CN117115025A - Image rapid restoration method, device and medium based on multiframe blind deconvolution - Google Patents
Image rapid restoration method, device and medium based on multiframe blind deconvolution Download PDFInfo
- Publication number
- CN117115025A CN117115025A CN202311078013.6A CN202311078013A CN117115025A CN 117115025 A CN117115025 A CN 117115025A CN 202311078013 A CN202311078013 A CN 202311078013A CN 117115025 A CN117115025 A CN 117115025A
- Authority
- CN
- China
- Prior art keywords
- image
- restoration
- target image
- point spread
- spread function
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 81
- 230000008569 process Effects 0.000 claims abstract description 29
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 25
- 230000006870 function Effects 0.000 claims description 79
- 238000005457 optimization Methods 0.000 claims description 23
- 238000004364 calculation method Methods 0.000 claims description 15
- 238000004590 computer program Methods 0.000 claims description 12
- 238000002939 conjugate gradient method Methods 0.000 claims description 11
- 238000013519 translation Methods 0.000 claims description 9
- 230000021615 conjugation Effects 0.000 claims description 6
- 238000001228 spectrum Methods 0.000 claims description 6
- 230000009466 transformation Effects 0.000 claims description 3
- 230000017105 transposition Effects 0.000 claims description 3
- 238000011084 recovery Methods 0.000 claims description 2
- 238000002474 experimental method Methods 0.000 description 4
- 238000004088 simulation Methods 0.000 description 3
- 230000003044 adaptive effect Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR 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; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
- G06T7/37—Determination of transform parameters for the alignment of images, i.e. image registration using transform domain methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR 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]
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (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)
- Computer Vision & Pattern Recognition (AREA)
- Image Processing (AREA)
Abstract
The application discloses a multi-frame blind deconvolution-based image quick restoration method, equipment and medium, which comprises the following steps: step S100: registering the observation images; step S200: initializing and setting a target image and a point spread function; step S300: the method for recovering the image by multi-frame blind deconvolution specifically comprises the following steps: step S301, judging whether the image meets the standard, specifically, if the image meets the standard, transmitting the point spread function and the target image from the GPU to the CPU, and outputting the point spread function and the target image in an image form; if the standard is not met, the step S302 is carried out; step S302: optimizing the target image and the point spread function, and proceeding to step S303; step S303: the number of restoration times is updated, specifically, the number of restoration times is increased by 1, and the process returns to step S301. The application has the advantages that the method greatly improves the operation speed of the algorithm while not reducing the restoration quality of the multi-frame blind deconvolution algorithm on the image, and realizes the quick restoration of the image under the single GPU condition of a single desktop.
Description
Technical Field
The application relates to the technical field of image restoration, in particular to an image rapid restoration method, device and medium based on multi-frame blind deconvolution.
Background
The multi-frame blind deconvolution algorithm can realize high-resolution image reconstruction of blurred images with few frames (the number of frames is lower than 20), and plays a key role in space target observation. The traditional multi-frame blind deconvolution algorithm uses a conjugate gradient method in the calculation process, the existing conjugate gradient method has a triple-cycle structure, the calculation amount is large, the calculation time is long, the single restoration task usually needs minutes or even hours, and the practical application is greatly limited.
In view of the foregoing, there is a need for a method, apparatus and medium for fast image restoration based on multi-frame blind deconvolution to solve the problems in the prior art.
Disclosure of Invention
The application aims to provide a multi-frame blind deconvolution-based image quick restoration method, equipment and medium, and the specific technical scheme is as follows:
the image quick restoration method based on multi-frame blind deconvolution is characterized by comprising the following steps:
step S100: registering an observation image, specifically, reading in the observation image, and registering the observation image;
step S200: the initialization setting, namely, transmitting the registered observation image into the GPU from the CPU to perform the initialization setting, wherein the initialization setting comprises setting the initial value of a target image and a point spread function, constructing a cost function, selecting a parameterization method and an optimization algorithm;
step S300: the method for recovering the image by multi-frame blind deconvolution specifically comprises the following steps:
step S301, judging whether the target image reaches the standard, specifically, checking whether the target image reaches the standard or whether the restoration process reaches the maximum restoration times; if the target image reaches the standard or the restoration process reaches the maximum restoration times, transmitting the point spread function and the target image from the GPU to the CPU, and outputting the point spread function and the target image in an image form; if the target image does not reach the standard and the restoration process does not reach the maximum restoration times, the step S302 is entered;
step S302: optimizing a target image and a point spread function, namely, keeping the target image unchanged, and optimizing the point spread function by using a macopt-based optimization conjugate gradient method; maintaining the point spread function unchanged, optimizing the target image by using a macopt-based optimized conjugate gradient method, and entering step S303;
step S303: the number of restoration times is updated, specifically, the number of restoration times is increased by 1, and the process returns to step S301.
Preferably, in step S100, on the basis of removing the background and the flat field, a phase correlation algorithm based on fourier transform is used to perform translational registration on the observed image, and the specific process is as follows:
calculate the first frame observation image g 1 Fourier transform G of (2) 1 Then calculate the Fourier transform G of the w-th frame observation image w The normalized power spectrum R (u, v) of the first and w-th frame images is shown in formula 1):
wherein, (x) 0 ,y 0 ) Representing the amount of translation between the first frame and the w frame; * Represents conjugation; (u, v) represents a spatial frequency; j is an imaginary unit;
performing fourier inverse transformation on formula 1) to obtain formula 2):
r(x,y)＝FFT -1 (R)＝δ(x-x 0 ,y-y 0 ) 2)
wherein FFT -1 Representing an inverse fourier transform; r (x, y) is the inverse fourier transform of the normalized power spectrum R (u, v); delta (x-x) 0 ,y-y 0 ) The representation center is located at (x) 0 ,y 0 ) Is a dirac function of (c); equation 2) represents a function containing sharp pulses whose positions represent the magnitude of the translation values based on which the translation registration of the observed image is achieved.
Preferably, in step S200, the expression of the cost function is as follows:
wherein J (x, y) represents a cost function; sigma (sigma) 2 Representing the variance of observed image noise; g w (x, y) represents a w-th frame observation image; f (x, y) represents a target image; h is a w (x, y) represents a point spread function of the w-th frame observation image;representing a convolution operation; />To derive a symbol; />Representation->Is a binary norm of (2); lambda (lambda) 0 Is a regularization coefficient.
Preferably, in step S200, the initial value of the target image is an average value of the observed image, and the initial value of the point spread function is an autocorrelation of the observed image, where the expression is as follows:
wherein f 0 (x, y) represents an initial value of the target image, W represents the number of observed image frames,representing the initial value of the point spread function of the w-th frame, for example>Representing a convolution calculation.
Preferably, in step S200, the parameterization method is used to parameterize the target image and the point spread function, where the expression is as follows:
wherein h (x, y) represents a point spread function;representing the parameterized variable of the target image; phi (phi) 2 (x, y) represents the parameterized variable of the point spread function.
Preferably, in step S301, it is checked whether the target image meets the standard or whether the restoration process reaches the maximum restoration number, specifically:
setting the restoration frequency k, wherein the initial value is 1, adding 1 after completing one image restoration, setting the maximum restoration frequency as k_maxiter, setting a gradient convergence standard epsilon, checking whether k > k_maxiter exists or whether the norm of the cost function gradient is smaller than epsilon, stopping image restoration if one of the two conditions is met, and continuing the next image restoration if the k > k_maxiter or the norm of the cost function gradient is not met.
Preferably, in step S301, the cost function gradient y k The expression of (2) is as follows:
wherein a is k And b k Is a range of two endpoints, s, containing minima k And t k A is respectively k And b k Is a direction discrimination value of (a).
Preferably, in step S302, the method for optimizing the point spread function and the target image by using the conjugate gradient method is the same, and the specific procedure is as follows:
point spread function h for kth image restoration k And the target image f k Collectively denoted as x in the respective optimization process k The objective of the optimization is to minimize the cost function, namely:
min{J(x k )|x k ∈R n } 7)
the optimized conjugate gradient algorithm based on macopt is to find the x which minimizes the cost function by an iterative method k The expression is as follows:
x k+1 ＝x k +α k d k 8)
wherein d k Represents the search direction, alpha, of the kth iteration k A search step representing the kth iteration;
solving for d k The expression of (2) is:
beta in formula 9) k Representing the conjugation direction, the expression is:
wherein T represents transposition;
in formula 6), s k And t k The calculation formula of (2) is as follows:
for two end points a k And b k Solving by adopting an iterative method, setting the upper limit of the optimization frequency of the iterative optimization process as i_maxiter, recording the optimization frequency i, and if the initial value is 1, a k And b k The calculation formula is as follows:
a k ＝x k +2 i+1 d k l k 13)
b k ＝x k +2 -i-1 d k l k 14)
wherein l k In relation to the search step, the expression is:
wherein m represents the single-side pixel number of the image, and m is the total pixel number of the image; the kth iteration step alpha k Calculated from the following formula:
in formulae 13) and 14), if s k t k <0, then a k And b k A minimum value exists between the two, and searching is stopped; otherwise, the iteration number is increased by 1, and a) is continued according to the formulas 13) and 14) k And b k Searching until s occurs k t k <0, or the maximum searching times i_maxiter is reached, stopping searching;
thus far, the point spread function h of the kth image restoration is completed k Or a target image f k Is described.
In addition, the application also includes a computer device comprising:
at least one processor;
and a memory communicatively coupled to the at least one processor;
wherein the memory is used for storing a computer program;
the processor is configured to implement the image quick restoration method as described above when executing the computer program.
In addition, the application further comprises a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the computer program realizes the image quick restoration method when being executed by a processor.
The technical scheme of the application has the following beneficial effects:
according to the application, the target image with the minimum cost function and the point spread function are estimated by optimizing the conjugate gradient algorithm, so that the calculation coincidence of step length search in the estimation process is reduced, the convergence speed of the algorithm is accelerated, and the image restoration efficiency is improved.
In addition to the objects, features and advantages described above, the present application has other objects, features and advantages. The present application will be described in further detail with reference to the drawings.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and together with the description serve to explain the application. In the drawings:
FIG. 1 is a flow chart showing the steps of a method for rapid image restoration in a preferred embodiment of the present application;
fig. 2 is a graph comparing images of experiments in a preferred embodiment of the present application.
Detailed Description
Embodiments of the application are described in detail below with reference to the attached drawings, but the application can be implemented in a number of different ways, which are defined and covered by the claims.
Examples:
referring to fig. 1, the embodiment discloses an image rapid restoration method based on multi-frame blind deconvolution, which comprises the following steps:
step S100: registering an observation image, specifically, reading in the observation image, and registering the observation image;
step S200: the initialization setting, namely, transmitting the registered observation image into the GPU from the CPU to perform the initialization setting, wherein the initialization setting comprises setting the initial value of a target image and a point spread function, constructing a cost function, selecting a parameterization method and an optimization algorithm;
step S300: the method for recovering the image by multi-frame blind deconvolution specifically comprises the following steps:
step S301, judging whether the target image reaches the standard, specifically, checking whether the target image reaches the standard or whether the restoration process reaches the maximum restoration times; if the target image reaches the standard or the restoration process reaches the maximum restoration times, transmitting the point spread function and the target image from the GPU to the CPU, and outputting the point spread function and the target image in an image form; if the target image does not reach the standard or the restoration process does not reach the maximum restoration times, the step S302 is entered;
step S302: optimizing a target image and a point spread function, namely, keeping the target image unchanged, and optimizing the point spread function by using a macopt-based optimization conjugate gradient method; maintaining the point spread function unchanged, optimizing the target image by using a macopt-based optimized conjugate gradient method, and entering step S303;
step S303: the number of restoration times is updated, specifically, the number of restoration times is increased by 1, and the process returns to step S301.
Preferably, in step S100, on the basis of removing the background and the flat field, a phase correlation algorithm based on fourier transform is used to perform translational registration on the observed image, and the specific process is as follows:
calculate the first frame observation image g 1 Fourier transform G of (2) 1 Then calculate the Fourier transform G of the w-th frame observation image w The normalized power spectrum R (u, v) of the first and w-th frame images is shown in formula 1):
wherein, (x) 0 ,y 0 ) Representing the amount of translation between the first frame and the w frame; * Represents conjugation; (u, v) represents a spatial frequency; j is an imaginary unit;
performing fourier inverse transformation on formula 1) to obtain formula 2):
r(x,y)＝FFT -1 (R)＝δ(x-x 0 ,y-y 0 ) 2)
wherein FFT -1 Representing an inverse fourier transform; r (x, y) is the inverse fourier transform of the normalized power spectrum R (u, v); delta (x-x) 0 ,y-y 0 ) The representation center is located at (x) 0 ,y 0 ) Is a dirac function of (c); equation 2) represents a function containing sharp pulses whose positions represent the magnitude of the translation values based on which the translation registration of the observed image is achieved.
Preferably, in step S200, the expression of the cost function is as follows:
wherein J (x, y) represents a cost function; sigma (sigma) 2 Representing the variance of observed image noise; g w (x, y) represents a w-th frame observation image; f (x, y) represents a target image; h is a w (x, y) represents a point spread function of the w-th frame observation image;representing a convolution operation; />To derive a symbol; />Representation->Is a binary norm of (2); lambda (lambda) 0 Is a regularization coefficient.
Preferably, in step S200, the initial value of the target image is an average value of the observed images, and the initial value of the point spread function is an autocorrelation of the observed images of each frame, and the expression is as follows:
wherein f 0 (x, y) represents an initial value of the target image, W represents the number of observed image frames,representing the initial value of the point spread function of the w-th frame, for example>Representing a convolution calculation.
Preferably, in step S200, the parameterization method is used to parameterize the target image and the point spread function, where the expression is as follows:
wherein h (x, y) represents a point spread function;representing the parameterized variable of the target image; phi (phi) 2 (x, y) represents the parameterized variable of the point spread function.
Preferably, in step S300, it is checked whether the target image meets the standard or whether the restoration process reaches the maximum restoration number, specifically:
setting the restoration frequency k, wherein the initial value is 1, adding 1 after completing one image restoration, setting the maximum restoration frequency as k_maxiter, setting a gradient convergence standard epsilon, checking whether k > k_maxiter exists or whether the norm of the cost function gradient is smaller than epsilon, stopping image restoration if one of the two conditions is met, and continuing the next image restoration if the k > k_maxiter or the norm of the cost function gradient is not met.
Preferably, in step S300, the cost function gradient y k The expression of (2) is as follows:
wherein a is k And b k Is a range of two endpoints, s, containing minima k And t k A is respectively k And b k Is a direction discrimination value of (a).
Preferably, in step S302, the method for optimizing the point spread function and the target image by using the conjugate gradient method is the same, and the specific procedure is as follows:
point spread function h for kth image restoration k And the target image f k Collectively denoted as x in the respective optimization process k The objective of the optimization is to minimize the cost function, namely:
min{J(x k )|x k ∈R n } 7)
the optimized conjugate gradient algorithm based on macopt is to find the x which minimizes the cost function by an iterative method k The expression is as follows:
x k+1 ＝x k +α k d k 8)
wherein d k Represents the search direction, alpha, of the kth iteration k A search step representing the kth iteration;
solving for d k The expression of (2) is:
beta in formula 9) k Representing the conjugation direction, the expression is:
wherein T represents transposition;
in formula 6), s k And t k The calculation formula of (2) is as follows:
for two end points a k And b k Solving by adopting an iterative method, setting the upper limit of the optimization frequency of the iterative optimization process as i_maxiter, recording the optimization frequency i, and if the initial value is 1, a k And b k The calculation formula is as follows:
a k ＝x k +2 i+1 d k l k 13)
wherein l k In relation to the search step, the expression is:
wherein m represents the single-side pixel number of the image, and m is the total pixel number of the image; the kth iteration step alpha k Calculated from the following formula:
in formulae 13) and 14), if s k t k <0, then a k And b k A minimum value exists between the two, and searching is stopped; otherwise, the iteration number is increased by 1, and a) is continued according to the formulas 13) and 14) k And b k Searching until s occurs k t k <0, or the maximum searching times i_maxiter is reached, stopping searching;
thus far, the point spread function h of the kth image restoration is completed k Or a target image f k Is described.
The image rapid restoration method disclosed by the embodiment adopts an optimized conjugate gradient algorithm, and the target image with the minimum cost function and the point spread function are estimated by solving the optimized conjugate gradient algorithm, so that the calculation load of step search in the estimation process is reduced, the convergence speed of the algorithm is accelerated, and the calculation speed of PC equipment is remarkably improved.
In order to further explain the advantages of the image quick restoration method in this embodiment, an experiment of performing image restoration under the same operation environment by adopting a traditional multi-frame blind deconvolution algorithm and the image quick restoration method disclosed in this embodiment is described below, and the specific experimental process is as follows:
in the experiment, a degraded image which is generated by the simulation of the adaptive optical simulation software CAOS and corrected by the adaptive optical system is selected as an observation image, and the observation image is restored, wherein simulation parameters of the degraded image generated by the CAOS are shown in table 1.
TABLE 1 simulation parameters of degraded images
The development environment used in this experiment is shown in table 2.
Table 2 multiframe blind deconvolution acceleration algorithm development environment
Selection of restoration parameters: maximum number of recoveries k _maxiter =60, in each restoration process, the number of optimizations of the target image is 2, the number of optimizations of the point spread function is 1, the maximum search number i maxiter =10, variance σ of observed image noise 2 =1, regularization coefficient λ 0 ＝0.1。
Further, as shown in fig. 2, fig. 2 (a) shows an original image, fig. 2 (b) shows an observed image, fig. 2 (c) shows a restoration result of a conventional multi-frame blind deconvolution algorithm, the minimum mean square error NMSE value of the restored image and the original image is 0.058, and the algorithm takes 815.42s. Fig. 2 (d) shows the restoration result of the image quick restoration method of this embodiment, the NMSE value is 0.058, and the algorithm takes 4.27s.
According to the restoration results, the image quick restoration method can greatly improve the operation speed of the algorithm without reducing the restoration quality of the multi-frame blind deconvolution algorithm on the image, and can realize accurate and quick image restoration under the single-desktop single-GPU condition.
In addition, the embodiment also discloses a computer device, which comprises:
at least one processor;
and a memory communicatively coupled to the at least one processor;
wherein the memory is used for storing a computer program;
the processor is configured to implement the image quick restoration method as described above when executing the computer program.
In addition, the embodiment also discloses a computer readable storage medium, and the computer readable storage medium stores a computer program, and the computer program realizes the image quick restoration method when being executed by a processor.
The above description is only of the preferred embodiments of the present application and is not intended to limit the present application, but various modifications and variations can be made to the present application by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application.
Claims (10)
1. The image quick restoration method based on multi-frame blind deconvolution is characterized by comprising the following steps:
step S100: registering an observation image, specifically, reading in the observation image, and registering the observation image;
step S200: the initialization setting, namely, transmitting the registered observation image into the GPU from the CPU to perform the initialization setting, wherein the initialization setting comprises setting the initial value of a target image and a point spread function, constructing a cost function, selecting a parameterization method and an optimization algorithm;
step S300: the method for recovering the image by multi-frame blind deconvolution specifically comprises the following steps:
step S301, judging whether the target image reaches the standard, specifically, checking whether the target image reaches the standard or whether the restoration process reaches the maximum restoration times; if the target image reaches the standard or the restoration process reaches the maximum restoration times, transmitting the point spread function and the target image from the GPU to the CPU, and outputting the point spread function and the target image in an image form; if the target image does not reach the standard and the restoration process does not reach the maximum restoration times, the step S302 is entered;
step S302: optimizing a target image and a point spread function, namely, keeping the target image unchanged, and optimizing the point spread function by using a macopt-based optimization conjugate gradient method; maintaining the point spread function unchanged, optimizing the target image by using a macopt-based optimized conjugate gradient method, and entering step S303;
step S303: the number of restoration times is updated, specifically, the number of restoration times is increased by 1, and the process returns to step S301.
2. The method according to claim 1, wherein in step S100, the observed image is subjected to translational registration by using a fourier transform-based phase correlation algorithm on the basis of removing background and flat fields, and the specific procedure is as follows:
calculate the first frame observation image g 1 Fourier transform G of (2) 1 Then calculate the Fourier transform G of the w-th frame observation image w The normalized power spectrum R (u, v) of the first and w-th frame images is shown in formula 1):
wherein, (x) 0 ,y 0 ) Representing the amount of translation between the first frame and the w frame; * Represents conjugation; (u, v) represents a spatial frequency; j is an imaginary unit;
performing fourier inverse transformation on formula 1) to obtain formula 2):
r(x,y)＝FFT -1 (R)＝δ(x-x 0 ,y-y 0 ) 2)
wherein FFT -1 Representing an inverse fourier transform; r (x, y) is the inverse fourier transform of the normalized power spectrum R (u, v); delta (x-x) 0 ,y-y 0 ) The representation center is located at (x) 0 ,y 0 ) Is a dirac function of (c); equation 2) represents a function containing sharp pulses whose positions represent the magnitude of the translation values based on which the translation registration of the observed image is achieved.
3. The image quick restoration method according to claim 2, wherein in step S200, the expression of the cost function is as follows:
wherein J (x, y) represents a cost function; sigma (sigma) 2 Representing the variance of observed image noise; g w (x, y) represents a w-th frame observation image; f (x, y) represents a target image; h is a w (x, y) represents a point spread function of the w-th frame observation image;representing a convolution operation;/>to derive a symbol; />Representation->Is a binary norm of (2); lambda (lambda) 0 Is a regularization coefficient.
4. The method of image quick recovery according to claim 3, wherein in step S200, the initial value of the target image is an average value of the observed image, and the initial value of the point spread function is an autocorrelation of the observed image, expressed as follows:
wherein f 0 (x, y) represents an initial value of the target image, W represents the number of observed image frames,representing the initial value of the point spread function of the w-th frame, for example>Representing a convolution calculation.
5. The method according to claim 4, wherein in step S200, the parameterization method is used to parameterize the target image and the point spread function, and the expression is as follows:
wherein h (x, y) represents a point spread function;representing the parameterized variable of the target image; phi (phi) 2 (x, y) represents the parameterized variable of the point spread function.
6. The method for rapid restoration of an image according to claim 5, wherein in step S301, it is checked whether the target image meets the standard or whether the restoration process reaches the maximum restoration number, specifically:
setting the restoration frequency k, wherein the initial value is 1, adding 1 after completing one image restoration, setting the maximum restoration frequency as k_maxiter, setting the gradient convergence standard epsilon, checking whether k > k_maxiter exists or whether the norm of the cost function gradient is smaller than epsilon, and if one of the two conditions is met, continuing the next image restoration.
7. The method for rapid image restoration according to claim 6, wherein the cost function gradient y k The expression of (2) is as follows:
wherein a is k And b k Is a range of two endpoints, s, containing minima k And t k A is respectively k And b k Is a direction discrimination value of (a).
8. The method for rapid image restoration according to claim 7, wherein in step S302, the method for optimizing the point spread function and the target image by using the conjugate gradient method is the same, and the specific procedure is as follows:
point spread function h for kth image restoration k And the target image f k Collectively denoted as x in the respective optimization process k The objective of the optimization is to minimize the cost function, namely:
min{J(x k )|x k ∈R n } 7)
the optimized conjugate gradient algorithm based on macopt is to find the x which minimizes the cost function by an iterative method k The expression is as follows:
x k+1 ＝x k +α k d k 8)
wherein d k Represents the search direction, alpha, of the kth iteration k A search step representing the kth iteration;
solving for d k The expression of (2) is:
beta in formula 9) k Representing the conjugation direction, the expression is:
wherein T represents transposition;
in formula 6), s k And t k The calculation formula of (2) is as follows:
for two end points a k And b k Solving by adopting an iterative method, setting the upper limit of the optimization frequency of the iterative optimization process as i_maxiter, recording the optimization frequency i, and if the initial value is 1, a k And b k The calculation formula is as follows:
a k ＝x k +2 i+1 d k l k 13)
b k ＝x k +2 -i-1 d k l k 14)
wherein l k In relation to the search step, the expression is:
wherein m represents the single-side pixel number of the image, and m is the total pixel number of the image; the kth iteration step alpha k Calculated from the following formula:
in formulae 13) and 14), if s k t k <0, then a k And b k A minimum value exists between the two, and searching is stopped; otherwise, the iteration number is increased by 1, and a) is continued according to the formulas 13) and 14) k And b k Searching until s occurs k t k <0, or the maximum searching times i_maxiter is reached, stopping searching;
thus far, the point spread function h of the kth image restoration is completed k Or a target image f k Is described.
9. A computer device, comprising:
at least one processor;
and a memory communicatively coupled to the at least one processor;
wherein the memory is used for storing a computer program;
the processor is configured to implement the image quick restoration method according to any one of claims 1-8 when executing the computer program.
10. A computer readable storage medium, wherein a computer program is stored on the computer readable storage medium, the computer program implementing the image quick restoration method according to any one of claims 1 to 8 when executed by a processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311078013.6A CN117115025B (en) | 2023-08-25 | 2023-08-25 | Image rapid restoration method, device and medium based on multiframe blind deconvolution |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311078013.6A CN117115025B (en) | 2023-08-25 | 2023-08-25 | Image rapid restoration method, device and medium based on multiframe blind deconvolution |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117115025A true CN117115025A (en) | 2023-11-24 |
| CN117115025B CN117115025B (en) | 2024-04-05 |
Family
ID=88803385
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311078013.6A Active CN117115025B (en) | 2023-08-25 | 2023-08-25 | Image rapid restoration method, device and medium based on multiframe blind deconvolution |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117115025B (en) |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6470097B1 (en) * | 1999-01-22 | 2002-10-22 | Siemens Corporation Research, Inc. | Total variational blind image restoration from image sequences |
| JP2008165657A (en) * | 2006-12-29 | 2008-07-17 | Seiko Epson Corp | Image restoration device, image restoration method, and image restoration program |
| CN102063716A (en) * | 2011-01-13 | 2011-05-18 | 耿则勋 | Multiframe iteration blind deconvolution image restoration method based on anisotropic constraint |
| CN102208100A (en) * | 2011-05-31 | 2011-10-05 | 重庆大学 | Total-variation (TV) regularized image blind restoration method based on Split Bregman iteration |
| CN106530251A (en) * | 2016-11-03 | 2017-03-22 | 北京航星机器制造有限公司 | Blind convolutional motion fuzzy image restoration method |
| CN107369134A (en) * | 2017-06-12 | 2017-11-21 | 上海斐讯数据通信技术有限公司 | A kind of image recovery method of blurred picture |
| CN107680040A (en) * | 2017-09-26 | 2018-02-09 | 北京空间飞行器总体设计部 | A kind of blind convolution super-resolution reconstruction method of multiframe based on bayesian criterion and device |
| CN107817493A (en) * | 2017-10-25 | 2018-03-20 | 电子科技大学 | A kind of 2D near fields diameter radar image method for reconstructing based on compressed sensing |
| US20200175731A1 (en) * | 2018-11-30 | 2020-06-04 | Canon Medical Systems Corporation | Method and apparatus for acceleration of iterative reconstruction of a computed tomography image |
| CN113379647A (en) * | 2021-07-08 | 2021-09-10 | 湘潭大学 | Multi-feature image restoration method for optimizing PSF estimation |
| CN114037636A (en) * | 2021-11-29 | 2022-02-11 | 无锡学院 | Multi-frame blind restoration method for correcting image by adaptive optical system |
| CN114972081A (en) * | 2022-05-12 | 2022-08-30 | 浙江大学 | Blind restoration-based image restoration method under complex optical imaging condition |
-
2023
- 2023-08-25 CN CN202311078013.6A patent/CN117115025B/en active Active
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6470097B1 (en) * | 1999-01-22 | 2002-10-22 | Siemens Corporation Research, Inc. | Total variational blind image restoration from image sequences |
| JP2008165657A (en) * | 2006-12-29 | 2008-07-17 | Seiko Epson Corp | Image restoration device, image restoration method, and image restoration program |
| CN102063716A (en) * | 2011-01-13 | 2011-05-18 | 耿则勋 | Multiframe iteration blind deconvolution image restoration method based on anisotropic constraint |
| CN102208100A (en) * | 2011-05-31 | 2011-10-05 | 重庆大学 | Total-variation (TV) regularized image blind restoration method based on Split Bregman iteration |
| CN106530251A (en) * | 2016-11-03 | 2017-03-22 | 北京航星机器制造有限公司 | Blind convolutional motion fuzzy image restoration method |
| CN107369134A (en) * | 2017-06-12 | 2017-11-21 | 上海斐讯数据通信技术有限公司 | A kind of image recovery method of blurred picture |
| CN107680040A (en) * | 2017-09-26 | 2018-02-09 | 北京空间飞行器总体设计部 | A kind of blind convolution super-resolution reconstruction method of multiframe based on bayesian criterion and device |
| CN107817493A (en) * | 2017-10-25 | 2018-03-20 | 电子科技大学 | A kind of 2D near fields diameter radar image method for reconstructing based on compressed sensing |
| US20200175731A1 (en) * | 2018-11-30 | 2020-06-04 | Canon Medical Systems Corporation | Method and apparatus for acceleration of iterative reconstruction of a computed tomography image |
| CN113379647A (en) * | 2021-07-08 | 2021-09-10 | 湘潭大学 | Multi-feature image restoration method for optimizing PSF estimation |
| CN114037636A (en) * | 2021-11-29 | 2022-02-11 | 无锡学院 | Multi-frame blind restoration method for correcting image by adaptive optical system |
| CN114972081A (en) * | 2022-05-12 | 2022-08-30 | 浙江大学 | Blind restoration-based image restoration method under complex optical imaging condition |
Non-Patent Citations (4)
| Title |
|---|
| DONG WENDE ET AL.: "Blind Deconvolution for Poissonian Blurred Image With Total Variation and L0-Norm Gradient Regularizations", IEEE TRANSACTIONS ON IMAGE PROCESSING, vol. 30, 24 November 2020 (2020-11-24), pages 1030 - 1043, XP011825635, DOI: 10.1109/TIP.2020.3038518 * |
| 卜欣彤: "基于PSF优化估计的图像盲复原方法研究", 中国优秀硕士学位论文全文数据库 (信息科技辑), 15 May 2019 (2019-05-15), pages 138 - 1451 * |
| 姜南李威 等: "引入泽尼克多项式的极大似然盲解卷积算法初值的优化选取", 光学与光电技术, vol. 12, no. 06, 10 December 2014 (2014-12-10), pages 25 - 29 * |
| 张镇韬: "基于GPU的多帧盲解卷积图像复原并行化研究", 中国优秀硕士学位论文全文数据库 (信息科技辑), 15 April 2018 (2018-04-15), pages 138 - 3256 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117115025B (en) | 2024-04-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP1801754B1 (en) | Degradation information restoring method and device | |
| Lagendijk et al. | Maximum likelihood image and blur identification: a unifying approach | |
| Xu et al. | A fast patch-dictionary method for whole image recovery | |
| EP1801753B1 (en) | Degradation information restoring method and device | |
| CN111047544B (en) | Saturated image deblurring method based on nonlinear degradation model | |
| US20050147313A1 (en) | Image deblurring with a systolic array processor | |
| CN113240596A (en) | Color video recovery method and system based on high-order tensor singular value decomposition | |
| CN110809126A (en) | Video frame interpolation method and system based on adaptive deformable convolution | |
| CN107424119B (en) | Super-resolution method of single image | |
| Ng et al. | A fast MAP algorithm for high-resolution image reconstruction with multisensors | |
| CN107451961B (en) | Method for recovering sharp image under multiple fuzzy noise images | |
| CN117115025B (en) | Image rapid restoration method, device and medium based on multiframe blind deconvolution | |
| Ng et al. | A new approach to constrained total least squares image restoration | |
| Chen et al. | Efficient discrete spatial techniques for blur support identification in blind image deconvolution | |
| CN115311164B (en) | Synthetic beam effect elimination method for sparse interference array observation imaging | |
| Chuan et al. | Box-constrained total-variation image restoration with automatic parameter estimation | |
| CN107644406B (en) | Image denoising method based on improved orthogonal matching pursuit | |
| CN107895387B (en) | MRI image reconstruction method and device | |
| CN112712486B (en) | Radio astronomical image reconstruction method based on activation function | |
| CN112422133B (en) | Binary sparse signal recovery method for subtraction matching pursuit and application thereof | |
| CN115021759A (en) | Binary sparse signal recovery method and system based on binary least square method | |
| Sottek et al. | An efficient approach to extrapolation and spectral analysis of discrete signals | |
| Sahnoun et al. | Sparse multiresolution modal estimation | |
| CN112991140B (en) | Envelope alignment rapid implementation method for GPU parallel acceleration | |
| Mourad et al. | Automatic threshold estimation for iterative shrinkage algorithms used with compressed sensing |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |