CN104574315B - Optical system imaging recovering method based on light intensity transmission matrix - Google Patents
Optical system imaging recovering method based on light intensity transmission matrix Download PDFInfo
- Publication number
- CN104574315B CN104574315B CN201510009310.4A CN201510009310A CN104574315B CN 104574315 B CN104574315 B CN 104574315B CN 201510009310 A CN201510009310 A CN 201510009310A CN 104574315 B CN104574315 B CN 104574315B
- Authority
- CN
- China
- Prior art keywords
- matrix
- dimensional
- light intensity
- dimentional
- vector
- 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
Abstract
The invention relates to an optical system imaging recovering method based on a light intensity transmission matrix. The optical system imaging recovering method includes the steps that S1, each row of elements in a two-dimensional image matrix are sequentially connected end to end and arrayed to be one-dimensional image vectors; S2, according to a PSF matrix of an optical system and the one-dimensional image vectors, the two-dimensional light intensity transmission matrix of the optical system is solved, and inversion mathematical calculation is carried out on the two-dimensional light intensity transmission matrix to obtain a two-dimensional light intensity transmission inverse matrix; S3, the one-dimensional image vectors are multiplied with the two-dimensional light intensity transmission inverse matrix to obtain one-dimensional recovering object vectors, and two-dimensional processing is carried out on the one-dimensional recovering object vectors to obtain a two-dimensional recovering object matrix. Imaging is recovered through matrix operation, the ill-conditioned problem of a deconvolution method and the high frequency missing problem of a frequency domain recovering method in the previous imaging recovering technology are avoided, the error between the recovering object matrix and an original object matrix is very small, and therefore the recovering accuracy is high.
Description
Technical field
The invention belongs to optical system imaging restores and image quality lift technique field, it is related to a kind of using two-dimentional light intensity
Transmission matrix is by optical system as matrix reverts to the new method with picture matrix data amount identical reconstruction matrix.
Background technology
During optical system imaging, due to diffraction and the presence of aberration, each point on object plane can be imaged in image planes
For a disc of confusion, so as to cause image blur, image quality to be degenerated, as inconsistent with thing.By being imaged recovery technique to picture
Matter is lifted, and can improve the definition of picture, so that as being lifted with the similarity degree of thing.But in conventional image restoration
Cheng Zhong, generally using deconvolution restored method and frequency spectrum restored method, both approaches all have the shortcomings that certain.First, exist
The point spread function (PSF) that the light distribution function of picture can be with the light distribution function of thing with optical system in spatial domain makees convolution algorithm
Try to achieve, the operation expression of the convolution is the integral equation of Fredholm types.In the case of known picture with PSF, from this
The computing that reconstruction is solved in Fredholm type integral equations is ill-conditioning problem, it is impossible to obtain accurate analytic solutions, so, warp
The reset error of product restored method is larger.Secondly, in a frequency domain as frequency spectrum the optics of optical system can be multiplied by with thing frequency spectrum
Transmission function (OTF) is tried to achieve, and in the case of known picture with OTF, the principle of frequency domain restored method is by as making Fourier transformation
Picture frequency spectrum is obtained, then picture frequency is composed divided by OTF acquisition reconstruction frequency spectrums, finally making inverse Fourier transform to reconstruction frequency spectrum can obtain
To reconstruction.However, OTF has cut-off frequency, generally, the cut-off frequency of OTF will be less than thing frequency spectrum highest frequency.
In optical system imaging, the spectrum component in thing frequency spectrum higher than OTF cut-off frequencies is truncated, and picture frequency spectrum is only ended comprising OTF
Spectrum component within frequency.So, frequency domain restored method can only restore the thing frequency spectrum within OTF cut-off frequencies, cause to restore
The radio-frequency component disappearance of thing, reset error are larger.
The content of the invention
It is an object of the invention to provide a kind of optical system imaging restored method based on light intensity transmission matrix, the method profit
Imaging is realized with matrix operation to restore, it is to avoid the ill-conditioning problem and frequency domain of the Deconvolution Method in conventional imaging recovery technique
The high frequency disappearance problem of restored method, reconstruction matrix are minimum with the error of original thing matrix, and recovery accuracy is high.
The purpose of the present invention is achieved through the following technical solutions:
A kind of optical system imaging restored method based on light intensity transmission matrix, comprises the steps:
The first step, the one-dimensional of two-dimensional image matrix are processed:
One-dimensional picture vector is arranged as by each row element in two-dimensional image matrix end to end successively.
Second step, two-dimentional light intensity transmission matrix and two-dimentional light intensity transmit the solution of inverse matrix:
For a certain known optical system, optical system is solved according to the PSF matrixes and one-dimensional picture vector of the optical system
The two-dimentional light intensity transmission matrix of system;The mathematical computations inverted by two-dimentional light intensity transmission matrix are just obtained into two-dimentional light intensity transmission
Inverse matrix.
3rd step, the solution of reconstruction matrix and reset error are calculated:
One-dimensional picture vector is multiplied by into two-dimentional light intensity transmission inverse matrix and tries to achieve one-dimensional reconstruction vector, by one-dimensional reconstruction vector
Two-dimentional reconstruction matrix is obtained after two dimensionization.Known two-dimentional original thing matrix is subtracted each other with two-dimentional reconstruction matrix and is obtained
Reset error matrix, and then it is equal with the error of corresponding element in two-dimentional reconstruction matrix to obtain known two-dimentional original thing matrix
Root value RMSE, RMSESize be used for evaluate imaging restored method recovery accuracy.
Optical system imaging restored method based on light intensity transmission matrix proposed by the present invention known two-dimensional image matrix with
During two-dimentional PSF Matrix Solvings two dimension reconstruction matrix, one-dimensional picture vector can be multiplied by two-dimentional light intensity transmission inverse matrix and obtain one-dimensional
The vector two dimension chemical conversion of one-dimensional reconstruction two-dimentional reconstruction matrix just can be tried to achieve accurate restoration result by reconstruction vector.It is this
The two-dimentional reconstruction matrix that method is obtained is minimum with the error of known two-dimentional original thing matrix, and recovery accuracy is high.
Description of the drawings
Fig. 1 is optical system spatial domain imaging schematic diagram;
Fig. 2 is the imaging schematic diagram that image planes are rotated 180 degree around optical axis;
Fig. 3 is the one-dimensional Principle of Process figure that the unknown original thing matrix of two dimension is converted into one-dimensional unknown original thing vector;
Fig. 4 is the one-dimensional Principle of Process figure that two-dimensional image matrix is converted into one-dimensional picture vector;
Fig. 5 is the graph of a relation of vectorial one-dimensional unknown material, one-dimensional picture vector and two-dimentional light intensity transmission matrix;
Fig. 6 is the one-dimensional picture vector after the two-dimensional image matrix and one-dimensional being made up of 5 row, 5 column element;
Fig. 7 be according in Fig. 6 as the two-dimentional unknown original thing matrix being made up of 5 row, 5 column element and one-dimensional that matrix is set up
One-dimensional unknown material vector after change;
Fig. 8 is the PSF matrixes being made up of 3 row, 3 column element;
Element as of the Fig. 9 (a) for the unknown original thing matrix of two dimension in Fig. 71,1(corresponding in one-dimensional unknown original thing vector
First element α1) light distribution data that Jing optical systems are formed in image planes, it is (b) element a1,1To the light as matrix each element
Strong transmission coefficient;
Element as of the Figure 10 (a) for the unknown original thing matrix of two dimension in Fig. 71,2(corresponding in one-dimensional unknown original thing vector
Second element α2) light distribution data that Jing optical systems are formed in image planes, it is (b) element a1,2To as matrix each element
Light intensity transmission coefficient;
Element as of the Figure 11 (a) for the unknown original thing matrix of two dimension in Fig. 72,2(corresponding in one-dimensional unknown original thing vector
The 7th element α7) light distribution data that Jing optical systems are formed in image planes, it is (b) element a2,2To as matrix each element
Light intensity transmission coefficient;
Element as of the Figure 12 (a) for the unknown original thing matrix of two dimension in Fig. 75,5(corresponding in one-dimensional unknown original thing vector
Last element α25) light distribution data that Jing optical systems are formed in image planes, it is (b) element a5,5To as matrix each element
Light intensity transmission coefficient;
Figure 13 is the two-dimentional light intensity transmission matrix tried to achieve with picture vector in Fig. 6 by PSF matrixes in Fig. 8;
Figure 14 (a) is the known original thing of alphabetical F patterns, (b) is 7 row, 7 column matrix of the original thing of alphabetical F patterns
Figure 15 is optical system PSF matrix;
Figure 16 (a) be alphabetical F things Jing optical systems into picture, (b) be picture 7 row, 7 column matrix;
Figure 17 is the two-dimentional light intensity transmission matrix data obtained by the picture Matrix Calculating in the PSF matrixes and Figure 16 (b) in Figure 15
Graphics;
Figure 18 is the two-dimentional light intensity transmission inverse matrix tried to achieve by two-dimentional light intensity transmission matrix in Figure 17;
Figure 19 (a) is the reconstruction after restoring to the picture in Figure 16, is (b) two-dimentional reconstruction matrix;
Figure 20 is the error moments that the reconstruction matrix for deducting Figure 19 (b) by the known original thing matrix of Figure 14 (b) is obtained
Battle array;
Figure 21 is containing original thing known to 400 row, 400 column element;
Figure 22 is the optical system PSF matrix containing 5 row, 5 column element;
Figure 23 is the picture containing 400 row, 400 column element;
Figure 24 be the picture in Figure 23 is restored after the reconstruction that obtains;
Figure 25 is the error matrix data three-dimensional figure containing 400 row, 400 column element.
Specific embodiment
Below in conjunction with the accompanying drawings technical scheme is further described, but is not limited thereto, it is every to this
Inventive technique scheme is modified or equivalent, without deviating from the spirit and scope of technical solution of the present invention, all should cover
In protection scope of the present invention.
The present invention is provided a kind of reverting to the two-dimensional image matrix of optical system using two-dimentional light intensity transmission matrix and is counted with which
According to the new method of amount identical two dimension reconstruction matrix.The solution of two-dimentional reconstruction matrix is in known two-dimensional image matrix and two dimension
Carry out under conditions of PSF matrixes, during the entire process of two-dimentional reconstruction matrix is solved, two-dimentional original thing matrix is unknown bar
Part.In order to illustrate the principle of imaging restored method proposed by the present invention, need to set up one it is complete with two-dimensional image matrix data amount
Identical two-dimensional matrix is used as the unknown original thing matrix of two dimension.The element symbol a of the unknown original thing matrix of two dimensionM, nRepresent, its
In, m, n are the position number of certain element in the unknown original thing matrix of two dimension.
The unknown original thing matrix of two dimension is on all four with the meaning of two-dimentional original thing matrix, and only the former is all
Element is unknown, and all elements of the latter are known, and the dimension of the two is identical with the dimension of two-dimensional image matrix.
The two-dimentional light intensity transmission matrix solved in this restored method and the element a of the unknown original thing matrix of two dimensionM, nIt is unrelated, only with two
Dimension is relevant as matrix and PSF matrixes.Evaluate if desired for the recovery accuracy to this method, then need to provide known two dimension original
Beginning thing matrix is being analyzed to reset error.Comprise the following steps that:
The first step:The one-dimensional of two-dimensional image matrix is processed.
Current image stores into the form of e-file, and gray level image is represented with a two-dimensional matrix, in matrix
The relative position of a point in the positional representation image planes of each element, the numerical value of each element represent the light intensity of the point.
Color document image represents that with a three-dimensional matrice this three-dimensional matrice is equivalent to three two-dimensional matrixs, each two-dimensional matrix point
Not Biao Shi image each point three kinds of colors of red, green, blue.When processing coloured image, can be according to the method difference of gray level image process
The data to three kinds of colors of red, green, blue are processed, therefore the present invention is only illustrated to the restored method of gray level image.
To illustrate the principle of this method, first the imaging process of optical system is described.As shown in figure 1, two-dimentional thing
The light intensity Jing optical system 3 of a certain object point 2 on face 1 is imaged as a disc of confusion 5, the light intensity of disc of confusion 5 in two-dimentional image planes 4
Distribution can be accurately described by the data of the point spread function (PSF) 6 centered on corresponding picture point.In the light that turns light path of not putting english
In system, image planes are to stand upside down relative to object plane, i.e., image planes have rotated 180 degree angle relative to object plane.For the ease of analysis and
Process, image planes can be rotated 180 degree centered on optical axis, as shown in Fig. 2 this processing method does not change the imaging of optical system
Rule, in actual optical system, can turn light path realization by putting english.As shown in Fig. 2 on two-dimentional object plane 1 each point light intensity number
According to can by Fig. 3 in two-dimentional unknown original thing matrix 7 represent that equally, the light intensity data of each point can be by Fig. 4 in two-dimentional image planes 4
Two-dimensional image matrix 9 represent.Certain element a of the unknown original thing matrix of two dimensionM, nTo certain element b of two-dimensional image matrixQ, w's
Light intensity transmission coefficient dM, n, q, wMust be four-dimensional parameter, then by dM, n, q, wThe matrix of composition necessarily four-matrix, and four-matrix
It is not easy to analyze and processes, being also not easy to draw represents.And for the element α in one-dimensional thing vectoriTo one-dimensional picture vector
In element βjLight intensity transmission coefficient pI, jIt is two-dimensional parameter, by pI, jThe light intensity transmission matrix of composition is two-dimensional matrix, and two
Dimension matrix is easy to analyze, is processed, and also allowing for drawing represents.
The solution of reconstruction matrix in this restored method is entered under conditions of known two-dimensional image matrix with two dimension PSF matrixes
Capable, the two-dimentional original thing matrix during the two-dimentional reconstruction matrix of solution is unknown.Need to set up one with two-dimensional image square
The identical two-dimensional matrix of battle array data volume is used as the unknown original thing matrix of two dimension.
As shown in figure 3, the one-dimensional of thing process be exactly by the every a line in two-dimentional unknown 7 each row element of original thing matrix according to
It is secondary end to end to be arranged in order as a row vector.As shown in figure 4, it is exactly by 9 each row of two-dimensional image matrix that the one-dimensional of picture is processed
Every a line in element is end to end successively to be arranged in order as a row vector.Thing vector is pressed again with the element in picture vector
After tandem numbering, one-dimensional unknown original thing vector 8 and one-dimensional picture vector 10 are just defined.
As shown in Figure 3, Figure 4, to the unknown original thing matrix of two dimension, after carrying out one-dimensional as matrix, two dimension is unknown original
Certain element a of thing matrixM, nRespective element α being changed in one-dimensional unknown original thing vectori, certain unit of two-dimensional image matrix
Plain bQ, wRespective element β being changed in one-dimensional picture vectorj, following relation should be met:
Wherein, m, n are the sequence number of certain element in the unknown original thing matrix of two dimension, and i is in one-dimensional unknown original thing vector
The sequence number of certain element, q, w are the sequence number of certain element in two-dimensional image matrix, and j is the sequence number of certain element in one-dimensional picture vector,
M, N are the line number of the unknown original thing matrix of two dimension, columns, Q, W total line number, total columns for two-dimensional image matrix, and i, j, m, n,
Q, w, M, N, Q, W are positive integer.Due to recovering and two-dimensional image matrix function in imaging recovery technique of the present invention
According to etc. big two-dimentional reconstruction matrix, the element of two-dimentional reconstruction matrix and two-dimensional image matrix as many, so, M is equal with Q
, N be equal with W.
Second step:Two-dimentional light intensity transmission matrix and two-dimentional light intensity transmit the solution of inverse matrix.
According to the basic theories of Fourier Optics, optical imaging system belongs to incoherent imaging system, two-dimentional object plane
On the light intensity that sends of each object point be focused in image planes by optical system and form a disc of confusion for covering multiple picture points,
Disc of confusion is centrally located at the picture point (i.e. in Fig. 2 with object point two-dimensional position identical picture point) corresponding with object point, the disc of confusion
Light distribution can be accurately described by the PSF centered on the picture point.It should be noted that with every in linear space-variant optical system
PSF centered on individual picture point is different, and the PSF in linear empty constant optical system centered on each picture point is phase
With.So, need in the imaging recuperation of linear space-variant optical system it is known as matrix in each pixel element be
The PSF matrixes of the heart, and in the imaging recuperation of linear empty constant optical system, it is only necessary to some in known picture matrix
PSF matrixes centered on pixel element.
Due to certain element a of the unknown original thing matrix of two dimensionM, nIt is transferred to certain element b of two-dimensional image matrixQ, wLight
Strong transmission coefficient dM, n, q, wCan be from bM, nCentered on PSF extracting datas, and dM, n, q, wWith the unit in one-dimensional unknown material vector
Plain αiElement β in as vectorjLight intensity transmission coefficient pI, j(element i.e. in two-dimentional light intensity transmission matrix 11) be it is equal,
I.e.:
Wherein, the relation and meaning in parameter i, j, the relation of m, n, q, w, M, N, Q, W satisfaction and meaning and formula (1), (2)
Justice is identical.Therefore, as long as the unknown original thing matrix of the good two dimension of record, as the four-dimensional light intensity transmission coefficient between matrix each element
dM, n, q, w, one-dimensional unknown original thing vector each element just can be obtained to the two-dimentional light intensity transmission coefficient of one-dimensional picture vector each element
pI, j, the coefficient is the respective element in the two-dimentional light intensity transmission matrix 11 shown in Fig. 5.According to the method, obtain successively one-dimensional
Unknown original thing vector each element is just obtained two-dimentional light intensity transmission matrix to the light intensity transmission coefficient of one-dimensional picture vector each element
11.Due to the transmission coefficient p in two-dimentional light intensity transmission matrixI, jThe data being in PSF, so, two-dimentional light intensity transmission matrix with
The element a of the unknown original thing matrix of two dimensionM, nUnrelated, two-dimentional light intensity transmission matrix completely can be according to two-dimensional image matrix and two dimension
PSF Matrix Solvings.
After solving two-dimentional light intensity transmission matrix 11, one-dimensional unknown original thing vector 8 is multiplied by two-dimentional light intensity transmission matrix 11
Just establish equal to the relation of one-dimensional picture vector 10, as shown in Figure 5.This product calculation relation can be write as the form of formula (4)
A P=B (4);
Wherein, A is one-dimensional unknown original thing vector, and meaning is equal to the one-dimensional unknown original thing vector 8 in Fig. 5;P is two
Dimension light intensity transmission matrix, meaning are equal to the two-dimentional light intensity transmission matrix 11 in Fig. 5;B is that one-dimensional picture is vectorial, and meaning is equal to figure
Picture vector 10 in 5.
The mathematical operation inverted by two-dimentional light intensity transmission matrix is just obtained into two-dimentional light intensity transmission inverse matrix.Two-dimentional light
Strong transmission matrix has following relation with two-dimentional light intensity transmission inverse matrix:
P·P-1=E (5);
Wherein, matrix P is two-dimentional light intensity transmission matrix, matrix P-1Inverse matrix is transmitted for two-dimentional light intensity, matrix E is and P ranks
Number identical unit matrix.
For matrix P, its Methods of Finding Inverse Matrix is as follows:
Wherein, determinants of | the P | for matrix P, matrix P*For the adjoint matrix that the algebraic complement of matrix P each element is constituted
Battle array.
3rd step, the solution of two-dimentional reconstruction matrix and reset error are calculated:
The one-dimensional picture vector that the first step is obtained is multiplied by into the two-dimentional light intensity transmission inverse matrix that second step obtains and just obtains one
Dimension reconstruction vector, i.e.,
C=B P-1(7),
B is that one-dimensional picture is vectorial, P-1Inverse matrix is transmitted for two-dimentional light intensity, C is one-dimensional reconstruction vector.
One-dimensional reconstruction vector C two dimensionizations are just obtained into two-dimentional reconstruction matrix, such as shown in formula (8)
Wherein, aK, l' it is element in two-dimentional reconstruction matrix, αc' it is element in one-dimensional reconstruction vector C, k, l are
The sequence number of certain element in two-dimentional reconstruction matrix, c are the sequence number of certain element in one-dimensional reconstruction vector, and K, L are multiple for two dimension
Total line number of original matrix, total columns, K are equal to the W that Q, L in formula (2) are equal in formula (2), and c, k, l, K, L are just
Integer.
In order to evaluate the recovery accuracy of the imaging restored method, needs are imaged using known two-dimentional original thing,
Imaging is obtained two-dimentional reconstruction matrix according to the method described above, two-dimentional original thing matrix is deducted two-dimentional reconstruction matrix can
Two-dimentional reset error matrix E is obtained, such as shown in formula (9), reset error root-mean-square value RMSECan be calculated by formula (10):
Wherein, eK, lFor the element of two-dimentional error matrix,For the element of known two-dimentional original thing matrix, aK, l' be
The element of two-dimentional reconstruction matrix, k, l are element in known two-dimentional original thing matrix, two-dimentional reconstruction matrix and error matrix
Sequence number, K, L are total line number, total columns of known two-dimentional original thing matrix, two-dimentional reconstruction matrix and error matrix, and k,
L, K, L are positive integer, and K is equal to the Q in formula (2), and L is equal to the W in formula (2), and c, k, l, K, L are positive integer.
So far, the optical system imaging restored method based on light intensity transmission matrix of the present invention has all been illustrated
Finish, below with three examples illustrating principle, recuperation, restoration result and the reset error of methods described.
Example one:
In order to absolutely prove the method for solving of two-dimentional light intensity transmission matrix of the present invention, Fig. 6 gives and is arranged by 5 rows 5
The example of the one-dimensional picture vector after the two-dimensional image matrix and one-dimensional of element composition.Fig. 7 is given according to two-dimensional image matrix in Fig. 5
It is unknown original after the two-dimentional unknown original thing matrix containing 5 row, 5 column element big with two-dimensional image matrix etc. set up and one-dimensional
Thing vector, in solution procedure, the unknown original thing matrix of two dimension is not with all elements in one-dimensional unknown original thing vector
Know.Fig. 8 gives the PSF matrix examples being made up of 3 row, 3 column element, and the optical system is linear sky invariant system.Two dimension is not
Know the effect of each element Jing optical systems PSF in original thing matrix, disc of confusion can be formed in image planes, but in actual optical system
In system, due to the restriction of diaphragm and image planes size, exceeding because of the diffusion of PSF as the part at edge cannot be recorded in detector
On, so, cannot record in as matrix beyond the part as matrix in disc of confusion.Fig. 9 (a) is given two in Fig. 6
Tie up the element a of unknown original thing matrix1,1(corresponding to first element α in one-dimensional unknown original thing vector1) Jing optical systems
In the light distribution data that image planes are formed, Fig. 9 (b) gives a1,1To the light intensity transmission coefficient of two-dimensional image matrix each element.Figure 10
A () gives the element a of the unknown original thing matrix of two dimension1,2(corresponding to second element in one-dimensional unknown original thing vector
α2) the light distribution data that formed in image planes of Jing optical systems, Figure 10 (b) gives a1,2To the light intensity of two-dimensional image matrix each element
Transmission coefficient.The like, Figure 11 (a) gives the element a2 of the unknown original thing matrix of two dimension, and 2 (correspond to one-dimensional unknown original
The 7th element α in beginning thing vector7) the light distribution data that formed in image planes of Jing optical systems, Figure 11 (b) gives a2,2Arrive
The light intensity transmission coefficient of two-dimensional image matrix each element.Figure 12 (a) gives the element a of the unknown original thing matrix of two dimension5,5(correspondence
Last element α in one-dimensional unknown original thing vector25) the light distribution data that formed in image planes of Jing optical systems, Figure 12
B () gives a5,5To the light intensity transmission coefficient of two-dimensional image matrix each element.On the whole, the matrix in Fig. 9 (a)~Figure 12 (a)
For the light intensity diffusion matrix that the element of the unknown original thing matrix of two dimension is formed in image planes, Fig. 9 (b)~Figure 12 (b) is two-dimentional unknown
Light intensity diffusion coefficient matrix of the element of original thing matrix to two-dimensional image matrix each element.Two-dimensional image matrix is equal to all two dimensions not
Know the light intensity diffusion matrix sum of form that the element of original thing matrix formed in image planes as shown in Fig. 9 (a)~Figure 12 (a).Can
To find out, light intensity diffusion coefficient matrix shown in Fig. 9 (b)~Figure 12 (b) and the two-dimentional unknown original thing matrix element shown in Fig. 6
Unrelated, the PSF matrixes that light intensity diffusion matrix completely can be as shown in Figure 8 are asked by formula (3) with the two-dimensional image matrix shown in Fig. 6
.The light intensity diffusion coefficient matrix that Fig. 9 (b)~Figure 12 (b) is given is the unknown original thing matrix element a of two dimensionM, nTo two-dimensional image
Matrix element bQ, wFour-dimensional light intensity transmission coefficient dM, n, q, wSet, as the unknown original thing matrix of two dimension has 25 units
Element, the number of light intensity diffusion coefficient matrix is also 25.For example, what the light intensity diffusion coefficient matrix shown in Fig. 9 (b) was represented is
a1,1To b1,1、b1,2、b2,1、b2,2The four-dimensional strong transmission coefficient d of light1,1,1,1、d1,1,1,2、d1,1,2,1、d1,1,2,2, according to formula (1),
(2) understand, a1,1Corresponding to the element α in one-dimensional unknown original thing vector1, and b1,1、b1,2、b2,1、b2,2Correspond respectively to one-dimensional
As the element β in vector1、β2、β6、β7, understood according to formula (3), element α in one-dimensional unknown original thing vector1To one-dimensional picture to
Element β in amount1、β2、β6、β7Between two-dimentional light intensity transmission coefficient p1,1、p1,2、p1,6、p1,7Respectively equal to d1,1,1,1、d1,1,1,2、
d1,1,2,1、d1,1,2,2, and α1To except β1、β2、β6、β7Other one-dimensional two-dimentional light intensity diffusion coefficients as vector element in addition are
0, so far, the first row element p of two-dimentional light intensity transmission coefficient1, j(j=1~25) all solve out.For α2~α25, press
The all elements p of the 2nd~25 row in two-dimentional light intensity transmission matrix can be obtained according to same method2, j~p25, j(j=1~25).
By all light intensity transmission coefficient pI, jWrite as the form of matrix by subscript position number, just can be tried to achieve the two-dimentional light shown in Figure 13
Strong transmission matrix.As can be seen that the two-dimentional light intensity transmission matrix shown in Figure 13 has 25 rows, 25 row, its all elements is Fig. 8
The element a of the two-dimentional unknown original thing matrix in the data in shown PSF matrixes, with Fig. 7M, nIt is unrelated.
Example two:
In order to further illustrate the meter of the method for solving and reset error of two-dimentional light intensity transmission matrix in imaging recovery technique
Calculate, Figure 14 (a) gives the known two-dimentional original thing of the alphabetical F patterns containing 7 rows, 7 column elements, and Figure 14 (b) gives alphabetical F
Two-dimentional original thing matrix known to pattern, in recuperation, it is believed that the original thing matrix is unknown, answering after reconstruction
In the calculating process of former error, the original thing matrix is known.Figure 15 gives certain optical system PSF matrix data, the light
System is linear sky invariant system.Figure 16 (a) gives the picture of alphabetical F things, and Figure 16 (b) gives two-dimensional image matrix, can be with
Find out, due to optical system imaging quality it is poor, as obscuring very much, and two-dimensional image matrix data and known two-dimentional original thing square
Battle array data differ greatly.Two-dimentional light intensity transmission matrix is built according to the method described in formula (3) and example one, be obtained 49 rows,
The two-dimentional light intensity transmission matrix of 49 row, as the element number of the matrix is excessive, inconvenience is listed as Figure 13 one by one, and Figure 17 gives
The graphics of the two-dimentional light intensity transmission matrix data drawn with matlab softwares is gone out, wherein x coordinate represents the row sequence of the matrix
Number, the y coordinate representation matrix column sequence number, z coordinate represents the numerical value of the matrix element.As can be seen that two-dimentional light intensity transmission square
Battle array has 3 data strips, and intermediate strap data value is maximum, and both sides Rearrangments, band number are equal to the line number of PSF data.
According to formula (6) to Figure 17 in the mathematical operation that carries out finding the inverse matrix of two-dimentional light intensity transmission matrix two-dimentional light intensity transmission is obtained
Inverse matrix, Figure 18 give the graphics that inverse matrix data are transmitted with the two-dimentional light intensity that matlab softwares draw, wherein x coordinate table
Show row sequence number, the y coordinate representation matrix column sequence number of the matrix, z coordinate represents the numerical value of the matrix element.According to formula
(2) by the two-dimensional image matrix shown in Figure 16 (b) one-dimensional chemical conversion one-dimensional picture vector, one-dimensional picture vector is multiplied by by figure according to formula (7)
Two-dimentional light intensity transmission inverse matrix shown in 18 is just obtained one-dimensional reconstruction vector, one-dimensional reconstruction is vectorial according to formula (8)
Two dimensionization is just obtained two-dimentional reconstruction matrix, shown in reconstruction and two-dimentional reconstruction matrix such as Figure 19 (a), (b).According to formula
(9) known two-dimentional original thing matrix in Figure 14 (b) is deducted into the two-dimentional reconstruction matrix in Figure 19 (b) and two-dimentional recovery is obtained
Error matrix, the data of two-dimentional reset error matrix are as shown in figure 20.Two-dimentional reset error can be calculated according to formula (10)
Root-mean-square value RMSEFor 8.6787 × 10-14, it is seen then that the imaging restored method described in this patent has minimum reset error,
In the case of known optical systems PSF, accurately can restore two-dimentional original thing matrix by two-dimensional image matrix.
Example three:
In order to verify recovery effect of this restored method to the picture matrix of big data quantity, Figure 21 give containing 400 rows,
The known original thing of 400 column elements, Figure 22 gives the PSF matrixes of optical system, and (optical system is linear sky invariant system
System), Figure 23 gives the original thing Jing optical system imagings shown in Figure 21.As can be seen that due to optical system imaging matter
Amount is very poor, and original thing imaging is obscured very much.The two-dimensional matrix of the picture in Figure 23 is carried out one-dimensional first to be contained
The one-dimensional picture vector of 160000 elements, then according to formula (3) and the method described in example one build two-dimentional light intensity transmission square
Battle array, is obtained 160000 rows, the two-dimentional light intensity transmission matrixs of 160000 row, due to matrix element number excessively, matlab numbers
Value analysis software cannot draw the datagraphic of the matrix.According to formula (6) to 160000 rows, the two-dimentional light intensity of 160000 row
Transmission matrix finding the inverse matrix is obtained the two-dimentional light intensity transmission inverse matrix of 160000 rows, 160000 row, similarly, since the matrix
Element number is excessive, and matlab numerical analysis softwares cannot draw the datagraphic that two-dimentional light intensity transmits inverse matrix.According to formula
(7) one-dimensional picture vector is multiplied by into two-dimentional light intensity transmission inverse matrix, one-dimensional reconstruction vector is obtained.According to formula (8), will be one-dimensional
Reconstruction vector two dimensionization is just obtained two-dimentional reconstruction matrix.Two-dimentional reconstruction is as shown in figure 24, it can be seen that in Figure 24
Reconstruction there is no difference with the original thing in Figure 21.According to formula (9) by two-dimentional original thing matrix known in Figure 21
The two-dimentional reconstruction matrix for deducting Figure 24 is obtained two-dimentional reset error matrix, is drawn using the data of two-dimentional reset error matrix
Reset error distribution map as shown in figure 25, wherein x coordinate represents the row sequence number of the matrix, the y coordinate representation matrix column sequence
Number, z coordinate represents the numerical value of the matrix element.As can be seen that each element and two dimension in known two-dimentional original thing matrix
The difference of each element in reconstruction matrix is less than 6 × 10-12.Reset error root-mean-square value can be calculated according to formula (10)
RMSEFor 2.6293 × 10-13.It can be seen that, the recovery technique described in this patent still has minimum recovery for the picture of big data quantity
Error, in the case of known optical systems PSF, the big data quantity such as accurately can restore by the two-dimensional image matrix of big data quantity
Two-dimentional original thing matrix.
Claims (6)
1. a kind of optical system imaging restored method based on light intensity transmission matrix, it is characterised in that the step of methods described such as
Under:
The first step, the one-dimensional of two-dimensional image matrix are processed:
When the one-dimensional of two-dimensional image matrix is processed, one is set up with the identical matrix of two-dimensional image matrix data amount as two
Unknown original thing matrix is tieed up, to the unknown original thing matrix of two dimension, after carrying out one-dimensional as matrix, the unknown original thing matrix of two dimension
Certain element aM, nRespective element α being changed in one-dimensional unknown original thing vectori, certain element b of two-dimensional image matrixQ, wTurn
Respective element β being changed in one-dimensional picture vectorj, following relation should be met:
Wherein, m, n are the sequence number of certain element in the unknown original thing matrix of two dimension, and i is certain in one-dimensional unknown original thing vector
The sequence number of element, q, w are the sequence number of certain element in two-dimensional image matrix, and j is the sequence number of certain element in one-dimensional picture vector, M, N
For the line number of the unknown original thing matrix of two dimension, columns, Q, W are total line number, the total columns as matrix, and i, j, m, n, q, w, M, N,
Q, W are positive integer, and M is equal with Q, N is equal with W;
Second step, two-dimentional light intensity transmission matrix and two-dimentional light intensity transmit the solution of inverse matrix:
The two-dimentional light intensity transmission matrix of optical system is solved according to the PSF matrixes and one-dimensional picture vector of optical system;By two-dimentional light
The mathematical computations inverted by strong transmission matrix, obtain two-dimentional light intensity transmission inverse matrix;
The solution procedure of described two-dimentional light intensity transmission matrix is as follows:
From the PSF extracting datas of optical system by the unknown original thing matrix of two dimension certain element representation light intensity aM, nTransmission
To light intensity b of certain element representation of two-dimensional image matrixQ, wFour-dimensional light intensity transmission coefficient dM, n, q, w, dM, n, q, wWith one-dimensional unknown original
Element α in beginning thing vectoriTo the element β in one-dimensional picture vectorjTwo-dimentional light intensity transmission coefficient pI, jIt is equal, pI, jFor two
Element in dimension light intensity transmission matrix, i.e.,:
Wherein, m, n are the sequence number of certain element in the unknown original thing matrix of two dimension, and i is certain in one-dimensional unknown original thing vector
The sequence number of element, q, w are the sequence number of certain element in two-dimensional image matrix, and j is the sequence number of certain element in one-dimensional picture vector, M, N
For the line number of the unknown original thing matrix of two dimension, columns, Q, W total line number, total columns for two-dimensional image matrix, and i, j, m, n, q, w,
M, N, Q, W are positive integer, and M is equal with Q, N is equal with W;Obtain one-dimensional unknown original thing vector according to the method described above successively each
Element is just obtained two-dimentional light intensity transmission matrix to the light intensity transmission coefficient of one-dimensional picture vector each element;
3rd step, the solution of reconstruction matrix:
One-dimensional picture vector is multiplied by into two-dimentional light intensity transmission inverse matrix and tries to achieve one-dimensional reconstruction vector, by one-dimensional reconstruction vector two dimension
Two-dimentional reconstruction matrix is obtained after change.
2. the optical system imaging restored method based on light intensity transmission matrix according to claim 1, it is characterised in that institute
State in step 2, two-dimentional light intensity transmission matrix has following relation with two-dimentional light intensity transmission inverse matrix:
P·P-1=E;
Wherein, matrix P is two-dimentional light intensity transmission matrix, matrix P-1Inverse matrix is transmitted for two-dimentional light intensity, matrix E is and P exponent number phases
Same unit matrix.
3. the optical system imaging restored method based on light intensity transmission matrix according to claim 2, it is characterised in that two
Dimension light intensity transmission matrix P of matrix-1Ask method as follows:
Wherein, determinants of | the P | for matrix P, matrix P*For the adjoint matrix that the algebraic complement of matrix P each element is constituted.
4. the optical system imaging restored method based on light intensity transmission matrix according to claim 1,2 or 3, its feature exist
In the step 3, the one-dimensional picture vector that the first step is obtained is multiplied by into the two-dimentional light intensity transmission inverse matrix that second step obtains and just may be used
One-dimensional reconstruction vector is obtained, i.e.,
B·P-1=C;
B is that one-dimensional picture is vectorial, P-1Inverse matrix is transmitted for two-dimentional light intensity, C is one-dimensional reconstruction vector.
5. the optical system imaging restored method based on light intensity transmission matrix according to claim 1, it is characterised in that institute
State in step 3, one-dimensional reconstruction vector is just obtained into two-dimentional reconstruction matrix according to equation below two dimensionization:
Wherein, aK, l' it is element in two-dimentional reconstruction matrix, αc' it is element in one-dimensional reconstruction vector C, k, l are two dimension
The sequence number of certain element in reconstruction matrix, c are the sequence number of certain element in one-dimensional reconstruction vector, and K, L are two-dimentional reconstruction
Total line number of matrix, total columns, K are equal to Q, and L is equal to W, Q, W total line number, total columns for two-dimensional image matrix, and c, k, l, K, L
It is positive integer.
6. the optical system imaging restored method based on light intensity transmission matrix according to claim 1, it is characterised in that institute
State in step 3, known two-dimentional original thing matrix is subtracted each other with two-dimentional reconstruction matrix and be obtained error matrix, Jin Erqiu
Go out error mean square root RMS of known two-dimentional original thing matrix and corresponding element in two-dimentional reconstruction matrixE, RMSESize
For evaluating the recovery accuracy of imaging restored method.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510009310.4A CN104574315B (en) | 2015-01-08 | 2015-01-08 | Optical system imaging recovering method based on light intensity transmission matrix |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510009310.4A CN104574315B (en) | 2015-01-08 | 2015-01-08 | Optical system imaging recovering method based on light intensity transmission matrix |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104574315A CN104574315A (en) | 2015-04-29 |
CN104574315B true CN104574315B (en) | 2017-03-22 |
Family
ID=53090296
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510009310.4A Active CN104574315B (en) | 2015-01-08 | 2015-01-08 | Optical system imaging recovering method based on light intensity transmission matrix |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104574315B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105608058A (en) * | 2015-12-23 | 2016-05-25 | 哈尔滨工业大学 | Finite-dimension eigen frequency domain analysis method for optical imaging system |
CN105571835B (en) * | 2015-12-24 | 2017-11-10 | 哈尔滨工业大学 | A kind of imaging measurement method of the photo electric imaging system point spread function based on matrix operation |
CN112528514B (en) * | 2020-12-21 | 2024-02-23 | 北京机电工程研究所 | High-precision sub-pixel star spot remodelling method and device |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103971341A (en) * | 2014-05-16 | 2014-08-06 | 北京理工大学 | Out-of-focus iris image restoration method based on frequency domain parameter estimation |
CN104091314A (en) * | 2014-07-22 | 2014-10-08 | 西北工业大学 | Turbulence-degraded image blind restoration method based on edge prediction and sparse ratio regular constraints |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR101667798B1 (en) * | 2010-03-29 | 2016-10-20 | 삼성전자 주식회사 | Image Processing device and image processing method |
-
2015
- 2015-01-08 CN CN201510009310.4A patent/CN104574315B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103971341A (en) * | 2014-05-16 | 2014-08-06 | 北京理工大学 | Out-of-focus iris image restoration method based on frequency domain parameter estimation |
CN104091314A (en) * | 2014-07-22 | 2014-10-08 | 西北工业大学 | Turbulence-degraded image blind restoration method based on edge prediction and sparse ratio regular constraints |
Non-Patent Citations (3)
Title |
---|
Blind image deconvolution;Kurdur D 等;《IEEE Single Processing Magazine》;19960531;43-64 * |
一种自适应正则化技术的图像复原方法;苗晴;《电子设计工程》;20140920;第22卷(第18期);169-171,175 * |
图像盲复原算法研究现状及其展望;张航 等;《中国图象图形学报》;20041031;第9卷(第10期);1145-1152 * |
Also Published As
Publication number | Publication date |
---|---|
CN104574315A (en) | 2015-04-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103077505B (en) | Based on the image super-resolution rebuilding method of dictionary learning and documents structured Cluster | |
CN110119780B (en) | Hyper-spectral image super-resolution reconstruction method based on generation countermeasure network | |
CN105488776B (en) | Super-resolution image reconstruction method and device | |
CN109767000B (en) | Neural network convolution method and device based on Winograd algorithm | |
CN109064396A (en) | A kind of single image super resolution ratio reconstruction method based on depth ingredient learning network | |
CN111080567A (en) | Remote sensing image fusion method and system based on multi-scale dynamic convolution neural network | |
CN104574315B (en) | Optical system imaging recovering method based on light intensity transmission matrix | |
CN110139046B (en) | Tensor-based video frame synthesis method | |
CN1327810C (en) | Method for removing Gibbs ring pseudo-image generated in MRI reconstruction process | |
CN105225207B (en) | A kind of compressed sensing imaging and image rebuilding method based on observing matrix | |
CN108288256A (en) | A kind of multispectral mosaic image restored method | |
CN112381711B (en) | Training and quick super-resolution reconstruction method for light field image reconstruction model | |
CN108198149B (en) | A kind of image deblurring method | |
CN102609920B (en) | Colorful digital image repairing method based on compressed sensing | |
CN106470293B (en) | Image processing apparatus and image processing method | |
CN107170018A (en) | Constitution optimization method based on compressed sensing calculation matrix in image reconstruction | |
Li et al. | A pseudo-siamese deep convolutional neural network for spatiotemporal satellite image fusion | |
CN107451954A (en) | Iterated pixel interpolation method based on image low-rank property | |
CN113008371B (en) | Hyperspectral imaging method for deep learning dispersion-based fuzzy solution | |
He et al. | Deep frequency-recurrent priors for inverse imaging reconstruction | |
CN104794691A (en) | Definition reconstruction method of single out-of-focus image using generalized Gaussian model | |
CN103824262A (en) | Improved rapid image restoration processing method using R-L iterative algorithm | |
CN116862787A (en) | Image restoration method, system, equipment and storage medium based on tensor singular value decomposition | |
CN113450319B (en) | Super-resolution reconstruction image quality evaluation method based on KLT technology | |
CN104008532B (en) | A kind of infrared image based on prior probability model removes striping methods |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant |