CN111709892A - Rapid blind restoration method and device for space target turbulence degraded image - Google Patents

Rapid blind restoration method and device for space target turbulence degraded image Download PDF

Info

Publication number
CN111709892A
CN111709892A CN202010542938.1A CN202010542938A CN111709892A CN 111709892 A CN111709892 A CN 111709892A CN 202010542938 A CN202010542938 A CN 202010542938A CN 111709892 A CN111709892 A CN 111709892A
Authority
CN
China
Prior art keywords
image
restoration
degraded
turbulence
preliminary
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
Application number
CN202010542938.1A
Other languages
Chinese (zh)
Other versions
CN111709892B (en
Inventor
王静
徐安林
李若娴
刘忠领
高昆
邓蓉
万昊
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.)
Beijing Institute of Technology BIT
Beijing Institute of Environmental Features
63921 Troops of PLA
Original Assignee
Beijing Institute of Technology BIT
Beijing Institute of Environmental Features
63921 Troops of PLA
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 Beijing Institute of Technology BIT, Beijing Institute of Environmental Features, 63921 Troops of PLA filed Critical Beijing Institute of Technology BIT
Priority to CN202010542938.1A priority Critical patent/CN111709892B/en
Publication of CN111709892A publication Critical patent/CN111709892A/en
Application granted granted Critical
Publication of CN111709892B publication Critical patent/CN111709892B/en
Active 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
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/13—Edge detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/70—Denoising; Smoothing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/50—Depth or shape recovery
    • 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
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Landscapes

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

Abstract

The invention relates to a method and a device for quickly and blindly restoring a space target turbulence degraded image, wherein the method comprises the following steps: acquiring a space target turbulence degradation image; solving a parameter value of an atmospheric turbulence optical transfer function based on the frequency spectrum characteristics of the degraded image; obtaining a preliminary restored image by using a wiener filtering restoration method according to the atmospheric turbulence optical transfer function and the obtained space target turbulence degraded image; and enhancing the boundary of the initial restored image by utilizing a truncation total variation method to obtain a restored enhanced image. The method fully utilizes the frequency spectrum characteristics of the space target turbulence degraded image, efficiently estimates the fuzzy atmospheric turbulence optical transfer function of the image, performs fast blind restoration by using wiener filtering, further enhances the initial restoration result, does not need a large number of iterative algorithms, and has clearer restored target boundary, thereby solving the problems of poor applicability and low calculation efficiency of the traditional restoration algorithm to the space target turbulence degraded image.

Description

Rapid blind restoration method and device for space target turbulence degraded image
Technical Field
The invention relates to the technical field of image restoration, in particular to a method and a device for quickly and blindly restoring a space target turbulence degraded image.
Background
The current image restoration method mostly aims at a noise-free natural blurred image, has certain requirements on the signal to noise ratio of the image, however, a space target image subjected to turbulence degradation has the characteristics of low signal to noise ratio, low contrast ratio and the like, and the blind restoration method applied to the natural light image is difficult to be directly applied to restoration of the visible light space target turbulence degradation image.
Although algorithms for space target turbulence degradation images exist in the prior art at present, many parameters need to be adjusted before calculation, and a large number of iterative algorithms need to be adopted in calculation, so that the algorithms need to spend a large amount of time, and are too heavy in manual work and calculation load, and are not suitable for quick recovery requirements. The image frequency spectrum can reflect the real characteristics of the image, so the estimation error is small, and the restoration attempt is performed by using the frequency spectrum characteristics in the restoration of the natural light image at present, but the process needs to use a clear image in the estimation of the point spread function, which is difficult to realize in the restoration of the visible light space target turbulence degradation image.
Therefore, in view of the above disadvantages, it is desirable to provide a fast blind restoration method that can be applied to a spatially target turbulence degraded image.
Disclosure of Invention
The invention aims to solve the technical problems that the existing restoration algorithm is not strong in applicability to a space target turbulence degradation image and low in calculation efficiency, and provides a quick blind restoration method capable of being applied to the space target turbulence degradation image aiming at the defects in the prior art.
In order to solve the technical problem, the invention provides a method for fast blind restoration of a space target turbulence degradation image, which comprises the following steps:
step S1, obtaining a space target turbulence degradation image;
step S2, solving the parameter value of the atmospheric turbulence optical transfer function based on the frequency spectrum characteristics of the degraded image;
s3, obtaining a preliminary restored image by using a wiener filtering restoration method according to the atmospheric turbulence optical transfer function and the obtained space target turbulence degraded image;
and step S4, enhancing the boundary of the preliminary restoration image by utilizing a truncation total variation method to obtain a restoration enhanced image.
Preferably, the step S2 includes:
estimating a pre-degradation image deformation function ln | F' (0, v) |, wherein the expression is as follows:
Figure BDA0002539654980000021
wherein ln | G' (0, v) | represents a deformation function of the degraded image, N represents the abscissa of the origin of the positive axis direction of the spectrum fitting straight line of the pre-degraded image corresponding to the low frequency part in the degraded image spectrum, N is greater than or equal to 0, N represents a half of the image width, a represents the slope of the spectrum fitting straight line of the pre-degraded image corresponding to the low frequency part in the degraded image spectrum, b represents the intercept of the spectrum fitting straight line of the pre-degraded image corresponding to the low frequency part in the degraded image spectrum, and v represents the discrete frequency;
calculating optics from estimated values of a degraded image deformation function ln | G '(0, v) | and a pre-degraded image deformation function ln | F' (0, v) |Transfer function variant- α | v!2βThe expression is as follows:
-α|v|2β≈ln|G′(0,v)|-ln|F′(0,v)|；
using minimization of L1Norm method fitting- α | v2βObtaining coefficients of the fitting curve as α parameter values of the optical transfer function of the atmospheric turbulence, and obtaining the optical transfer function of the atmospheric turbulence as
Figure BDA0002539654980000022
Preferably, the step S4 specifically includes:
and smoothing the preliminary restoration image by utilizing a truncation total variation method to obtain a smooth image with reserved edges, and obtaining a restored enhanced image according to the preliminary restoration image and the smooth image.
Preferably, in step S4, the smoothed image is obtained according to the following formula:
Figure BDA0002539654980000031
where f denotes an input preliminary restored image, s denotes a smoothed image, (sx, sy) denotes a gradient of s, and γ denotes a balance parameter.
Preferably, in step S4, obtaining a restored enhanced image according to the preliminary restored image and the smoothed image specifically includes:
subtracting the smooth image according to the preliminary restored image to obtain image edge information;
multiplying the image edge information and then overlapping the image edge information with the smooth image to obtain a restored enhanced image.
The invention also provides a device for fast blind restoration of a space target turbulence degradation image, which comprises:
the image acquisition unit is used for acquiring a space target turbulence degradation image;
the transfer function determining unit is used for solving the parameter value of the atmospheric turbulence optical transfer function based on the frequency spectrum characteristics of the degraded image;
the preliminary restoration unit is used for obtaining a preliminary restoration image by utilizing a wiener filtering restoration method according to the atmospheric turbulence optical transfer function and the acquired space target turbulence degradation image;
and the restoration enhancing unit is used for enhancing the boundary of the preliminary restoration image by utilizing a truncation total variation method to obtain a restoration enhanced image.
Preferably, the transfer function determination unit is configured to perform the following operations:
estimating a pre-degradation image deformation function ln | F' (0, v) |, wherein the expression is as follows:
Figure BDA0002539654980000032
wherein ln | G' (0, v) | represents a deformation function of the degraded image, N represents the abscissa of the origin of the positive axis direction of the spectrum fitting straight line of the pre-degraded image corresponding to the low frequency part in the degraded image spectrum, N is greater than or equal to 0, N represents a half of the image width, a represents the slope of the spectrum fitting straight line of the pre-degraded image corresponding to the low frequency part in the degraded image spectrum, b represents the intercept of the spectrum fitting straight line of the pre-degraded image corresponding to the low frequency part in the degraded image spectrum, and v represents the discrete frequency;
calculating an optical transfer function deformation form- α | v | from an estimated value of a degraded image deformation function ln | G '(0, v) | and a pre-degraded image deformation function ln | F' (0, v)2βThe expression is as follows:
-α|v|2β≈ln|G′(0,v)|-ln|F′(0,v)|；
using minimization of L1Norm method fitting- α | v2βObtaining coefficients of the fitting curve as α parameter values of the optical transfer function of the atmospheric turbulence, and obtaining the optical transfer function of the atmospheric turbulence as
Figure BDA0002539654980000041
Preferably, the resilience enhancing unit is specifically configured to perform the following operations:
and smoothing the preliminary restoration image by utilizing a truncation total variation method to obtain a smooth image with reserved edges, and obtaining a restored enhanced image according to the preliminary restoration image and the smooth image.
Preferably, the restoration enhancing unit obtains the smoothed image according to the following formula:
Figure BDA0002539654980000042
where f denotes an input preliminary restored image, s denotes a smoothed image, and(s) denotes a reference imagex,sy) Is the gradient representation of s and gamma represents the equilibrium parameter.
Preferably, the restoration enhancing unit is configured to, when obtaining a restoration enhanced image according to the preliminary restoration image and the smoothed image, specifically perform the following operations:
subtracting the smooth image according to the preliminary restored image to obtain image edge information;
multiplying the image edge information and then overlapping the image edge information with the smooth image to obtain a restored enhanced image.
The implementation of the method and the device for fast blind restoration of the space target turbulence degradation image has the following beneficial effects:
1. the scheme of the invention considers the characteristics of the space target turbulence degraded image, fully utilizes the frequency spectrum characteristics of the space target turbulence degraded image, and utilizes wiener filtering to obtain a primary recovered image;
2. the scheme of the invention also utilizes a truncation total variation method to carry out smooth denoising on the primary restored image, and the obtained smooth image can retain important edge information, thereby realizing the enhancement of the boundary of the primary restored image and improving the image quality;
3. according to the scheme, the space target turbulence degradation image is subjected to fast blind restoration based on the frequency spectrum characteristics and the truncation total variation method, a large number of iterative algorithms are not needed, and the problems that the applicability of the existing restoration algorithm to the space target turbulence degradation image is not strong and the calculation efficiency is low are solved.
Drawings
FIG. 1 is a flowchart of a method for fast blind restoration of a spatially target turbulence degradation image according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a spatial target turbulence degradation image fast blind restoration apparatus according to a second embodiment of the present invention;
FIG. 3 is an acquired spatially targeted turbulence degraded image;
FIG. 4 is a spectral diagram of the spatially target turbulence degradation image of FIG. 3;
FIG. 5 is a spectral graph of a pre-degraded image reconstructed by combining the degraded image with a simplified model;
FIG. 6 is- α | v!2βA graph of the calculation result transformed with the discrete frequency v;
FIG. 7 is a graph of the results of the extraction of v ∈ [ -S, S ] (S < N) of FIG. 6;
FIG. 8 is a graph of minimizing L for the data of FIG. 71Norm method fitting- α | v2βA graph of results of (1);
FIG. 9 is the preliminary restored image of FIG. 3;
FIG. 10 is the smoothed image of FIG. 9;
fig. 11 is the restored enhanced image of fig. 9.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be obtained by a person skilled in the art without any inventive step based on the embodiments of the present invention, are within the scope of the present invention.
Example one
As shown in fig. 1, a method for fast blind restoration of a spatially target turbulence degradation image according to an embodiment of the present invention includes the following steps:
first in step S1, a spatial target turbulence degradation image is acquired.
Subsequently in step S2, the parameter values of the atmospheric turbulence optical transfer function are solved based on the spectral characteristics of the degraded image.
In some preferred embodiments, the step S2 includes:
estimating a pre-degradation image deformation function ln | F' (0, v) |, wherein the expression is as follows:
Figure BDA0002539654980000061
wherein ln | G' (0, v) | represents a deformation function of the degraded image, N represents the abscissa of the origin of the positive axis direction of the spectrum fitting straight line of the pre-degraded image corresponding to the low frequency part in the degraded image spectrum, N is greater than or equal to 0, N represents a half of the image width, a represents the slope of the spectrum fitting straight line of the pre-degraded image corresponding to the low frequency part in the degraded image spectrum, b represents the intercept of the spectrum fitting straight line of the pre-degraded image corresponding to the low frequency part in the degraded image spectrum, and v represents the discrete frequency;
calculating an optical transfer function deformation form- α | v | according to the estimated values of the degraded image deformation function ln | G '(0, v) | and the pre-degraded image deformation function ln | F' (0, v) |2βThe expression is as follows:
-α|v|2β≈ln|G′(0,v)|-ln|F′(0,v)|；
using minimization of L1Norm method fitting- α | v2βObtaining coefficients of the fitting curve as α parameter values of the optical transfer function of the atmospheric turbulence, and obtaining the optical transfer function of the atmospheric turbulence as
Figure BDA0002539654980000062
The above steps are described in principle as follows:
after obtaining the space target turbulence degradation image, determining a degradation process function, where the degradation process function in this embodiment is expressed as:
g(x,y)＝h(x,y)*f(x,y)+n(x,y)
in the above formula, f (x, y) represents a pre-degradation image function, g (x, y) represents a degradation image function, h (x, y) represents a point spread function, n (x, y) represents a spatial expression of additive noise, and symbol x represents convolution.
Performing Fast Fourier Transform (FFT) on the degradation process function to obtain a frequency domain transform function, wherein the frequency domain transform function is expressed by the following formula:
G(u,v)＝H(u,v)F(u,v)+N(u,v)
here, G (u, v) ═ FFT (G (x, y)), H (u, v) ═ FFT (H (x, y)), F (u, v) ═ FFT (F (x, y)), and N (u, v) ═ FFT (N (x, y)).
Then, the frequency spectrum of the frequency domain transformation function is normalized to obtain the frequency spectrum normalization function, which is expressed as the following formula:
|G'(u,v)|＝|H(u,v)||F'(u,v)|+|N'(u,v)|
where, | G ' (u, v) | G (u, v) |/max (| G (u, v) |), | F ' (u, v) | | F (u, v) |/max (| G (u, v) |), | N ' (u, v) | N (u, v) |/max (| G (u, v) |).
Then, according to the frequency spectrum normalization function and the frequency spectrum of the degraded image, calculating a parameter value of the atmospheric turbulence optical transfer function, and determining the atmospheric turbulence optical transfer function, wherein the specific process comprises the following steps:
selecting an atmospheric turbulence optical transfer function obeying Gaussian distribution
Figure BDA0002539654980000071
Wherein α and β are function parameters, β in the invention selects an empirical value of β -5/6.
In the invention, the atmospheric turbulence optical transfer function is substituted into a frequency spectrum normalization function, natural logarithms are taken at two sides of an equation, and the equation can be approximately expressed as ln | G' (u, v) | approximately-closed- α (u, v) | approximately-closed2+v2)βAnd + ln | F' (u, v) |, considering that the point spread function has radial symmetry, selecting a straight line u passing through the origin of the (u, v) plane as 0, and rewriting an atmospheric turbulence space turbulence optical transfer function expression to obtain an optical transfer function deformation formula:
-α|v|2β≈ln|G′(0,v)|-ln|F′(0,v)|，
wherein- α | v2βRepresenting the optical transfer function deformation form, ln | G '(0, v) | represents the degraded image deformation function, and ln | F' (0, v) | represents the pre-degraded image deformation function, which is a known term.
The normalized Fourier frequency spectrum of a large number of space target turbulence degradation images is obtained from an image subjected to natural logarithm operation, so that for one space target turbulence degradation image, the frequency spectrum before degradation is higher than the frequency spectrum after degradation, the high-frequency part before and after degradation is slightly changed, the middle-low frequency part is obviously pressed down, and the frequency spectrum of the image shot in a large visual field and a long distance is nearly oblique at the middle-low frequency part. The invention adopts an empirical formula to estimate the image deformation function ln | F' (0, v) before degradation. And taking the high-frequency part data of the degraded image as the high-frequency part data of the image before degradation, and simplifying the evaluation of the middle-low frequency part data of the image before degradation by an approximate isosceles triangle model. ln | F' (0, v) | is expressed by the following equation:
Figure BDA0002539654980000081
wherein ln | G' (0, v) | represents a degraded image deformation function, specifically, a normalized frequency spectrum of the degraded image is taken as a natural logarithm. The abscissa of the starting point in the positive axis direction of the image spectrum fitting straight line before degradation corresponding to the middle and low frequency parts of the degraded image spectrum is represented as n, wherein n is more than or equal to 0, and the ordinate of the starting point is represented as:
Figure BDA0002539654980000082
because the frequency spectrum on the frequency domain coordinate axis is symmetrical relative to the origin, N represents half of the image width, a represents the slope of the pre-degradation image frequency spectrum fitting straight line corresponding to the low frequency part in the degradation image frequency spectrum, generally, a is-1, b represents the intercept of the pre-degradation image frequency spectrum fitting straight line corresponding to the low frequency part in the degradation image frequency spectrum, the intercept is obtained by the abscissa and the ordinate of the starting point of the straight line, and v represents the discrete frequency.
Substituting the pre-degradation image deformation function l | F' (0, v) | into the above optical transfer function deformation equation to obtain:
Figure BDA0002539654980000083
known as, - α | v2βAlong with the variation of the discrete frequency v, when the curve of-N is more than or equal to v and less than or equal to 0 is monotonously increased, and when the curve of v is more than or equal to 0 and less than or equal to N is monotonously decreased, only the origin is attachedThe recent data estimate point spread function is valid, so setting the threshold S and S < N, i.e., taking values within two peaks and valleys, taking data in the range of-S < v < S from- α | v2βExtracting to obtain- α | v-2βThe estimated point spread function can be ensured to meet certain precision by effectively truncating the data.
Reuse minimization L1Performing curve fitting on the effective truncation data by using a norm method to obtain a value of a function parameter α, wherein the function parameter β is a prior value 5/6, so as to determine the atmospheric turbulence optical transfer function H (u, v)1The mode of curve fitting by the norm method is simpler and more efficient, and the calculation amount is further reduced.
In step S3, a preliminary restored image is obtained by using a wiener filtering restoration method according to the atmospheric turbulence optical transfer function and the acquired spatial target turbulence degraded image.
Specifically, after obtaining the optical transfer function H (u, v) of the atmospheric turbulence, restoring the image by using a wiener filtering restoration method, wherein a specific expression is as follows:
Figure BDA0002539654980000091
wherein the content of the first and second substances,
Figure BDA0002539654980000092
k is a preset parameter and is more than or equal to 0.001 and less than or equal to 0.01.
And then, obtaining a primary restoration image by utilizing inverse Fourier transform, wherein the image is represented as:
Figure BDA0002539654980000093
the method considers the characteristics of the space target turbulence degraded image, fully utilizes the frequency spectrum characteristics of the space target turbulence degraded image, and utilizes the wiener filtering to obtain the preliminary recovery image.
Finally, in step S4, the boundary of the preliminary restored image is enhanced by the truncated total variation method, so as to obtain a restored enhanced image.
In some preferred embodiments, the step S4 specifically includes: and smoothing the preliminary restoration image obtained in the step S3 by using a truncation full-variational method to obtain a smoothed image with a retained edge, and obtaining a restored enhanced image according to the preliminary restoration image and the smoothed image.
Specifically, the input preliminary restoration image is represented as f, the smoothed image is represented as s, and the truncated total variation method can be represented as:
s(x)＝∫(T(sx)+T(sy))dxdy，
wherein
Figure BDA0002539654980000094
(sx,sy) For gradient representation of s, s (x) represents a truncated fully-variant function, the truncated fully-variant algorithm penalizes only gradients with magnitude less than a threshold. And performing edge-preserving image denoising operation on the primary restored image by using a smoothing method of truncation total variation, wherein the truncation total variation only offsets partial gradient and can preserve important edges.
In some preferred embodiments, the smoothed image is obtained according to the following formula:
Figure BDA0002539654980000101
where f denotes an input preliminary restored image, s denotes a smoothed image, and(s) denotes a reference imagex,sy) Is the gradient representation of s and gamma represents the equilibrium parameter. Solving this equation yields a smooth image with preserved edges. By the method, important edge information of the image can be retained while smooth denoising is realized.
In some preferred embodiments, after obtaining the smoothed image, image edge information is obtained by subtracting the preliminary restored image and the smoothed image. And finally, multiplying the image edge information and then overlapping the image edge information with the smooth image to obtain a restored enhanced image.
In the invention, smoothing is carried out by a truncation total variation method, and the edge information of the image is obtained by making a difference with a primary restored image, so that the influence of noise on the target significant edge can be avoided to a great extent, and the obtained edge information can be strengthened. In addition, different from the global image enhancement adopted in the prior art, the enhancement mode only enhances important edge information, and the method realizes the spatial target enhancement by utilizing the method of overlapping multiplied image edge information and a smooth image, and can adjust the enhancement degree of the edge.
According to the method, after the preliminary restored image is obtained by utilizing the wiener filtering, the image is subjected to smooth denoising by utilizing a truncation total variation method, the problem of edge preservation and smoothness can be solved, the obtained smooth image retains important edge information, the boundary of the preliminary restored image is enhanced by utilizing the image edge information, and the quality of the restored image is obviously improved.
Example two
As shown in fig. 2, a second embodiment of the present invention provides a fast blind spatial target degraded turbulence image restoration apparatus, which mainly includes: an image acquisition unit 100, a transfer function determination unit 200, a preliminary restoration unit 300, and a restoration enhancement unit 400.
An image acquisition unit 100 for acquiring a spatially targeted turbulence degraded image.
A transfer function determination unit 200 for solving parameter values of the atmospheric turbulence optical transfer function based on the spectral features of the degraded image.
In some preferred embodiments, the transfer function determining unit 200 is configured to perform the following operations:
estimating a pre-degradation image deformation function ln | F' (0, v) |, wherein the expression is as follows:
Figure BDA0002539654980000111
wherein ln | G' (0, v) | represents a deformation function of the degraded image, N represents the abscissa of the origin of the positive axis direction of the spectrum fitting straight line of the pre-degraded image corresponding to the low frequency part in the degraded image spectrum, N is greater than or equal to 0, N represents a half of the image width, a represents the slope of the spectrum fitting straight line of the pre-degraded image corresponding to the low frequency part in the degraded image spectrum, b represents the intercept of the spectrum fitting straight line of the pre-degraded image corresponding to the low frequency part in the degraded image spectrum, and v represents the discrete frequency;
calculating an optical transfer function deformation form- α | v | from an estimated value of a degraded image deformation function ln | G '(0, v) | and a pre-degraded image deformation function ln | F' (0, v)2βThe expression is as follows:
-α|v|2β≈ln|G′(0,v)|-ln|F′(0,v)|；
using minimization of L1Norm method fitting- α | v2βObtaining coefficients of the fitting curve as α parameter values of the optical transfer function of the atmospheric turbulence, and obtaining the optical transfer function of the atmospheric turbulence as
Figure BDA0002539654980000112
And the preliminary restoration unit 300 is configured to obtain a preliminary restoration image by using a wiener filtering restoration method according to the atmospheric turbulence optical transfer function and the acquired spatial target turbulence degradation image.
And a restoration enhancing unit 400, configured to enhance the boundary of the preliminary restoration image by using a truncated full-variational method, so as to obtain a restored enhanced image.
In some preferred embodiments, the restoration enhancement unit 400 is specifically configured to perform the following operations: and smoothing the preliminary restoration image by utilizing a truncation total variation method to obtain a smooth image, and obtaining a restoration enhanced image according to the preliminary restoration image and the smooth image.
In some preferred embodiments, the restoration enhancement unit 400 is configured to obtain a smoothed image according to the following formula:
Figure BDA0002539654980000121
where f denotes an input preliminary restored image, s denotes a smoothed image, and(s) denotes a reference imagex,sy) Is the gradient representation of s and gamma represents the equilibrium parameter.
In some more preferred embodiments, the restoration enhancement unit 400 is configured to perform the following operations to obtain a restoration enhanced image: subtracting the smooth image according to the preliminary restored image to obtain image edge information; and finally, multiplying the image edge information and then overlapping the image edge information with the smooth image to obtain a restored enhanced image.
The operations performed by the image obtaining unit 100, the transfer function determining unit 200, the preliminary restoration unit 300 and the restoration enhancing unit 400 correspond to the operations of the steps S1, S2, S3 and S4, respectively, and are not described herein again.
It should be understood that the principle of the spatial target turbulence degradation image fast blind restoration device of the present invention is the same as the previous spatial target turbulence degradation image fast blind restoration method, and therefore the specific explanation of the embodiment of the spatial target turbulence degradation image fast blind restoration method is also applicable to the device.
In conclusion, the scheme of the invention fully considers the characteristics of the space target turbulence degraded image, and obtains the preliminary restored image through wiener filtering by utilizing the frequency spectrum characteristics of the space target turbulence degraded image. The invention also utilizes a truncation total variation method to carry out smooth denoising, effectively solves the problem of edge protection and smoothness, obtains a smooth image and retains important edge information, and further utilizes the image edge information to realize the enhancement of the primary restored image, thereby improving the quality of the restored image. The method carries out fast blind restoration on the space target turbulence degradation image based on the frequency spectrum characteristics and the truncation total variation method, does not need a large number of iterative algorithms, and solves the problems of poor applicability and poor calculation efficiency of the existing restoration algorithm to the space target turbulence degradation image.
Herein, according to the inventionReferring to fig. 3, a space target turbulence degradation image g (x, y) is obtained, the size of the image is 2M × 2n, the abscissa in fig. 4 to 8 is discrete frequency, referring to fig. 4, a result graph of natural logarithm of normalized frequency spectrum of the degradation image of fig. 3 is obtained, fig. 5 is a spectrum graph of a clear image before degradation is reconstructed by combining the degradation image and a simplified model in fig. 3, and referring to fig. 6, the result graph is- α | v | M2β graph of calculation results transformed with the discrete frequency v, it can be seen from FIG. 6 that, near the origin, when-N ≦ v ≦ 0, the curve monotonically increases, and when 0 ≦ v ≦ N, the curve monotonically decreases, thereby setting the threshold S, e.g., taking 50 in this example, and then-50 will be used<v<Data in the 50 range is ventilated from- α | v2βExtracting to obtain- α | v-2βWith particular reference to fig. 7. Reuse minimization L1The norm method performs curve fitting on the valid truncated data in fig. 7, and the result of the curve fitting is shown in fig. 8. The preliminary restored image obtained by the method of the present invention is shown in fig. 9, the edge of the target in the image is fuzzy, so that the smooth image obtained by using the truncation total variation method is subjected to smooth denoising, the obtained smooth image is shown in fig. 10, finally, the image edge information is obtained according to the smooth image and the preliminary restored image for enhancement, the obtained restored enhanced image is shown in fig. 11, it can be seen that the contour of the target in the image is clear, and the restoration result is obviously improved.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.

Claims (10)

1. A method for fast blind restoration of a space target turbulence degradation image is characterized by comprising the following steps:
step S1, obtaining a space target turbulence degradation image;
step S2, solving the parameter value of the atmospheric turbulence optical transfer function based on the frequency spectrum characteristics of the degraded image;
s3, obtaining a preliminary restored image by using a wiener filtering restoration method according to the atmospheric turbulence optical transfer function and the obtained space target turbulence degraded image;
and step S4, enhancing the boundary of the preliminary restoration image by utilizing a truncation total variation method to obtain a restoration enhanced image.
2. The method for fast blind restoration of a spatially targeted turbulence degraded image according to claim 1, wherein the step S2 includes:
estimating a pre-degradation image deformation function ln | F' (0, v) |, wherein the expression is as follows:
Figure FDA0002539654970000011
wherein ln | G' (0, v) | represents a deformation function of the degraded image, N represents the abscissa of the origin of the positive axis direction of the spectrum fitting straight line of the pre-degraded image corresponding to the low frequency part in the degraded image spectrum, N is greater than or equal to 0, N represents a half of the image width, a represents the slope of the spectrum fitting straight line of the pre-degraded image corresponding to the low frequency part in the degraded image spectrum, b represents the intercept of the spectrum fitting straight line of the pre-degraded image corresponding to the low frequency part in the degraded image spectrum, and v represents the discrete frequency;
calculating an optical transfer function deformation form- α | v | according to the estimated values of the degraded image deformation function ln | G '(0, v) | and the pre-degraded image deformation function ln | F' (0, v) |2βThe expression is as follows:
-α|v|2β≈ln|G′(0,v)|-ln|F′(0,v)|；
using minimization of L1Norm method fitting- α | v2βObtaining coefficients of the fitting curve as α parameter values of the optical transfer function of the atmospheric turbulence, and obtaining the optical transfer function of the atmospheric turbulence as
Figure FDA0002539654970000012
3. The method for fast blind restoration of a spatially turbulent degraded image according to claim 1, wherein the step S4 specifically comprises:
and smoothing the preliminary restoration image by utilizing a truncation total variation method to obtain a smooth image with reserved edges, and obtaining a restored enhanced image according to the preliminary restoration image and the smooth image.
4. The method for fast blind restoration of a spatially target turbulence degradation image according to claim 3, wherein the step S4 is implemented by obtaining a smooth image according to the following formula:
Figure FDA0002539654970000021
where f denotes an input preliminary restored image, s denotes a smoothed image, and(s) denotes a reference imagex,sy) Is the gradient representation of s and gamma represents the equilibrium parameter.
5. The method for fast blind restoration of a spatially target turbulence degraded image according to claim 3 or 4, wherein the obtaining of the restored enhanced image according to the preliminary restored image and the smoothed image in step S4 specifically includes:
subtracting the smooth image according to the preliminary restored image to obtain image edge information;
multiplying the image edge information and then overlapping the image edge information with the smooth image to obtain a restored enhanced image.
6. A fast blind restoration device for a spatially target turbulence degradation image is characterized by comprising:
the image acquisition unit is used for acquiring a space target turbulence degradation image;
the transfer function determining unit is used for solving the parameter value of the atmospheric turbulence optical transfer function based on the frequency spectrum characteristics of the degraded image;
the preliminary restoration unit is used for obtaining a preliminary restoration image by utilizing a wiener filtering restoration method according to the atmospheric turbulence optical transfer function and the acquired space target turbulence degradation image;
and the restoration enhancing unit is used for enhancing the boundary of the preliminary restoration image by utilizing a truncation total variation method to obtain a restoration enhanced image.
7. The apparatus for fast blind restoration of a spatially target turbulence degradation image according to claim 6, wherein the transfer function determining unit is configured to:
estimating a pre-degradation image deformation function ln | F' (0, v) |, wherein the expression is as follows:
Figure FDA0002539654970000031
wherein ln | G' (0, v) | represents a deformation function of the degraded image, N represents the abscissa of the origin of the positive axis direction of the spectrum fitting straight line of the pre-degraded image corresponding to the low frequency part in the degraded image spectrum, N is greater than or equal to 0, N represents a half of the image width, a represents the slope of the spectrum fitting straight line of the pre-degraded image corresponding to the low frequency part in the degraded image spectrum, b represents the intercept of the spectrum fitting straight line of the pre-degraded image corresponding to the low frequency part in the degraded image spectrum, and v represents the discrete frequency;
calculating an optical transfer function deformation form- α | v | according to the estimated values of the degraded image deformation function ln | G '(0, v) | and the pre-degraded image deformation function ln | F' (0, v) |2βThe expression is as follows:
-α|v|2β≈ln|G′(0,v)|-ln|F′(0,v)|；
using minimization of L1Norm method fitting- α | v2βObtaining coefficients of the fitting curve as α parameter values of the optical transfer function of the atmospheric turbulence, and obtaining the optical transfer function of the atmospheric turbulence as
Figure FDA0002539654970000032
8. The apparatus for fast blind restoration of a spatially turbulent degraded image according to claim 7, wherein the restoration enhancing unit is specifically configured to:
and smoothing the preliminary restoration image by utilizing a truncation total variation method to obtain a smooth image with reserved edges, and obtaining a restored enhanced image according to the preliminary restoration image and the smooth image.
9. The apparatus for fast blind restoration of a spatially target turbulence degradation image according to claim 8, wherein the restoration enhancing unit obtains a smooth image according to the following formula:
Figure FDA0002539654970000033
where f denotes an input preliminary restored image, s denotes a smoothed image, and(s) denotes a reference imagex,sy) Is the gradient representation of s and gamma represents the equilibrium parameter.
10. The apparatus according to claim 8 or 9, wherein the restoration enhancing unit is configured to, when obtaining the restoration enhanced image according to the preliminary restoration image and the smoothed image, specifically perform the following operations:
subtracting the smooth image according to the preliminary restored image to obtain image edge information;
multiplying the image edge information and then overlapping the image edge information with the smooth image to obtain a restored enhanced image.
CN202010542938.1A 2020-06-15 2020-06-15 Rapid blind restoration method and device for space target turbulence degradation image Active CN111709892B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010542938.1A CN111709892B (en) 2020-06-15 2020-06-15 Rapid blind restoration method and device for space target turbulence degradation image

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010542938.1A CN111709892B (en) 2020-06-15 2020-06-15 Rapid blind restoration method and device for space target turbulence degradation image

Publications (2)

Publication Number Publication Date
CN111709892A true CN111709892A (en) 2020-09-25
CN111709892B CN111709892B (en) 2023-06-23

Family

ID=72540595

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010542938.1A Active CN111709892B (en) 2020-06-15 2020-06-15 Rapid blind restoration method and device for space target turbulence degradation image

Country Status (1)

Country Link
CN (1) CN111709892B (en)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110228120A1 (en) * 2010-03-16 2011-09-22 Canon Kabushiki Kaisha Image processing apparatus, image pickup apparatus and image processing program
CN104820969A (en) * 2015-04-03 2015-08-05 西安交通大学 Real-time blind image restoration method
CN111210390A (en) * 2019-10-15 2020-05-29 杭州电子科技大学 Motion blur restoration method based on Golay sequence complementary code word set

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110228120A1 (en) * 2010-03-16 2011-09-22 Canon Kabushiki Kaisha Image processing apparatus, image pickup apparatus and image processing program
CN104820969A (en) * 2015-04-03 2015-08-05 西安交通大学 Real-time blind image restoration method
CN111210390A (en) * 2019-10-15 2020-05-29 杭州电子科技大学 Motion blur restoration method based on Golay sequence complementary code word set

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
ZEYANG DOU等: "Image Smoothing via Truncated Total Variation", pages 27337 - 27344 *

Also Published As

Publication number Publication date
CN111709892B (en) 2023-06-23

Similar Documents

Publication Publication Date Title
US8380000B2 (en) Methods of deblurring image and recording mediums having the same recorded thereon
Chen et al. Spatially adapted total variation model to remove multiplicative noise
CN112419162A (en) Image defogging method and device, electronic equipment and readable storage medium
CN111968062B (en) Dark channel prior specular highlight image enhancement method and device and storage medium
Kumari et al. Single image fog removal using gamma transformation and median filtering
Lee et al. Motion deblurring using edge map with blurred/noisy image pairs
Zhang et al. Underwater image enhancement via multi-scale fusion and adaptive color-gamma correction in low-light conditions
CN111709892A (en) Rapid blind restoration method and device for space target turbulence degraded image
KR101942198B1 (en) Terminal devise, Method and Recording medium for enhancing image based on Retinex model
GUAN et al. A dual-tree complex wavelet transform-based model for low-illumination image enhancement
Li et al. A low-illuminance image enhancement method in YUV color space
Song et al. Enhancing Low-Light Color Image via L 0 Regularization and Reweighted Group Sparsity
Zheng et al. An illumination adaptive underwater image enhancement method
CN113920136A (en) Improved dark channel prior defogging algorithm
Han et al. An improved adaptive correction algorithm for non-uniform illumination panoramic image
Duan et al. Local contrast stretch based tone mapping for high dynamic range images
Hu et al. The improved algorithm for the defect of the Retinex Image Enhancement
Wang et al. An improved infrared image contrast enhancement method
Wei et al. MULTISCALE RETINEX COLOR RESTORATION WITH ADAPTIVE GAMMA CORRECTION FOR FOGGY IMAGE ENHANCEMENT
Chen et al. A structure-preserving image restoration method with high-level ensemble constraints
CN113160073B (en) Remote sensing image haze removal method combining rolling deep learning and Retinex theory
Dong et al. Application and Research of Dark Channel Defogging Algorithm in Video Logging Image Enhancement
Zhu et al. Underwater image color correction and adaptive contrast algorithm improvement based on fusion algorithm
Chen et al. Low-light image enhancement based on Double component processing of Retinex model
Li et al. An image dehazing method based on atmospheric veil

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