CN111127338B - Automatic aberration correction calculation method based on image sharpness evaluation optimization - Google Patents
Automatic aberration correction calculation method based on image sharpness evaluation optimization Download PDFInfo
- Publication number
- CN111127338B CN111127338B CN201911229320.3A CN201911229320A CN111127338B CN 111127338 B CN111127338 B CN 111127338B CN 201911229320 A CN201911229320 A CN 201911229320A CN 111127338 B CN111127338 B CN 111127338B
- Authority
- CN
- China
- Prior art keywords
- aberration
- image
- optimization
- aberration correction
- coefficient
- 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.)
- Active
Links
- 230000004075 alteration Effects 0.000 title claims abstract description 303
- 238000012937 correction Methods 0.000 title claims abstract description 188
- 238000005457 optimization Methods 0.000 title claims abstract description 160
- 238000011156 evaluation Methods 0.000 title claims abstract description 97
- 238000004364 calculation method Methods 0.000 title claims abstract description 35
- 239000002245 particle Substances 0.000 claims abstract description 119
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 68
- 238000000034 method Methods 0.000 claims abstract description 48
- 238000013178 mathematical model Methods 0.000 claims abstract description 16
- 101100269850 Caenorhabditis elegans mask-1 gene Proteins 0.000 claims description 34
- 230000006870 function Effects 0.000 claims description 27
- 238000012545 processing Methods 0.000 claims description 13
- 238000001914 filtration Methods 0.000 claims description 5
- 230000008569 process Effects 0.000 claims description 4
- 206010010071 Coma Diseases 0.000 claims description 3
- 230000001133 acceleration Effects 0.000 claims description 3
- 201000009310 astigmatism Diseases 0.000 claims description 3
- 238000010168 coupling process Methods 0.000 claims description 3
- 238000005859 coupling reaction Methods 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- NRNCYVBFPDDJNE-UHFFFAOYSA-N pemoline Chemical compound O1C(N)=NC(=O)C1C1=CC=CC=C1 NRNCYVBFPDDJNE-UHFFFAOYSA-N 0.000 claims description 3
- 238000004088 simulation Methods 0.000 description 24
- 238000010586 diagram Methods 0.000 description 14
- 238000003384 imaging method Methods 0.000 description 10
- 230000003044 adaptive effect Effects 0.000 description 3
- 238000012014 optical coherence tomography Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000012634 optical imaging Methods 0.000 description 2
- 238000011084 recovery Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000012952 Resampling Methods 0.000 description 1
- 239000012472 biological sample Substances 0.000 description 1
- 238000013170 computed tomography imaging Methods 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002595 magnetic resonance imaging Methods 0.000 description 1
- 238000000386 microscopy Methods 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 239000000523 sample Substances 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
Images
Classifications
-
- G06T5/73—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- 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
Abstract
The invention discloses a method for automatically correcting computed aberration based on image sharpness evaluation optimization, which comprises the following steps: firstly, establishing an aberration correction mathematical model based on image definition evaluation optimization; then establishing a phase filter mathematical model based on image clear evaluation optimization; then solving the aberration correction coefficient value by using the aberration calculation BPSO algorithm: the BPSO algorithm comprises a back propagation optimization algorithm and a particle swarm random optimization algorithm, and the aberration correction coefficient is utilized to solve a clear image after aberration correction. The automatic aberration calculation correction method based on the image definition evaluation optimization has the advantages that the evaluation index can be flexibly selected, the automatic aberration calculation correction is rapidly realized, and the correction result accuracy is high.
Description
Technical Field
The invention relates to the field of optical imaging, in particular to a method for calculating automatic aberration correction based on image definition evaluation optimization.
Background
With the technological progress, biomedical imaging techniques, such as ultrasonic imaging, X-ray imaging, CT, and magnetic resonance imaging, play an important role in the research of bioscience, the diagnosis of medical diseases, and the like. Optical Coherence Tomography (OCT) is a new biomedical imaging technology in the last 90 years, which has the characteristics of noninvasive, high resolution and fast imaging on biological samples, and has been widely favored.
However, when the imaging system has aberrations, the impulse response of the system is blurred: the peak value decreases and the main lobe width increases, thereby generating a blurred image with reduced sharpness and contrast. In order to solve this problem, static correction of aberrations can be achieved by appropriate optical design or a complicated lens group, but this method cannot flexibly correct aberrations at all imaging depths, and cannot satisfy adaptability to aberration correction specific to a sample for tomographic optical imaging. While the conventional Hardware-based Adaptive Optics (HAO) provides a means for correcting these aberrations by physically sensing and modifying the wavefront using a wavefront sensor or a spatial light modulator, etc., the HAO method generally has a complex optical structure and requires expensive components; some foreign researchers are dedicated to research different aberration correction mathematical models, wherein the mathematical models mainly include Digital Adaptive Optics (DAO) based on sub-apertures, which perform high-precision aberration correction on wavefronts by partitioning, but the sub-apertures are difficult to splice, the calculation time cost is huge, and images cannot be processed efficiently and quickly; while the mainstream Interference Synthetic Aperture Microscopy (ISAM) corrects the defocus amount by using a method of interpolation resampling in a wave number range, but the ISAM cannot eliminate other aberrations, so that the application of the ISAM is limited to a certain extent; based on a phase correction Model (FM) of an FM (Forward Model, FM), multiplying a defocusing phase correction factor calculated by using parameters of an imaging system to a two-dimensional Fourier equation of a detected signal to correct defocusing aberration, but the specific parameters of the imaging system need to be obtained, so that the method is not suitable for aberration blind correction and cannot eliminate other aberrations; the method can realize automatic correction of image aberration, does not need to acquire specific parameters of an imaging system, has a wide application range, can provide a substitute method for hardware-based Adaptive Optics, provides flexibility of data post-processing, and greatly saves time cost and hardware overhead.
Disclosure of Invention
The invention aims to provide an automatic aberration calculation correction method based on image definition evaluation optimization, which can flexibly select evaluation indexes, realize quick aberration calculation and automatic correction and have high correction result accuracy.
The technical solution for realizing the purpose of the invention is as follows: an automatic correction method for calculating aberration based on image definition evaluation optimization comprises the following steps:
an aberration-to-be-corrected image g (x, y) whose aberration is to be corrected is selected, and an aberration correction process therefor is modeled as a phase filtering process of a fourier domain of the aberration-to-be-corrected image g (x, y), that is
h(x,y)=ifft(G(u,v)e -jφ(u,v) ) (1)
Wherein ifft is inverse Fourier transform, G (u, v) is an aberration image to be corrected in a Fourier domain, h (x, y) is a corrected clear image, and phi (u, v) is a phase filter; j is an imaginary part;
determining the type of image aberration, selecting an aberration correction basis polynomialDefining aberration correction coefficient values, thereby establishing a mathematical model of the phase filter, i.e.
Wherein the number of different aberration type termsk=1,...,K,c k Is a value of a coefficient of the corresponding aberration;
3-1, obtaining an optimal coefficient value of back propagation optimization by using back propagation optimization in the aberration calculation BPSO algorithm, introducing the aberration image G (u, v) to be corrected in the Fourier domain into a back propagation optimization phase filter mask _1, and searching a polynomial of a correction base of the aberration image to be corrected by adopting the back propagation optimization in the aberration calculation BPSO algorithmThe back propagation aberration correction optimum coefficient value c _1 ═ c _1 1 ,c_1 2 ,...,c_1 K ] T And turning to the step 3-2;
step 3-2, taking the clear image h _1(x, y) after the back propagation optimization as an aberration image to be corrected, introducing a particle swarm random optimization phase filter mask _2, and searching a polynomial of an aberration correction base of the aberration image to be corrected by adopting a particle swarm random optimization algorithm in a calculation aberration BPSO algorithmThe particle swarm random optimization aberration correction optimal coefficient value c _2 ═ c _2 1 ,c_2 2 ,...,c_2 K ] T Turning to step 4;
Compared with the prior art, the invention has the remarkable advantages that:
(1) compared with other automatic correction methods for calculating aberration, the method can flexibly select clear evaluation indexes of the image, and different image processing requirements are met due to the diversity of the evaluation indexes; starting from a plurality of evaluation indexes, the image aberration correction accuracy can be improved.
(2) The aberration calculation automatic correction method based on the image definition evaluation optimization can realize quick aberration calculation automatic correction, the back propagation optimization algorithm in the BPSO algorithm realizes coarse correction of an aberration image to be corrected, a local optimal solution can be quickly found, the cycle iteration times of the particle swarm random optimization algorithm are reduced, and the time consumption of the algorithm is greatly shortened.
(3) The computed aberration automatic correction method based on image definition evaluation optimization has high correction result accuracy, the particle swarm random optimization algorithm in the BPSO algorithm realizes fine correction of an aberration image to be corrected, a local optimal solution can be skipped, a global optimal solution is found, and the aberration correction result accuracy of the image to be corrected is improved.
Drawings
FIG. 1 is a flowchart of an automatic aberration correction method based on image sharpness evaluation optimization according to the present invention.
Fig. 2 is a conceptual diagram of an algorithm for calculating the aberration BPSO.
Fig. 3 is a flowchart of a back propagation optimization algorithm in the algorithm for calculating aberration BPSO.
Fig. 4 is a flowchart of a random optimization algorithm for calculating the particle group in the aberration BPSO algorithm.
FIG. 5 is a simulation experiment chart of an aberration image to be corrected in a simulation experiment of the calculated aberration automatic correction method based on the clear evaluation optimization.
FIG. 6 is a diagram of a simulation experiment of a standard clear image in a simulation experiment of an automatic aberration correction method based on clear evaluation optimization according to the present invention.
Fig. 7 is a simulation experiment diagram of a standard phase filter in a simulation experiment of the calculated aberration automatic correction method based on the clear evaluation optimization.
Fig. 8 is a simulation experiment diagram of a back propagation optimized sharp image h _1(x, y) in a simulation experiment of the calculated aberration automatic correction method based on the sharp evaluation optimization.
Fig. 9 is a simulation experiment diagram of a backward propagation optimized phase filter mask _1 in a simulation experiment of the calculated aberration automatic correction method based on the clear evaluation optimization according to the present invention.
Fig. 10 is a simulation experiment diagram of the evaluation function S _1 changing with the increase of the number of loop iterations in the back propagation optimization in the simulation experiment of the calculated aberration automatic correction method based on the clear evaluation optimization according to the present invention.
Fig. 11 is a simulation experiment diagram of a clear image h _2(x, y) after random optimization of a particle group in a simulation experiment of the calculated aberration automatic correction method based on clear evaluation optimization according to the present invention.
Fig. 12 is a simulation experiment diagram of a random optimization phase filter mask _2 of a particle group in a simulation experiment of the calculated aberration automatic correction method based on the clear evaluation optimization according to the present invention.
Fig. 13 is a simulation experiment diagram showing that the evaluation function S _2 changes with the increase of the number of loop iterations in the random optimization of the particle group in the simulation experiment of the calculated aberration automatic correction method based on the clear evaluation optimization according to the present invention.
Fig. 14 is a simulation experiment diagram of a clear image h (x, y) after correction by a BPSO algorithm in a simulation experiment of the calculated aberration automatic correction method based on clear evaluation optimization according to the present invention.
FIG. 15 is a simulation experiment diagram of the BPSO algorithm corrected phase filter phi (u, v) in the simulation experiment of the calculated aberration automatic correction method based on the clear evaluation optimization.
FIG. 16 is a graph showing the recovery residual error of the BPSO algorithm corrected phase filter φ (u, v) with respect to the standard phase filter in the simulation experiment of the calculated aberration auto-correction method based on the sharpness evaluation optimization according to the present invention.
Detailed Description
In order to make the aforementioned objects, features and advantages of the present invention more comprehensible, embodiments accompanying figures of the present invention are described in detail below.
With reference to fig. 1, a method for calculating an automatic aberration correction based on image sharpness evaluation optimization includes the following steps:
an aberration-to-be-corrected image g (x, y) for which aberration is to be corrected is selected, and aberration correction processing therefor is modeled as phase filtering processing of the fourier domain of the aberration-to-be-corrected image g (x, y), that is
h(x,y)=ifft(G(u,v)e -jφ(u,v) ) (1)
Wherein ifft is inverse Fourier transform, G (u, v) is an aberration image to be corrected in a Fourier domain, h (x, y) is a corrected clear image, and phi (u, v) is a phase filter; j is an imaginary part;
determining the type of image aberration, selecting an aberration correction basis polynomialDefining aberration correction coefficient values, thereby establishing a mathematical model of the phase filter, i.e.
Wherein, the term number K of different aberration types is 1 k Is a coefficient value of the corresponding aberration;
aberration correcting base polynomialIs a Zernike circular domain orthogonal polynomial; the selection of terms of different aberration types corrects four aberrations, namely defocus, astigmatism, coma and spherical aberration.
And 3, solving the aberration correction optimal coefficient value by combining the conceptual diagram of the aberration calculation BPSO algorithm shown in the figure 2:
step 3-1, combining the flowchart of the back propagation optimization algorithm in the aberration calculation BPSO algorithm shown in FIG. 3, obtaining the back propagation optimization optimal coefficient value by using the back propagation optimization in the aberration calculation BPSO algorithm, introducing the aberration image G (u, v) to be corrected in the Fourier domain into a back propagation optimization phase filter mask _1, searching the aberration correction base polynomial of the aberration image to be corrected by using the back propagation optimization in the aberration calculation BPSO algorithmIs most effective in correcting back-propagating aberrationsThe best coefficient value c _1 ═ c _1 1 ,c_1 2 ,...,c_1 K ] T The method comprises the following steps:
step 3-1-1, defining coefficient forward deviation weight eta + Coefficient inverse offset weight η ═ 0.5 - Coefficient forward shift maximum boundary value Δ of 1.2 max Coefficient reverse deviation minimum boundary value delta of 50 min 0.001, aberration coefficient offset
The evaluation function S _1 is set as the image definition, and the specific expression is as follows:
wherein h is * (x, y) is a conjugate matrix of h (x, y).
Further, the maximum number of loop iterations t _1 max The value is taken for 500 times, and the step 3-1-2 is carried out.
Step 3-1-2, optimizing aberration correction coefficient value c _1 ═ c _1 using back propagation 1 ,c_1 2 ,...,c_1 K ] T Selecting aberration correction coefficient c _1 in the following loop iteration k The specific expression is as follows:
for k=1:length(c_1)
c_1 k
do the circle
end (3)
selecting aberration correction coefficient c _1 k If K meets the above conditions, entering loop iteration, and going to step 3-1-3; otherwise, turning to the step 3-1-10;
step 3-1-3, according to the loop iteration time t _1 and the maximum loop iteration time t _1 max Relation, judging the iteration condition of the loop when
t_1≥t_1 max (4)
The loop iteration is stopped to obtain the final selected aberration correction coefficient c _1 k Backward propagation optimized phase filtermask _1 and clear image h _1(x, y) after back propagation optimization, the step 3-1-2 is carried out, and the aberration correction coefficient c _1 updated in the next loop iteration is selected k+1 (ii) a Otherwise, the step 3-1-4 is carried out, and the aberration correction coefficient c _1 is continuously updated in a loop iteration mode k ;
Step 3-1-4, using the aberration correction coefficient c _1 selected in step 3-1-2 k And aberration coefficient offsetCalculating an evaluation function S _1 to optimize an aberration correction coefficient c _1 ═ c _1 for back propagation 1 ,c_1 2 ,...,c_1 K ] T Partial derivatives ofThe specific expression is as follows:
step 3-1-5, utilizing the forward deviation weight eta + Coefficient reverse bias weight η - Coefficient forward deviation maximum boundary value delta max Coefficient is reversely shifted by a minimum boundary value delta min Amount of deviation of each aberration coefficientAndcomputingUpdate value that changes as the number of loop iterations t _1 increasesThe specific expression is as follows:
turning to the step 3-1-6;
step 3-1 to 6, utilizingAndupdating the selected aberration correction coefficient c _1 k The specific expression is as follows:
records the updated selected aberration correction coefficient c _1 k Turning to the step 3-1-7;
step 3-1-7, utilizing the selected aberration correction coefficient c _1 after updating k And aberration correcting base polynomialEstablishing a back propagation optimization phase filter mask _1, wherein the specific expression is as follows:
recording the updated back propagation optimized phase filter mask _1, and turning to the step 3-1-8;
step 3-1-8, the image g (x, y) of the aberration to be corrected is optimized by using the back propagation optimization phase filter mask _1 obtained by updating in the step 3-1-7, and a clear image h _1(x, y) after back propagation optimization is obtained, wherein the specific expression is as follows:
h_1(x,y)=ifft{fft{g(x,y)}.*exp(-1*j*mask_1)} (9)
wherein fft is aberration diagram to be correctedThe image g (x, y) is subjected to two-dimensional Fourier transform processing, ift is two-dimensional inverse Fourier transform, and the selected aberration correction coefficient c _1 updated along with the increase of the number t of loop iterations is recorded k And the phase filter mask _1 is optimized through back propagation, and the step 3-1-9 is carried out;
step 3-1-9, recording the updated partial derivativeSpecific numerical value and sign of (2), and aberration coefficient offset amountSelected aberration correction coefficient c _1 k And the backward propagation optimization phase filter mask _1 and the backward propagation optimization back clear image h _1(x, y) are subjected to the next loop iteration:
t_1=t_1+1 (10)
and (4) transferring to the step 3-1-3.
Step 3-1-10, ending the back propagation optimization loop iteration to obtain the final back propagation aberration correction coefficient value c _1 ═ c _1 1 ,c_1 2 ,...,c_1 K ] T And the clear image h _1(x, y) after the back propagation optimization and the back propagation optimization phase filter mask _1 are switched to the step 3-2.
Step 3-2, combining the flow chart of the particle group random optimization algorithm in the aberration calculation BPSO algorithm shown in FIG. 4, taking the clear image h _1(x, y) after back propagation optimization as an aberration image to be corrected, introducing a particle group random optimization phase filter mask _2, and searching a polynomial of the aberration correction base of the aberration image to be corrected by adopting the particle group random optimization algorithm in the aberration calculation BPSO algorithmThe particle swarm random optimization aberration correction optimal coefficient value c _2 ═ c _2 1 ,c_2 2 ,...,c_2 K ] T The specific algorithm steps are as follows:
step 3-2-1, taking the clear image h _1(x, y) subjected to back propagation optimization as an aberration image to be corrected, initially defining a K-dimensional coefficient search space, and definingMaximum number of particle swarm loop iterations t _2 max ,t_2 max The particles form a set, wherein the t _2 th particle represents a vector with K dimensions, c _2 (t_2) Represents the value of the t _2 th particle and represents the random optimization aberration correction coefficient value of the particle swarm
c_2 (t_2) =(c_2 1 (t_2) ,c_2 2 (t_2) ,...,c_2 K (t_2) ) I.e. by
c_2 (t_2) =(c_2 1 (t_2) ,c_2 2 (t_2) ,...,c_2 K (t_2) ),t_2=1,2,...,t_2 max (12)
T _2 nd particle c _2 (t_2) The motion value of (2) is defined as a K-dimensional vector called the update step V (t_2) ,
V (t_2) =(V 1 (t_2) ,V 2 (t_2) ,...,V K (t_2) ),t_2=1,2,...,t_2 max (13)
Turning to the step 3-2-2;
3-2-2, calculating the fitness F of each particle in the particle swarm by using the evaluation function S _2 as an evaluation index, and turning to the step 3-2-3;
step 3-2-3, correcting the base polynomial to prevent aberrationCoupling correction occurs between the first particle and the second particle, the t _2 th particle c _2 generates the individual optimum and the global optimum of the particle swarm random optimization aberration correction coefficient according to the fitness (t_2) The optimum coefficient value searched by using the evaluation function S _2 as the evaluation index is called an individual optimum solution E best ,
The optimal coefficient value searched by the whole particle swarm by taking the evaluation function S _2 as an evaluation index is called as a global optimal solution G best ,
The step 3-2-4 is changed;
step 3-2-4, utilizing E best And G best Updating step size V (t_2) And randomly optimizing aberration correction coefficient value c _2 by particle swarm (t_2) =(c_2 1 (t_2) ,c_2 2 (t_2) ,...,c_2 K (t_2) ),
V (t-2) =W*V (t-2) +C 1 *R 1 *(E best -c_2 (t-2) )+C 2 *R 2 *(G best -c_2 (t-2) ) (16)
c_2 (t-2) =c_2 (t-2) +V (t-2) (17)
Wherein, C 1 Is an acceleration constant, controls the individual optimal solution E best For update step size V (t_2) Has an influence of 1.5, C 2 Is a learning rate, embodies a global optimal solution G best For update step size V (t_2) With a value of 1, W is an inertia constant with a value of 1, R 1 、R 2 Is [0,1 ]]The random number in the range is transferred to the step 3-2-5;
step 3-2-5, utilizing the updated step size V (t_2) And randomly optimizing aberration correction coefficient value c _2 by particle swarm (t_2) =(c_2 1 (t_2) ,c_2 2 (t_2) ,...,c_2 K (t_2) ) Calculating the fitness F of the updated particles by taking the evaluation function S _2 as an evaluation index, and turning to the step 3-2-6;
step 3-2-6, randomly optimizing aberration correction coefficient value c _2 by using updated particle fitness F and updated particle swarm (t_2) =(c_2 1 (t_2) ,c_2 2 (t_2) ,...,c_2 K (t_2) ) To update the individual optimal solution E best And global optimal solution G best And turning to the step 3-2-7;
step 3-2-7, judging the iteration time t _2 and the maximum particle swarm loop iteration time t _2 max The relationship is such that,when the temperature is higher than the set temperature
t_2>t_2 max (18)
Switching to the step 3-2-8, otherwise, switching to the step 3-2-4;
step 3-2-8, ending the particle swarm random optimization loop iteration to obtain the particle swarm random optimization aberration correction optimal coefficient value c _2 ═ c _2 1 ,c_2 2 ,...,c_2 K ] T The aberration correction optimum coefficient value c _2 is randomly optimized by the particle swarm [ c _2 ═ c _2 1 ,c_2 2 ,...,c_2 K ] T And aberration correcting base polynomialObtaining a particle swarm random optimization phase filter mask _2,
and (3) optimizing the back-propagation optimized clear image h _1(x, y) in the step 3-1-9 by using the particle swarm random optimization phase filter mask _2 to obtain the particle swarm random optimized clear image h _2(x, y), wherein the specific expression is as follows:
h_2(x,y)=ifft{fft{h_1(x,y)}.*exp(-1*j*mask_2)} (20)
and fft is to perform two-dimensional Fourier transform processing on the clear image h _1(x, y) subjected to the inverse propagation optimization, ifft is to perform two-dimensional inverse Fourier transform, and the step 4 is carried out.
In the step 3-2-2, the evaluation function S _2 is set as an image information entropy, and a specific expression is as follows:
in the step 3-2-2, the fitness F of the particle is calculated by taking the evaluation function S _2 as an evaluation index, and the specific expression is as follows:
F=S_2 (22)
using aberration correction optimum coefficient value c _1 ═ c _1 1 ,c_1 2 ,...,c_1 K ] T And c _2 ═ c \u2 1 ,c_2 2 ,...,c_2 K ] T And solving the clear image after correcting the aberration, which is concretely as follows:
using the back propagation optimum aberration correction coefficient value c _1 ═ c _1 1 ,c_1 2 ,...,c_1 K ] T And randomly optimizing the aberration correction by the particle swarm to obtain the optimal aberration correction coefficient c _2 ═ c _2 1 ,c_2 2 ,...,c_2 K ] T Solving aberration correction coefficient value c:
c=c_1+c_2=[c_1 1 +c_2 1 ,c_1 2 +c_2 2 ,...,c_1 K +c_2 K ] T (23)
and (3) solving the phase filter phi (u, v) by utilizing a back propagation optimization phase filter mask _1 and a particle swarm random optimization phase filter mask _2, wherein the specific expression is as follows:
φ(u,v)=mask_1+mask_2 (24)
solving the corrected clear image h (x, y) by utilizing the back propagation optimized clear image h _1(x, y) and the particle swarm random optimized clear image h _2(x, y), wherein the specific expression is as follows:
h(x,y)=h_2(x,y) (25)
and (5) turning to step 4.
Compared with other automatic correction methods for calculating aberration, the automatic correction method for calculating aberration based on image definition evaluation optimization can flexibly select image definition evaluation indexes, meets different image processing requirements due to the diversity of the evaluation indexes, and can improve the accuracy of image aberration correction starting from a plurality of evaluation indexes; the aberration calculation automatic correction method based on the image definition evaluation optimization can realize quick aberration calculation automatic correction, the back propagation optimization algorithm in the BPSO algorithm realizes coarse correction of an aberration image to be corrected, a local optimal solution can be quickly found, the cycle iteration times of the particle swarm random optimization algorithm are reduced, and the time consumption of the algorithm is greatly shortened; the computed aberration automatic correction method based on image definition evaluation optimization has high correction result accuracy, the particle swarm random optimization algorithm in the BPSO algorithm realizes fine correction of an aberration image to be corrected, a local optimal solution can be skipped, a global optimal solution is found, and the aberration correction result accuracy of the image to be corrected is improved.
Example 1
With reference to fig. 1, an algorithm for calculating aberration automatic correction based on image sharpness evaluation optimization specifically includes the following steps:
the aberration-to-be-corrected image g (x, y) for which aberration is to be corrected is selected, and as shown in fig. 5, aberration correction processing thereon is modeled as phase filtering processing in the fourier domain of the aberration-to-be-corrected image g (x, y), that is, processing for phase filtering in the fourier domain
h(x,y)=ifft(G(u,v)e -jφ(u,v) ) (1)
Wherein ifft is inverse Fourier transform, G (u, v) is an aberration image to be corrected in a Fourier domain, h (x, y) is a corrected clear image, and phi (u, v) is a phase filter; j is an imaginary part;
determining the type of image aberration, selecting an aberration correction basis polynomialDefining aberration correction coefficient values, thereby establishing a mathematical model of the phase filter, i.e.
Wherein, the term number K of different aberration types is 1 k Is a value of a coefficient of the corresponding aberration;
and 3, solving the aberration correction optimal coefficient value by combining the conceptual diagram of the aberration calculation BPSO algorithm shown in the figure 2:
step 3-1, combining the flowchart of the back propagation optimization algorithm in the aberration calculation BPSO algorithm shown in fig. 3, obtaining the back propagation optimization optimal coefficient value by using the back propagation optimization in the aberration calculation BPSO algorithm, introducing the aberration image G (u, v) to be corrected in the Fourier domain into a back propagation optimization phase filter mask _1, and searching the aberration correction base polynomial of the aberration image to be corrected by using the back propagation optimization in the aberration calculation BPSO algorithmIs corrected by the back propagation aberration with the optimum coefficient value c _1 ═ c _1 1 ,c_1 2 ,...,c_1 K ] T And turning to the step 3-2;
step 3-2, combining the flow chart of the particle group random optimization algorithm in the aberration calculation BPSO algorithm shown in FIG. 4, taking the clear image h _1(x, y) after back propagation optimization as an aberration image to be corrected, introducing a particle group random optimization phase filter mask _2, and searching a polynomial of the aberration correction base of the aberration image to be corrected by adopting the particle group random optimization algorithm in the aberration calculation BPSO algorithmThe particle swarm randomly optimizes the optimal aberration correction coefficient value c _2 ═ c _2 1 ,c_2 2 ,...,c_2 K ] T Turning to step 4;
In the step 2, the aberration is corrected by the base polynomialIs a Zernike circular domain orthogonal polynomial; the selection of terms of different aberration types corrects four aberrations, namely defocus, astigmatism, coma and spherical aberration.
Step 3-1, the concrete steps are as follows:
step 3-1-1, defining coefficient forward offset weightη + Coefficient inverse offset weight η ═ 0.5 - Coefficient forward shift maximum boundary value Δ of 1.2 max Coefficient reverse deviation minimum boundary value delta of 50 min 0.001, aberration coefficient offsetTurning to the step 3-1-2;
step 3-1-2, optimizing aberration correction coefficient value c _1 ═ c _1 using back propagation 1 ,c_1 2 ,...,c_1 K ] T Selecting aberration correction coefficient c _1 in the following loop iteration k The specific expression is as follows:
for k=1:length(c_1)
c_1 k
do the circle
end (3)
selecting aberration correction coefficient c _1 k If K meets the above conditions, entering loop iteration, and going to step 3-1-3; otherwise, turning to the step 3-1-10;
step 3-1-3, according to the loop iteration time t _1 and the maximum loop iteration time t _1 max Relation, judging the iteration condition of the loop when
t_1≥t_1 max (4)
The loop iteration is stopped to obtain the final selected aberration correction coefficient c _1 k The backward propagation optimization phase filter mask _1 and the backward propagation optimized clear image h _1(x, y) are switched to the step 3-1-2, and the aberration correction coefficient c _1 updated in the next loop iteration is selected k+1 (ii) a Otherwise, the step 3-1-4 is carried out, and the aberration correction coefficient c _1 is continuously updated in a loop iteration mode k ;
Step 3-1-4, using the aberration correction coefficient c _1 selected in step 3-1-2 k And aberration coefficient offsetCalculating an evaluation function S _1 to optimize an aberration correction coefficient c _1 ═ c _1 for back propagation 1 ,c_1 2 ,...,c_1 K ] T Partial derivative of (2)The specific expression is as follows:
step 3-1-5, utilizing the forward deviation weight eta + Coefficient reverse bias weight η - Coefficient forward deviation maximum boundary value delta max Coefficient reverse deviation minimum boundary value delta min Amount of deviation of each aberration coefficientAndcalculating outUpdate value that changes as the number of loop iterations t _1 increasesThe specific expression is as follows:
turning to the step 3-1-6;
step 3-1-6, utilizingAndupdating the selected aberration correction coefficient c _1 k In particularThe expression is as follows:
records the updated selected aberration correction coefficient c _1 k And turning to the step 3-1-7;
step 3-1-7, utilizing the selected aberration correction coefficient c _1 after updating k And aberration correcting base polynomialEstablishing a back propagation optimization phase filter mask _1, wherein the specific expression is as follows:
recording the updated back propagation optimized phase filter mask _1, and turning to the step 3-1-8;
step 3-1-8, optimizing the aberration image g (x, y) to be corrected by using the back propagation optimized phase filter mask _1 obtained by updating in step 3-1-7 to obtain a back propagation optimized clear image h _1(x, y), wherein the specific expression is as follows:
h_1(x,y)=ifft{fft{g(x,y)}.*exp(-1*j*mask_1)} (9)
wherein fft is two-dimensional Fourier transform processing of the aberration image g (x, y) to be corrected, ifft is two-dimensional inverse Fourier transform, and records the selected aberration correction coefficient c _1 updated along with the increase of the cycle iteration number t k And the phase filter mask _1 is optimized through back propagation, and the step 3-1-9 is carried out;
step 3-1-9, recording the updated partial derivativeSpecific numerical value and sign of (1), aberration coefficient offset amountSelected aberration correction coefficient c _1 k And reversely transmitAnd (3) broadcasting an optimized phase filter mask _1 and a back propagation optimized clear image h _1(x, y), and performing the next loop iteration:
t_1=t_1+1 (10)
turning to the step 3-1-3;
step 3-1-10, ending the back propagation optimization loop iteration to obtain the final back propagation aberration correction coefficient value c _1 ═ c _1 1 ,c_1 2 ,...,c_1 K ] T And the back propagation optimized clear image h _1(x, y) and the back propagation optimized phase filter mask _1 are switched to step 3-2 as shown in fig. 8-10.
In the step 3-1-1, the evaluation function S _1 is set to be the image definition, and the specific expression is as follows:
wherein h is * (x, y) is the conjugate matrix of h (x, y).
In the step 3-1-1, the maximum number of loop iterations t _1 max The value was 500 times.
Step 3-2, taking the clear image h _1(x, y) after the back propagation optimization as an aberration image to be corrected, introducing a particle swarm random optimization phase filter mask _2, adopting a particle swarm random optimization algorithm in a computed aberration BPSO algorithm, and searching for an aberration correction base polynomial of the aberration image to be correctedThe particle swarm randomly optimizes the optimal aberration correction coefficient value c _2 ═ c _2 1 ,c_2 2 ,...,c_2 K ] T The specific algorithm comprises the following steps:
step 3-2-1, taking the clear image h _1(x, y) subjected to back propagation optimization as an aberration image to be corrected, initially defining a K-dimensional coefficient search space, and defining the maximum number of particle swarm loop iterations t _2 max ,t_2 max The particles form a set, wherein the t _2 th particle represents a vector with K dimensions, c _2 (t_2) Represents the value of the t _2 particle and represents the random optimization of the particle swarmAberration correction coefficient value c _2 (t_2) =(c_2 1 (t_2) ,c_2 2 (t_2) ,...,c_2 K (t_2) ) I.e. by
c_2 (t_2) =(c_2 1 (t_2) ,c_2 2 (t_2) ,...,c_2 K (t_2) ),t_2=1,2,...,t_2 max (12)
T _2 nd particle c _2 (t_2) The moving value of (2) is defined as a K-dimensional vector called as an updating step V (t_2) ,
Turning to the step 3-2-2;
3-2-2, calculating the fitness F of each particle in the particle swarm by using the evaluation function S _2 as an evaluation index, and turning to the step 3-2-3;
step 3-2-3, correcting the base polynomial to prevent aberrationCoupling correction occurs between the first particle and the second particle, the t _2 th particle c _2 generates the individual optimum and the global optimum of the particle swarm random optimization aberration correction coefficient according to the fitness (t_2) The optimum coefficient value searched by using the evaluation function S _2 as the evaluation index is called an individual optimum solution E best ,
And the optimal coefficient value searched by the whole particle swarm by taking the evaluation function S _2 as an evaluation index is called as a global optimal solution G best ,
The step 3-2-4 is changed;
step 3-2-4, utilizing E best And G best Update the step size V (t_2) And randomly optimizing aberration correction coefficient value c _2 by particle swarm (t_2) =(c_2 1 (t_2) ,c_2 2 (t_2) ,...,c_2 K (t_2) ),
V (t-2) =W*V (t-2) +C 1 *R 1 *(E best -c_2 (t-2) )+C 2 *R 2 *(G best -c_2 (t-2) ) (16)
c_2 (t-2) =c_2 (t-2) +V (t-2) (17)
Wherein, C 1 Is an acceleration constant, controls the individual optimal solution E best For update step size V (t_2) Has an influence of 1.5, C 2 Is a learning rate and embodies a global optimal solution G best For update step size V (t_2) With a value of 1, W is an inertia constant value of 1, R is a value of 1 1 、R 2 Is [0,1 ]]The random number in the range is transferred to the step 3-2-5;
step 3-2-5, utilizing the updated step length V (t_2) And randomly optimizing aberration correction coefficient value c _2 by particle swarm (t_2) =(c_2 1 (t_2) ,c_2 2 (t_2) ,...,c_2 K (t_2) ) Calculating the fitness F of the updated particles by taking the evaluation function S _2 as an evaluation index, and turning to the step 3-2-6;
step 3-2-6, randomly optimizing aberration correction coefficient value c _2 by using updated particle fitness F and updated particle swarm (t_2) =(c_2 1 (t_2) ,c_2 2 (t_2) ,...,c_2 K (t_2) ) To update the individual optimal solution E best And global optimal solution G best And turning to the step 3-2-7;
step 3-2-7, judging the iteration times t _2 and the maximum particle swarm cyclic iteration times t _2 max Relationship when
t_2>t_2 max (18)
Switching to the step 3-2-8, otherwise, switching to the step 3-2-4;
step 3-2-8, ending the random optimization loop iteration of the particle swarm to obtain the particle swarmRandomly optimizing aberration correction optimum coefficient value c _2 ═ c _2 1 ,c_2 2 ,...,c_2 K ] T As shown in fig. 13, the evaluation function S _2 is changed such that the optimum aberration correction coefficient value c _2 is [ c _2 ] by the particle swarm optimization 1 ,c_2 2 ,...,c_2 K ] T And aberration correcting base polynomialObtaining a particle swarm random optimization phase filter mask _2 as shown in FIG. 12
Optimizing the post-optimization clear image h _1(x, y) in the step 3-1-9 by using the particle swarm random optimization phase filter mask _2 to obtain the post-particle swarm random optimization clear image h _2(x, y), as shown in fig. 11, the specific expression is as follows:
h_2(x,y)=ifft{fft{h_1(x,y)}.*exp(-1*j*mask_2)} (20)
and fft is to perform two-dimensional Fourier transform processing on the clear image h _1(x, y) subjected to the inverse propagation optimization, ifft is to perform two-dimensional inverse Fourier transform, and the step 4 is carried out.
In the step 3-2-2, the evaluation function S _2 is set as an image information entropy, and a specific expression is as follows:
in the step 3-2-2, the fitness F of the particle is calculated by taking the evaluation function S _2 as an evaluation index, and the specific expression is as follows:
F=S_2 (22)
using aberration correction optimum coefficient value c _1 ═ c _1 1 ,c_1 2 ,...,c_1 K ] T And c _2 ═ c _2 1 ,c_2 2 ,...,c_2 K ] T And solving the clear image after the image difference is corrected, which comprises the following steps:
optimum aberration correction coefficient using back propagationThe value c _1 ═ c _1 1 ,c_1 2 ,...,c_1 K ] T And randomly optimizing the aberration correction by the particle swarm to obtain the optimal aberration correction coefficient c _2 ═ c _2 1 ,c_2 2 ,...,c_2 K ] T Solving aberration correction coefficient value c:
c=c_1+c_2=[c_1 1 +c_2 1 ,c_1 2 +c_2 2 ,...,c_1 K +c_2 K ] T (23)
and (3) solving the phase filter phi (u, v) by utilizing a back propagation optimization phase filter mask _1 and a particle swarm random optimization phase filter mask _2, wherein the specific expression is as follows:
φ(u,v)=mask_1+mask_2 (24)
solving the corrected clear image h (x, y) by using the back propagation optimized clear image h _1(x, y) and the particle swarm random optimized clear image h _2(x, y), as shown in fig. 14, the specific expression is as follows:
h(x,y)=h_2(x,y) (25)。
in this embodiment 1, through a series of measures, automatic aberration correction is performed on an aberration image to be corrected, which has an aberration, to obtain a corrected sharp image, and to verify a correction result, as shown in fig. 7 and 14, the corrected sharp image is a standard sharp image and a BPSO algorithm-corrected sharp image, as shown in fig. 6 and 15, the corrected sharp image is a standard phase filter and a BPSO algorithm-corrected phase filter, as shown in fig. 16, a BPSO algorithm-corrected phase filter restored residual image is shown, and an RMS (root mean square) value of the restored residual image is 0.029, and an experimental result shows that an evaluation index can be flexibly selected for the calculated aberration automatic correction algorithm based on image sharpness evaluation optimization: the back propagation optimization and the particle swarm random optimization in the BPSO algorithm respectively select the image definition and the image information entropy as evaluation indexes to optimize the image, so that the accuracy of image aberration correction is improved from the viewpoint of a simulation result; the automatic aberration calculation correction method based on image sharpness evaluation optimization can realize quick automatic aberration calculation correction: the back propagation optimization algorithm in the BPSO algorithm realizes the rough correction of the aberration image to be corrected, reduces the cycle iteration times of the particle swarm random optimization algorithm and greatly shortens the time consumption of the algorithm; the automatic aberration correction calculation method based on image definition evaluation optimization has high correction result accuracy: the particle swarm random optimization algorithm in the BPSO algorithm realizes the fine correction of the aberration image to be corrected, the accuracy of the aberration correction result of the image to be corrected is improved, and the recovery residual RMS value is only 0.029.
Claims (7)
1. A method for automatically correcting calculated aberration based on image sharpness evaluation optimization is characterized by comprising the following steps:
step 1, establishing an aberration correction mathematical model based on image definition evaluation optimization:
an aberration-to-be-corrected image g (x, y) whose aberration is to be corrected is selected, and an aberration correction process therefor is modeled as a phase filtering process of a fourier domain of the aberration-to-be-corrected image g (x, y), that is
h(x,y)=ifft(G(u,v)e -jφ(u,v) ) (1)
Wherein ifft is inverse Fourier transform, G (u, v) is an aberration image to be corrected in a Fourier domain, h (x, y) is a corrected clear image, and phi (u, v) is a phase filter; j is an imaginary part;
step 2, establishing a phase filter mathematical model based on image definition evaluation optimization:
determining the type of image aberration, selecting an aberration correction basis polynomialDefining aberration correction coefficient values, thereby establishing a mathematical model of the phase filter, i.e.
Wherein, the term number K of different aberration types is 1 k Is a value of a coefficient of the corresponding aberration;
step 3, solving the aberration correction optimal coefficient value by using a calculation aberration BPSO algorithm:
step 3-1, obtaining the optimal coefficient value of back propagation optimization by using back propagation optimization in the aberration calculation BPSO algorithm, and correcting the aberration map to be corrected in the Fourier domainIntroducing a back propagation optimization phase filter mask _1 into the image G (u, v), and searching a aberration correction base polynomial of the aberration image to be corrected by adopting back propagation optimization in the aberration calculation BPSO algorithmThe back propagation aberration correction optimum coefficient value c _1 ═ c _1 1 ,c_1 2 ,...,c_1 K ] T The method comprises the following steps:
step 3-1-1, defining coefficient forward deviation weight eta + 0.5, coefficient inverse offset weight η - 1.2, coefficient forward offset maximum boundary value Δ max Coefficient reverse shift minimum boundary value Δ of 50 min 0.001, aberration coefficient offset
The evaluation function S _1 is set as the image definition, and the specific expression is as follows:
wherein h is * (x, y) is a conjugate matrix of h (x, y);
turning to the step 3-1-2;
step 3-1-2 of optimizing the aberration correction coefficient value c _1 ═ c _1 using back propagation 1 ,c_1 2 ,...,c_1 K ] T Selecting aberration correction coefficient c _1 in the following loop iteration k The specific expression is as follows:
k=1:length(c_1) (3)
selecting aberration correction coefficient c _1 k If K meets the above conditions, entering loop iteration, and going to step 3-1-3; otherwise, turning to the step 3-1-10;
step 3-1-3, according to the loop iteration time t _1 and the maximum loop iteration time t _1 max Relation, judging the iteration condition of the loop when
t_1≥t_1 max (4)
The loop iteration is stopped to obtain the final selected aberration correction coefficient c _1 k The backward propagation optimization phase filter mask _1 and the backward propagation optimized clear image h _1(x, y) are switched to the step 3-1-2, and the aberration correction coefficient c _1 updated in the next loop iteration is selected k+1 (ii) a Otherwise, the step 3-1-4 is carried out, and the aberration correction coefficient c _1 is continuously updated in a loop iteration mode k ;
Step 3-1-4, using the aberration correction coefficient c _1 selected in step 3-1-2 k And aberration coefficient offsetCalculating an evaluation function S _1 to optimize an aberration correction coefficient c _1 ═ c _1 with respect to back propagation 1 ,c_1 2 ,...,c_1 K ] T Partial derivatives ofThe specific expression is as follows:
step 3-1-5, utilizing the forward deviation weight eta + Coefficient reverse bias weight η - Coefficient forward deviation maximum boundary value delta max Coefficient is reversely shifted by a minimum boundary value delta min Amount of deviation of each aberration coefficientAndcomputingUpdate value that changes as the number of loop iterations t _1 increasesThe specific expression is as follows:
turning to the step 3-1-6;
step 3-1 to 6, utilizingAndupdating the selected aberration correction coefficient c _1 k The specific expression is as follows:
records the updated selected aberration correction coefficient c _1 k And turning to the step 3-1-7;
step 3-1-7, utilizing the selected aberration correction coefficient c _1 after updating k And aberration correcting base polynomialEstablishing a back propagation optimization phase filter mask _1, wherein the specific expression is as follows:
recording the updated back propagation optimized phase filter mask _1, and turning to the step 3-1-8;
step 3-1-8, the image g (x, y) of the aberration to be corrected is optimized by using the back propagation optimization phase filter mask _1 obtained by updating in the step 3-1-7, and a clear image h _1(x, y) after back propagation optimization is obtained, wherein the specific expression is as follows:
h_1(x,y)=ifft{fft{g(x,y)}.*exp(-1*j*mask_1)} (9)
wherein fft is two-dimensional Fourier transform processing of the aberration image g (x, y) to be corrected, ifft is two-dimensional inverse Fourier transform, and the selected aberration correction coefficient c _1 updated along with the increase of the number of loop iterations t is recorded k And a backward propagation optimization phase filter mask _1, and then the step 3-1-9 is carried out;
step 3-1-9, recording the updated partial derivativeSpecific numerical value and sign of (1), aberration coefficient offset amountSelected aberration correction coefficient c _1 k And the backward propagation optimization phase filter mask _1 and the backward propagation optimization back clear image h _1(x, y) are subjected to the next loop iteration:
t_1=t_1+1 (10)
turning to the step 3-1-3;
step 3-1-10, ending the back propagation optimization loop iteration to obtain the final back propagation aberration correction coefficient value c _1 ═ c _1 1 ,c_1 2 ,...,c_1 K ] T The clear image h _1(x, y) after the back propagation optimization and the back propagation optimization phase filter mask _1 are switched to the step 3-2;
step 3-2, taking the clear image h _1(x, y) after the back propagation optimization as an aberration image to be corrected, introducing a particle swarm random optimization phase filter mask _2, and searching a polynomial of an aberration correction base of the aberration image to be corrected by adopting a particle swarm random optimization algorithm in a calculation aberration BPSO algorithmThe particle swarm random optimization aberration correction optimal coefficient value c _2 ═ c _2 1 ,c_2 2 ,...,c_2 K ] T Turning to step 4;
step 4, correcting the optimal coefficient value c _1 ═ c _1 using the aberration 1 ,c_1 2 ,...,c_1 K ] T And c _2 ═ c _2 1 ,c_2 2 ,...,c_2 K ] T And solving the clear image after the image difference is corrected.
2. The method for automatically correcting the calculated aberration based on the image sharpness evaluation optimization according to claim 1, wherein: in the step 2, the aberration is corrected by the base polynomialIs a Zernike circular domain orthogonal polynomial; the selection of terms of different aberration types corrects four aberrations, namely defocus, astigmatism, coma and spherical aberration.
3. The method for automatically correcting the calculated aberration based on the image sharpness evaluation optimization according to claim 1, wherein: in the step 3-1-1, the maximum number of loop iterations t _1 max The value was 500 times.
4. The automatic correction method for aberration calculation based on image definition evaluation optimization according to claim 1, wherein in step 3-2, the back propagation optimized clear image h _1(x, y) is taken as an aberration image to be corrected, a particle swarm random optimization phase filter mask _2 is introduced, and a particle swarm random optimization algorithm in a calculated aberration BPSO algorithm is adopted to find a base polynomial for aberration correction of the aberration image to be correctedThe particle swarm randomly optimizes the optimal aberration correction coefficient value c _2 ═ c _2 1 ,c_2 2 ,...,c_2 K ] T The specific algorithm comprises the following steps:
step 3-2-1, taking the clear image h _1(x, y) after the back propagation optimization as an aberration image to be corrected, initially defining a K-dimensional coefficient search space,defining maximum iteration times t _2 of particle swarm loop max ,t_2 max The particles form a set, wherein the t _2 th particle represents a vector with K dimensions, c _2 (t_2) Represents the value of the t _2 th particle and represents the random optimization aberration correction coefficient value c _2 of the particle swarm (t_2) =(c_2 1 (t_2) ,c_2 2 (t_2) ,...,c_2 K (t_2) ) I.e. by
c_2 (t_2) =(c_2 1 (t_2) ,c_2 2 (t_2) ,...,c_2 K (t_2) ),t_2=1,2,...,t_2 max (12)
T _2 nd particle c _2 (t_2) The motion value of (2) is defined as a K-dimensional vector called the update step V (t_2) ,
Turning to the step 3-2-2;
3-2-2, calculating the fitness F of each particle in the particle swarm by using the evaluation function S _2 as an evaluation index, and turning to the step 3-2-3;
step 3-2-3, correcting the base polynomial to prevent aberrationCoupling correction occurs between the first particle and the second particle, the t _2 th particle c _2 generates the individual optimum and the global optimum of the particle swarm random optimization aberration correction coefficient according to the fitness (t_2) The optimum coefficient value searched by using the evaluation function S _2 as the evaluation index is called an individual optimum solution E best ,
And the optimal coefficient value searched by the whole particle swarm by taking the evaluation function S _2 as an evaluation index is called as a global optimal solution G best ,
The step 3-2-4 is converted;
step 3-2-4, utilizing E best And G best Update the step size V (t_2) And randomly optimizing aberration correction coefficient value c _2 by particle swarm (t_2) =(c_2 1 (t_2) ,c_2 2 (t_2) ,...,c_2 K (t_2) ),
V (t-2) =W*V (t-2) +C 1 *R 1 *(E best -c_2 (t-2) )+C 2 *R 2 *(G best -c_2 (t-2) ) (16)
c_2 (t-2) =c_2 (t-2) +V (t-2) (17)
Wherein, C 1 Is an acceleration constant, controls the individual optimal solution E best For update step size V (t_2) Has an influence of 1.5, C 2 Is a learning rate and embodies a global optimal solution G best For update step size V (t_2) With a value of 1, W is an inertia constant with a value of 1, R 1 、R 2 Is [0,1 ]]The random number in the range is transferred to the step 3-2-5;
step 3-2-5, utilizing the updated step size V (t_2) And randomly optimizing aberration correction coefficient value c _2 by particle swarm (t_2) =(c_2 1 (t_2) ,c_2 2 (t_2) ,...,c_2 K (t_2) ) Calculating the fitness F of the updated particles by taking the evaluation function S _2 as an evaluation index, and turning to the step 3-2-6;
step 3-2-6, randomly optimizing aberration correction coefficient value c _2 by using updated particle fitness F and updated particle swarm (t_2) =(c_2 1 (t_2) ,c_2 2 (t_2) ,...,c_2 K (t_2) ) To update the individual optimal solution E best And global optimal solution G best And turning to the step 3-2-7;
step 3-2-7, judging the iteration time t _2 and the maximum particle swarm loop iteration time t _2 max Relationship when
t_2>t_2 max (18)
Switching to the step 3-2-8, otherwise, switching to the step 3-2-4;
step 3-2-8, ending the particle swarm random optimization loop iteration to obtain the optimal aberration correction coefficient value c _2 ═ c _2 of the particle swarm random optimization 1 ,c_2 2 ,...,c_2 K ] T Randomly optimizing the aberration correction optimum coefficient value c _2 ═ c _2 by using a particle swarm 1 ,c_2 2 ,...,c_2 K ] T And aberration correcting base polynomialObtaining a particle swarm random optimization phase filter mask _2,
and (3) optimizing the back clear image h _1(x, y) in the step 3-1-9 by using a particle swarm random optimization phase filter mask _2 to obtain the back clear image h _2(x, y) of the particle swarm random optimization, wherein the specific expression is as follows:
h_2(x,y)=ifft{fft{h_1(x,y)}.*exp(-1*j*mask_2)} (20)
and fft is to perform two-dimensional Fourier transform processing on the clear image h _1(x, y) subjected to the inverse propagation optimization, ifft is to perform two-dimensional inverse Fourier transform, and the step 4 is carried out.
6. the method for automatically correcting the calculated aberration based on the image sharpness evaluation optimization according to claim 4, wherein: in the step 3-2-2, the fitness F of the particle is calculated by taking the evaluation function S _2 as an evaluation index, and the specific expression is as follows:
F=S_2 (22)。
7. the method according to claim 1, wherein the optimal aberration correction coefficient value c _1 ═ c _1 is used to correct aberration 1 ,c_1 2 ,...,c_1 K ] T And c _2 ═ c _2 1 ,c_2 2 ,...,c_2 K ] T And solving the clear image after the image difference is corrected, which comprises the following steps:
using the back propagation optimum aberration correction coefficient value c _1 ═ c _1 1 ,c_1 2 ,...,c_1 K ] T And randomly optimizing the aberration correction by the particle swarm to obtain the optimal aberration correction coefficient c _2 ═ c _2 1 ,c_2 2 ,...,c_2 K ] T Solving aberration correction coefficient value c:
c=c_1+c_2=[c_1 1 +c_2 1 ,c_1 2 +c_2 2 ,...,c_1 K +c_2 K ] T (23)
and (3) solving the phase filter phi (u, v) by utilizing a back propagation optimization phase filter mask _1 and a particle swarm random optimization phase filter mask _2, wherein the specific expression is as follows:
φ(u,v)=mask_1+mask_2 (24)
and (3) solving the corrected clear image h (x, y) by utilizing the back propagation optimized clear image h _1(x, y) and the particle swarm randomly optimized clear image h _2(x, y), wherein the specific expression is as follows:
h(x,y)=h_2(x,y) (25)。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911229320.3A CN111127338B (en) | 2019-12-04 | 2019-12-04 | Automatic aberration correction calculation method based on image sharpness evaluation optimization |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911229320.3A CN111127338B (en) | 2019-12-04 | 2019-12-04 | Automatic aberration correction calculation method based on image sharpness evaluation optimization |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111127338A CN111127338A (en) | 2020-05-08 |
CN111127338B true CN111127338B (en) | 2022-09-06 |
Family
ID=70497350
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911229320.3A Active CN111127338B (en) | 2019-12-04 | 2019-12-04 | Automatic aberration correction calculation method based on image sharpness evaluation optimization |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111127338B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112381731B (en) * | 2020-11-12 | 2021-08-10 | 四川大学 | Single-frame stripe image phase analysis method and system based on image denoising |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2012102887A2 (en) * | 2011-01-24 | 2012-08-02 | The Board Of Trustees Of The University Of Illinois | Computational adaptive optics for interferometric synthetic aperture microscopy and other interferometric imaging |
CN109919880A (en) * | 2019-03-18 | 2019-06-21 | 郑州轻工业学院 | A kind of infrared image enhancing method based on particle group optimizing |
-
2019
- 2019-12-04 CN CN201911229320.3A patent/CN111127338B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN111127338A (en) | 2020-05-08 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108550115B (en) | Image super-resolution reconstruction method | |
RU2716843C1 (en) | Digital correction of optical system aberrations | |
US9953402B2 (en) | Imaging system and method for imaging objects with reduced image blur | |
JP5414752B2 (en) | Image processing method, image processing apparatus, imaging apparatus, and image processing program | |
Denis et al. | Fast model of space-variant blurring and its application to deconvolution in astronomy | |
CN103310427B (en) | Image super-resolution and quality enhancement method | |
US20050204329A1 (en) | Methods and systems for designing electromagnetic wave filters and electromagnetic wave filters designed using same | |
CN111062889B (en) | Light intensity correction method for Fourier stacked microscopic imaging technology | |
CN113158487B (en) | Wavefront phase difference detection method based on long-short term memory depth network | |
CN111127338B (en) | Automatic aberration correction calculation method based on image sharpness evaluation optimization | |
US7635832B2 (en) | Hybrid diversity method utilizing adaptive diversity function for recovering unknown aberrations in an optical system | |
CN109407311A (en) | A kind of diffraction optics imaging system wavefront inversion algorithm based on phase difference method | |
CN112629678B (en) | Rapid phase recovery method for general shape diffraction-free iterative computation | |
CN104504657B (en) | Method and device for de-noising magnetic resonance diffusion tensor | |
CN112099229B (en) | High-speed self-adaptive optical closed-loop control method based on far field | |
CN108805916B (en) | Image registration method based on fractional order variation and fractional optical flow model and dual optimization | |
Chua et al. | Enhanced image super-resolution technique using convolutional neural network | |
CN107194874B (en) | Super-resolution imaging system and method based on bias image stabilization | |
Wilby et al. | A" Fast and Furious'" solution to the low-wind effect for SPHERE at the VLT | |
CN110322528B (en) | Nuclear magnetic resonance brain image vessel reconstruction method based on 3T and 7T | |
CN113096020A (en) | Calligraphy font creation method for generating confrontation network based on average mode | |
CN108205800B (en) | fast and effective pixelation deconvolution technology | |
JP2013034068A (en) | Image processing method, image processing apparatus and image processing program | |
Leung et al. | Centroid estimation by model-fitting from undersampled wavefront sensing images | |
KRAUZE et al. | Preprocessing methods for quantitative phase image stitching |
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 |