CN107220945B - Restoration method of multiple degraded extremely blurred image - Google Patents

Restoration method of multiple degraded extremely blurred image Download PDF

Info

Publication number
CN107220945B
CN107220945B CN201710234684.5A CN201710234684A CN107220945B CN 107220945 B CN107220945 B CN 107220945B CN 201710234684 A CN201710234684 A CN 201710234684A CN 107220945 B CN107220945 B CN 107220945B
Authority
CN
China
Prior art keywords
sub
degradation
degradation function
function
image
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.)
Expired - Fee Related
Application number
CN201710234684.5A
Other languages
Chinese (zh)
Other versions
CN107220945A (en
Inventor
刘丹平
何小敏
毛菀丁
王贤秋
胡小波
谭晓衡
印勇
胡学斌
蒋阳
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.)
Chongqing University
Original Assignee
Chongqing 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 Chongqing University filed Critical Chongqing University
Priority to CN201710234684.5A priority Critical patent/CN107220945B/en
Publication of CN107220945A publication Critical patent/CN107220945A/en
Application granted granted Critical
Publication of CN107220945B publication Critical patent/CN107220945B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/73—Deblurring; Sharpening
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/20—Image enhancement or restoration using local operators
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/77—Retouching; Inpainting; Scratch removal
    • 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
    • G06T2207/10—Image acquisition modality
    • G06T2207/10016—Video; Image sequence

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Image Processing (AREA)

Abstract

The invention discloses a method for restoring a multiply-degraded extremely-blurred image, which is used for solving the problem that the conventional criminal investigation software cannot effectively restore the blurred image. The method mainly takes multiple fuzzy factors caused by optical diffraction, quantization, defocusing and relative movement into consideration, initializes each sub-degradation function according to actually measured physical parameters, solves the optimal distribution of each sub-degradation function by using a variational Bayes algorithm in an alternate iteration mode, and synthesizes each sub-degradation function into a point spread function of a system by using the relationship between the sub-degradation functions and the system degradation function. And finally, restoring the image by utilizing an L-R algorithm.

Description

Restoration method of multiple degraded extremely blurred image
Technical Field
The invention belongs to the field of image restoration, relates to a restoration method of a multiple degraded extremely-blurred image, and is particularly suitable for restoration of an extremely-blurred criminal investigation image.
Background
With the gradual expansion of the application range, the video monitoring technology has become an important means for collecting crime evidences and extracting crime clues in the process of detecting various cases [1-4 ]. When the image is blurred, the Investigator uses Photoshop, police Video analysis platform, Video investor, "doctor shadow", "VCS", vReveal Premium, and "Sobel" software to process [1 ]. Generally, these software are effective for blurred images due to a single degradation cause, but it is difficult to obtain an ideal effect for extremely blurred images due to multiple degradations.
Early methods typically assumed PSFs to be of simple parametric form. In practice, PSFs are much more complex and should be described with a complex parametric model. The PSF is regarded as a matrix called fuzzy matrix by Dilip G.Warrier and Uday B.Desai, elements of the matrix adopt a random model, and a closed form fuzzy element average value expression is obtained by using average field approximation. Thus the PSF estimation algorithm using the non-functional model is developed rapidly: the early estimation method based on image complex cepstrum, the estimation method based on two-dimensional ARMA model, the multi-frame image sequence wiener filtering method and the estimation method based on singular value decomposition which appears recently, etc.
Recently, there are many good algorithms for estimating the PSF based on the statistical characteristics of natural images. Fergus et al used a mixed Gaussian model for the gray scale gradients of natural images, combined with variational Bayes to estimate the PSF, and then deconvolved using an L-R algorithm. This approach works better for smaller PSFs. Later, Krishan and Fergus adopt a super Laplacian prior model for the heavy tail distribution of the gray gradient of a natural image, and apply a cross iteration scheme, wherein the scheme utilizes an algorithm of a lookup table to carry out rapid optimization, but the restored image still has an obvious step effect. Levin et al generalize these methods and propose an optimization method of edge likelihood based on maximum a posteriori probability that proves to be more robust than the usual a posteriori probability. Although these methods work well, the amount of computation is too large. Because the latent image needs to be marginalized, optimizing the energy function usually requires a rather complex iterative numerical algorithm, such as alternating iterations of PSF estimation and image restoration in the optimization scheme. However, if the initial blur kernel is not well set in a matching manner or with an appropriate size, it will generally not converge to a true global minimum.
Disclosure of Invention
The method for restoring the extremely-blurred image with multiple degradations is provided by taking a test value of a scene as an initial value, and each sub-degradation function is accurately estimated by applying a variational Bayes theory, so that the PSF of an imaging system is accurately estimated, and the restoration of the extremely-blurred criminal investigation image is realized. Due to careful treatment of multiple degradation factors, the image restoration effect is good.
A restoration method of a multiple-degraded extremely-blurred image, S1, acquiring data;
s01, acquiring external parameters:
s02, obtaining a matrix g (x, y) of the blurred image and solving a gradient cepstrum C of the matrix g▽g；
S2, establishing a system degradation function h (x, y) and each sub degradation function;
s3, initializing each sub-degradation function, and solving initial gradient cepstrum of each sub-degradation function;
s4, taking the initial gradient cepstrum of each sub-degradation function and the gradient cepstrum of the fuzzy image as the input of variational Bayes, and solving the optimal distribution of each sub-degradation function;
and S5, solving the point spread function h (x, y) of the system by utilizing the relation between each sub-degradation function and the point spread function of the system.
And S6, taking h (x, y) and g (x, y) as input of the L-R algorithm, and obtaining a clear image matrix f (x, y).
Further, the acquiring of the external parameter in S01 specifically includes:
s01, acquiring external parameters: obtaining the working parameters of the camera, such as focal length F, diameter D of the lens, aperture coefficient F and CCD imaging pixel size wx、wyAnd the object distance z, the image distance v, the number of pixel points K, P, the length l of each pixel point of the photosensitive element, the movement distance d of the object within the exposure time tau, and the movement direction theta of the target.
Further, it is defined that the respective degradation functions of S2 specifically include:
the sub-degradation function of the airy disk mode caused by optical diffraction is h1(x, y);
the sub-degradation function of the image sensor at quantization is h2(x, y);
the sub-degradation function of camera defocus is h3(x, y);
the sub-degradation function of the camera moving linearly at a constant speed relative to the target is h4(x, y).
Further defining, the sub-degradation function h1(x, y) of the airy disk mode caused by optical diffraction is specifically:
Figure GDA0002586312070000031
the sub-degradation function of the image sensor during quantization is h2(x, y) specifically:
Figure GDA0002586312070000032
the sub-degradation function of camera defocus is h3(x, y) specifically:
Figure GDA0002586312070000033
the sub-degradation function of the camera and the target which move linearly relatively at a constant speed is h4(x, y), which is specifically as follows:
Figure GDA0002586312070000034
further, the S3 specifically includes:
substituting the parameter values obtained in the step S01 into the corresponding sub-degradation functions in the step S2 to initialize each sub-degradation function, so as to obtain initial values h10, h20, h30 and h40 of each sub-degradation function;
solving initial gradient cepstrum C of each sub-degradation functionh10、Ch20、Ch30、Ch40And the gradient cepstrum of the system degradation function h (x, y) is set as Ch。
Further, the S4 specifically includes:
cepstrum C of the initial gradients of the respective sub-degenerate functionsh10、Ch20、Ch30、Ch40Ch40、C▽gObtaining C as an input to variational Bayesh1、Ch2、Ch3、Ch4；
They satisfy the linear relationship:
Ch＝k1Ch1+k2Ch2+k3Ch3+k4Ch4wherein k is1、k2、k3、k4Is constant, andk1、k2、k3、k4at least one is not 0.
Further defined, the k parameter k1、k2、k3、k4A typical value is 1, and for a certain imaging system degradation situation, the value of the k parameter changes around 1, and the specific value is determined by experiments. The patent is not limited to four degradation factors, and if there are more than four degradation factors, the number of k parameters should be increased.
The invention has the beneficial effects that: by considering multiple degradation factors of image blur and accurately estimating each sub degradation function through the variational Bayes theory, the PSF of the imaging system is accurately estimated, and the restoration of the extremely blurred image is realized.
Detailed Description
Description of the drawings:
FIG. 1 is a schematic flow chart of the present invention.
Fig. 2 is an imaging optical path diagram of the camera.
Fig. 3 is a photographic picture of the camera during clear photographing.
Fig. 4 is a frame diagram of a blurred video of the camera.
FIG. 5 is a clear picture after processing by the method of the present invention
As shown in fig. 1, a method for restoring an extremely blurred image with multiple degradation factors includes the following steps:
s1, acquiring data;
s01, acquiring external parameters: obtaining the working parameters of the camera, such as focal length F, diameter D of the lens, aperture coefficient F and CCD imaging pixel size omegax、ωyThe object distance z, the image distance v, the number of pixel points K, P, the length l of each pixel point of the photosensitive element, the movement distance d of the object within the exposure time tau and the movement direction theta of the target;
s02, acquiring a blurred image matrix g (x, y);
s2, establishing a degradation function,
the system degradation function is h (x, y);
the sub-degradation function of the airy disk mode caused by optical diffraction is h1(x, y);
the sub-degradation function of the image sensor at quantization is h2(x, y);
the sub-degradation function of camera defocus is h3(x, y);
the sub-degradation function of the camera and the target which move linearly relatively at a constant speed is h4(x, y);
the sub-degradation function h1(x, y) of the airy disk mode caused by optical diffraction is specifically:
Figure GDA0002586312070000041
the sub-degradation function of the image sensor during quantization is h2(x, y) specifically:
Figure GDA0002586312070000051
the sub-degradation function of camera defocus is h3(x, y) specifically:
Figure GDA0002586312070000052
the sub-degradation function of the camera and the target which move linearly relatively at a constant speed is h4(x, y), which is specifically as follows:
Figure GDA0002586312070000053
s3, substituting the parameter values obtained in the S01 into the corresponding sub-degeneration functions in the S2 to initialize each sub-degeneration function, and obtaining initial values h10, h20, h30 and h40 of each sub-degeneration function;
solving initial gradient cepstrum C of each sub-degradation functionh10、Ch20、Ch30、Ch40The gradient cepstrum with the system degradation function h (x, y) is Ch；
Cepstrum C of the initial gradients of the respective sub-degenerate functionsh10、Ch20、Ch30、Ch40、CgObtaining C as an input to variational Bayesh1、Ch2、Ch3、Ch4；
They satisfy the linear relationship:
Ch＝k1Ch1+k2Ch2+k3Ch3+k4Ch4wherein k is1、k2、k3、k4Is a constant, and said k1、k2、k3、k4At least two are not 0, and the values are obtained by experiments.
And S4, taking h (x, y) and g (x, y) as input of the L-R algorithm, and obtaining a clear image matrix f (x, y).
The specific principle is as follows:
the degraded image g (x, y) of the sharp image f (x, y) can be expressed as:
Figure GDA0002586312070000054
where represents convolution operation, h (x, y) is the system Point Spread Function (PSF) upsilon (x, y) is the system noise. The gradient of the natural image f (x, y) follows a heavy-tailed distribution, whose distribution is represented by a mixed gaussian distribution:
Figure GDA0002586312070000055
wherein C is the number of Gaussian mixture distribution, picWeight, v, representing the c-th distributioncIs the variance of the c-th distribution. For the later computational aspect, the image is subjected to gradient cepstrum processing, and the cepstrum of the blurred image g (x, y) is defined as:
Cg(p,q)＝FFT-1{log[1+|G(u，v)|]} (3)
and (4) respectively carrying out gradient processing on the formula (3) in the x direction and the y direction to obtain a gradient cepstrum of the blurred image.
The system takes the image degradation caused by factors such as optical diffraction, quantification, defocusing, relative movement and the like into consideration, and constructs a model of a point spread function caused by multiple degradation factors. Assuming that the sub-degradation functions of the imaging process are denoted as h1, h2, h3, h4, the degraded image g (x, y) under the combined action of multiple degradations and noise v (x, y) is:
Figure GDA0002586312070000061
here, v (x, y) is zero-mean white gaussian noise, and h1(x, y) represents a sub-degradation function of the airy disk mode caused by optical diffraction:
Figure GDA0002586312070000062
wherein the content of the first and second substances,
Figure GDA0002586312070000063
j1() is a first order Bessel function; r is0The distance from the center of the airy disk to the main lobe; b ═ pi F/F (F is the lens focal length, F is the aperture factor).
h2(x, y) represents the sub-degradation function of the rectangular sensing cell when quantized:
Figure GDA0002586312070000064
wherein ω x and ω y are respectively expressed as the length and width of the rectangular pixel.
h3(x, y) represents the sub-degradation function at defocus:
Figure GDA0002586312070000065
where R is the defocus spot radius, expressed as:
R＝(1/f-1/v-1/z)Dv/2 (8)
wherein f is the focal length of the lens, z is the object distance, D is the radius of the convex lens, and v is the image distance.
h4(x, y) represents a sub-degradation function when the imaging system moves linearly relative to the target at a constant speed:
Figure GDA0002586312070000066
where θ is the blur angle and L is the length of the degradation function. And the theta is estimated by analyzing the target motion condition of the video frame, and the L can be actually measured on site according to the imaging proportional relation.
In a single frame image, the motion blur length is the distance of motion of the vehicle when exposed, and the motion time is the exposure time. Calculating the corresponding length of the motion blur length on the photosensitive element according to an imaging proportional relation method, wherein the principle is as follows:
as shown in the imaging optical path diagram of fig. 2, d is the distance that the vehicle moves when being photographed, i.e., the length of the motion blur, the included angle with the focal plane is α, k is the length of the motion blur corresponding to the length d of the motion blur on the photosensitive element, on which the distance between the motion blur edge and the normal is p, the focal length is f, and the object distance is z. According to the similar triangle proportion relation, the following can be obtained:
Figure GDA0002586312070000071
since k and p are length units, but the length is actually obtained by counting the number of pixels, let K, P be the actual number of pixels, and the length of each pixel of the photosensitive element is l, so k equals kl and p equals Pl.
Let us assume cepstrum C of h, h1, h2, h3, h4h、Ch1、Ch2、Ch3、Ch4Then the following relationship exists:
Ch＝k1Ch1+k2Ch2+k3Ch3+k4Ch4(11)
wherein k is1、k2、k3、k4Is a constant. For blurred images caused by four degradation factors, the k parameter k1、k2、k3、k4A typical value is 1, and for a certain imaging system degradation situation, the value of the k parameter changes around 1, and the specific value is determined by experiments. The patent is not limited to four degradation factors, if there are more than four degradation factorsThe number of k parameters should be increased and the estimation method is still effective.
As can be seen from the equation (1), the process of recovering a clear image when the system point spread function is unknown is blind deconvolution. Therefore, two steps are taken for image restoration. First, the point spread function of the system is estimated. Secondly, the image is restored by using an L-R algorithm. And (4) estimating a fuzzy kernel with higher precision by using a variational Bayesian theory.
Interpretation of point spread function of variational Bayesian estimation system:
given a blurred image g (x, y), the PSF (system point spread function) and sharp image f (x, y) are estimated by finding the maximum a posteriori probability given the prior information of f (x, y).
For equation (4), it is cumbersome to directly calculate the posterior distribution p (Θ | g) assuming the hidden variables Θ ═ f, h1, h2, h3, h4, and VB approximates the posterior distribution p (Θ | g) with a simpler distribution q (Θ | g) whose KL divergence is:
Figure GDA0002586312070000081
q (Θ | g) is solved by minimizing equation (12). Since the integral variables are Θ and p (g) is constant, equation (12) can be expressed as:
Figure GDA0002586312070000082
definition CKLAs a cost function, it can be expressed as:
Figure GDA0002586312070000083
assuming that q (Θ) and q (g) are independent of each other, q (Θ, g) is q (Θ) q (g), that is, formula (14) can be rewritten as:
Figure GDA0002586312070000084
by substituting Θ into equation (15), we can obtain:
Figure GDA0002586312070000085
assuming that the sub-degenerate functions are independent of each other, then
q(f,h1.h2,h3,h4)＝q(f)q(h1)q(h2)q(h3)q(h4) (17)
Bringing (17) into (16) yields:
Figure GDA0002586312070000086
further derivation of equation (18) yields:
CKL＝∫q(f)q(h1)q(h2)q(h3)q(h4)[lnq(f)+lnq(h1)+lnq(h2)+lnq(h3)+lnq(h4)-lnp(f)-lnp(h1)-lnp(h2)-lnp(h3)-lnp(h4)-lnp(g|f,h1.h2,h3,h4)]dΘ (19)
when the cost function (19) is minimized, the solution is carried out by adopting an alternate iteration method, and when a variable is solved, the rest variables are assumed to be constant.
In solving for q (h1), assuming q (f), q (h2), q (h3), and q (h4) are known constants, let:
Figure GDA0002586312070000087
assuming that the noise in the degradation model is white gaussian noise with intensity σ 2, then there are:
Figure GDA0002586312070000088
where N () represents a gaussian distribution, we can obtain:
Figure GDA0002586312070000091
equation (19) then integrates only h1 and the rewritable cost function is as follows:
Figure GDA0002586312070000092
adding Lagrange to (22) according to the constraint condition ^ h1dh1 ═ 1Lambd multiplierh1And (5) obtaining an extremum of the cost function to obtain q (h 1):
Figure GDA0002586312070000093
it can be seen that q (h1) is a function of f, h2, h3, h4, and therefore requires alternate updates to iterate. The same can be obtained:
Figure GDA0002586312070000094
wherein the content of the first and second substances,
Figure GDA0002586312070000095
Figure GDA0002586312070000096
wherein the content of the first and second substances,
Figure GDA0002586312070000097
Figure GDA0002586312070000098
wherein the content of the first and second substances,
Figure GDA0002586312070000099
therefore, each sub-degradation function value obtained through field test is used as an initial value, and the optimal distribution of each sub-degradation function can be obtained by using the equations (23) - (26) and adopting an alternate iteration method, so that the accurate value of each sub-degradation function can be obtained through calculating mathematical expectation.
Just because of the above relationship, the method for estimating the variational Bayes sub-degradation function needs to be carried out in the gradient cepstrum domain, and the method estimates each sub-degradation function Ch1、Ch2、Ch3、Ch4Then using the point spread function C of the formula (11) synthesis systemhAnd obtaining a system point spread function h. Finally, by utilizing the L-R algorithm,a sharp image f (x, y) is obtained.
The method comprises the following specific implementation steps:
firstly, obtaining parameters of focal length F, diameter D of lens, aperture coefficient F and CCD imaging pixel size wx、wyAnd the object distance z, the image distance v, the number of pixel points K, P, the length l of each pixel point of the photosensitive element, the movement distance d of the object within the exposure time tau, and the movement direction theta of the target.
And secondly, obtaining initial values h10, h20, h30 and h40 of each sub-degradation function by using the function modes of the formulas (5) to (10).
Thirdly, solving initial gradient cepstrum C of each sub-degradation functionh10、Ch20、Ch30、Ch40And gradient cepstrum C of the blurred image▽g。
The fourth step, with Ch10、Ch20、Ch30、Ch40、C▽gTaking the initial value and utilizing the super Laplacian distribution of the gray gradient of the natural image, and obtaining C through variational Bayesh1、Ch2、Ch3、Ch4。
And fifthly, obtaining a point spread function of the system by using the formula (11).
And sixthly, solving the distribution of the point spread function of the system in the time domain, and calculating the expectation to obtain the point spread function of the system.
And seventhly, restoring by using an L-R algorithm to obtain a clear image f (x, y).
For the investigation personnel, the point spread function estimated by the variational Bayes and the principle of the L-R restoration algorithm do not need to be understood, and the extremely-blurred image restoration system can be used by adopting the following steps:
(1) for an extremely blurred image obtained by monitoring, parameters of focal length F, diameter D of lens, aperture coefficient F and CCD imaging pixel size omega are inputx、ωyAnd the object distance z, the image distance v, the number of pixel points K, P, the length l of each pixel point of the photosensitive element, the movement distance d of the object within the exposure time tau, and the movement direction theta of the target.
(2) These parameters and the blurred image are input to the system, and a restored image is obtained by the restoration system.
In the process, the precision of the testing parameters F, D, F, z, v and D is required to reach 0.1mm, the precision of tau is required to reach 0.001s, and the precision of theta is required to reach 0.1 degree. The parametric testing requires rigorous procedures. The same result can be obtained if different persons run the software following this step.
As shown in fig. 4, a blurred video with a frame image in the region of 5:46 pm on 3/15/2017 is read from the hard disk of the monitoring system.
From the product specification, we know that the focal length F of the lens of the monitoring system is 6mm, the aperture coefficient F is 1.4, the size ω x of the imaging pixel is 6.4 μm, and the exposure time τ of the image sensor is 0.01 s. These vehicles were measured in situ at a distance of about 108m from the lens. The k parameters tested by the experiments are specifically: k1 ═ 1.2; k2 ═ 0.9; k3 ═ 0.8; k4 is 1.3.
The results obtained by the patented process are shown in fig. 5, where the signal-to-noise ratio is improved by about 9 dB.

Claims (4)

1. A restoration method of a multiple-degraded extremely blurred image is characterized in that:
s1, acquiring data:
s01, acquiring external parameters;
s02, acquiring a blurred image matrix g (x, y), and solving a gradient cepstrum of the blurred image matrix g (x, y);
s2, establishing a system degradation function h (x, y) and each sub degradation function;
each sub-degradation function of S2 specifically includes:
the sub-degradation function of the airy disk mode caused by optical diffraction is h1(x, y);
the sub-degradation function of the image sensor at quantization is h2(x, y);
the sub-degradation function of camera defocus is h3(x, y);
the sub-degradation function of the camera and the target which move linearly relatively at a constant speed is h4(x, y);
s3, initializing each sub-degradation function and obtaining an initial value of each sub-degradation function; solving initial gradient cepstrum of each sub-degradation function;
s3 specifically includes:
substituting the parameter values obtained in the step S01 into the corresponding sub-degradation functions in the step S2 to initialize each sub-degradation function, so as to obtain initial values h10, h20, h30 and h40 of each sub-degradation function;
solving initial gradient cepstrum C of each sub-degradation functionh10、Ch20、Ch30、Ch40And the gradient cepstrum with a system degradation function h (x, y) is set as Ch；
S4, taking the initial gradient cepstrum of each sub-degradation function and the gradient cepstrum of the fuzzy image as the input of variational Bayes, and solving the optimal distribution of each sub-degradation function;
the S4 specifically includes:
cepstrum C of the initial gradients of the respective sub-degenerate functionsh10、Ch20、Ch30、Ch40、C▽gObtaining C as an input to variational Bayesh1、Ch2、Ch3、Ch4；
S5, solving a point spread function h (x, y) of the system by utilizing the relation between each sub-degradation function and the point spread function of the system;
each sub-degradation function satisfies a linear relationship:
Ch＝k1Ch1+k2Ch2+k3Ch3+k4Ch4wherein k is1、k2、k3、k4Is a constant, and k1、k2、k3、k4At least one is not 0;
and S6, taking h (x, y) and g (x, y) as input of the L-R algorithm, and obtaining a clear image matrix f (x, y).
2. A method for restoring a multiple degraded extremely blurred image according to claim 1, wherein:
the acquiring of the external parameter in S01 specifically includes:
s01, acquiring external parameters: obtaining the working parameters of the camera, such as focal length F, diameter D of the lens and aperture coefficient F,CCD imaging pixel size wx、wyAnd the object distance z, the image distance v, the number of pixel points K, P, the length l of each pixel point of the photosensitive element, the movement distance d of the object within the exposure time tau, and the movement direction theta of the target.
3. A method for restoring a multiple degraded extremely blurred image according to claim 1, wherein:
among the sub-degradation functions of S2, the sub-degradation function h1(x, y) of the airy disk mode caused by optical diffraction is specifically:
Figure FDA0002586312060000021
the sub-degradation function of the image sensor during quantization is h2(x, y) specifically:
Figure FDA0002586312060000022
the sub-degradation function of camera defocus is h3(x, y) specifically:
Figure FDA0002586312060000023
the sub-degradation function of the camera and the target which move linearly relatively at a constant speed is h4(x, y), which is specifically as follows:
Figure FDA0002586312060000024
。
4. a method for restoring a multiple degraded extremely blurred image according to claim 1, wherein:
for blurred images caused by four degradation factors, k parameter k1、k2、k3、k4The typical value is 1, the value of the k parameter changes near 1 according to the degradation condition of the imaging system, and when the degradation factors are more than four, the number of the k parameters is correspondingly increased.
CN201710234684.5A 2017-04-12 2017-04-12 Restoration method of multiple degraded extremely blurred image Expired - Fee Related CN107220945B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710234684.5A CN107220945B (en) 2017-04-12 2017-04-12 Restoration method of multiple degraded extremely blurred image

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710234684.5A CN107220945B (en) 2017-04-12 2017-04-12 Restoration method of multiple degraded extremely blurred image

Publications (2)

Publication Number Publication Date
CN107220945A CN107220945A (en) 2017-09-29
CN107220945B true CN107220945B (en) 2020-09-22

Family

ID=59927558

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710234684.5A Expired - Fee Related CN107220945B (en) 2017-04-12 2017-04-12 Restoration method of multiple degraded extremely blurred image

Country Status (1)

Country Link
CN (1) CN107220945B (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111179158B (en) * 2019-12-30 2023-07-18 深圳市商汤科技有限公司 Image processing method, device, electronic equipment and medium
CN113313655B (en) * 2021-06-28 2022-09-23 合肥工业大学 Blind image deblurring method based on saliency mapping and gradient cepstrum technology
CN117058038B (en) * 2023-08-28 2024-04-30 北京航空航天大学 Diffraction blurred image restoration method based on even convolution deep learning
CN117058118B (en) * 2023-08-28 2024-04-05 北京航空航天大学 Method for evaluating image deterioration effect caused by plane shielding glass silk screen

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101968881A (en) * 2010-10-27 2011-02-09 东南大学 Motion blurring and defocusing composite blurring image restoration method
CN102708550A (en) * 2012-05-17 2012-10-03 浙江大学 Blind deblurring algorithm based on natural image statistic property

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR102023518B1 (en) * 2013-02-06 2019-09-23 삼성전자주식회사 Imaging apparatus, ultrasonic imaging apparatus, method for image processing and method for ultrasonic image processing

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101968881A (en) * 2010-10-27 2011-02-09 东南大学 Motion blurring and defocusing composite blurring image restoration method
CN102708550A (en) * 2012-05-17 2012-10-03 浙江大学 Blind deblurring algorithm based on natural image statistic property

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
"运动成像混合模糊的全变分图像复原";石明珠等;《光学 精密工程》;20110831;第19卷(第8期);正文第3页 *

Also Published As

Publication number Publication date
CN107220945A (en) 2017-09-29

Similar Documents

Publication Publication Date Title
Lee et al. Iterative filter adaptive network for single image defocus deblurring
JP6957197B2 (en) Image processing device and image processing method
Li et al. Aod-net: All-in-one dehazing network
Wang et al. Recent progress in image deblurring
Chakrabarti et al. Depth and deblurring from a spectrally-varying depth-of-field
CN110097509B (en) Restoration method of local motion blurred image
Kee et al. Modeling and removing spatially-varying optical blur
CN107220945B (en) Restoration method of multiple degraded extremely blurred image
Quan et al. Gaussian kernel mixture network for single image defocus deblurring
JP2013531268A (en) Measuring distance using coded aperture
CN108629741B (en) Fuzzy kernel estimation method based on L0 and L1 regular terms
Gilles et al. Atmospheric turbulence restoration by diffeomorphic image registration and blind deconvolution
Quan et al. Neumann network with recursive kernels for single image defocus deblurring
El-Henawy et al. A comparative study on image deblurring techniques
Gupta et al. Motion blur removal via coupled autoencoder
US11967096B2 (en) Methods and apparatuses of depth estimation from focus information
CN116029924A (en) Image processing method of infrared system by single-chip diffraction
CN114037636A (en) Multi-frame blind restoration method for correcting image by adaptive optical system
Ranipa et al. A practical approach for depth estimation and image restoration using defocus cue
Fazlali et al. Atmospheric turbulence removal in long-range imaging using a data-driven-based approach
Maik et al. Blind deconvolution using maximum a posteriori (MAP) estimation with directional edge based priori
Liu et al. Guided image deblurring by deep multi-modal image fusion
Li et al. Joint motion deblurring with blurred/noisy image pair
CN101131429A (en) Image restoration method for image quality degraded imaging system with optical aberration and small hole diffraction
HWANG et al. Multi-aperture image processing using deep learning

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
CF01 Termination of patent right due to non-payment of annual fee
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20200922

Termination date: 20210412