114972100B (zh) 噪声模型的估计方法和装置、图像处理方法和装置
CN112699826B (zh) 人脸检测方法、装置、计算机设备和存储介质
CN118747728A (zh) 图像去模糊方法、装置、计算机设备和可读存储介质
CN115082364B (zh) 基于散斑图叠加和平均的超记忆效应范围成像方法和装置
CN115035478B (zh) 对象统计方法、装置、设备及存储介质
US20230401670A1 (en) Multi-scale autoencoder generation method, electronic device and readable storage medium
CN117689564A (zh) 图像恢复方法、装置、计算机设备和存储介质
CN116883257A (zh) 图像去雾方法、装置、计算机设备和存储介质
CN117037244A (zh) 人脸安全检测方法、装置、计算机设备和存储介质
CN116978070A (zh) 资源转移方法、装置、设备和介质
CN115174811A (zh) 相机抖动检测方法、装置、设备、存储介质和程序产品
CN114140570A (zh) 一种获取纹理贴图的方法及装置
CN115861520B (zh) 高光检测方法、装置、计算机设备和存储介质
CN119316737B (zh) 图像处理方法、装置、电子设备和计算机可读存储介质
CN117292178B (zh) 目标检测方法、装置、计算机设备及计算机可读存储介质
TWI840833B (zh) 多尺度自編碼器生成方法、電腦設備及儲存介質
CN114781490B (zh) 多任务模型迁移训练方法、装置、计算机设备及存储介质

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: 24905310

Country of ref document: EP

Kind code of ref document: A1