CN111047521A - Infrared image nonuniformity parameterization correction optimization method based on image entropy - Google Patents

Infrared image nonuniformity parameterization correction optimization method based on image entropy Download PDF

Info

Publication number
CN111047521A
CN111047521A CN201911065298.3A CN201911065298A CN111047521A CN 111047521 A CN111047521 A CN 111047521A CN 201911065298 A CN201911065298 A CN 201911065298A CN 111047521 A CN111047521 A CN 111047521A
Authority
CN
China
Prior art keywords
image
entropy
column
value
calculating
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN201911065298.3A
Other languages
Chinese (zh)
Other versions
CN111047521B (en
Inventor
文高进
何林
钟灿
王哲
龙亮
黄浦
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Institute of Space Research Mechanical and Electricity
Original Assignee
Beijing Institute of Space Research Mechanical and Electricity
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing Institute of Space Research Mechanical and Electricity filed Critical Beijing Institute of Space Research Mechanical and Electricity
Priority to CN201911065298.3A priority Critical patent/CN111047521B/en
Publication of CN111047521A publication Critical patent/CN111047521A/en
Application granted granted Critical
Publication of CN111047521B publication Critical patent/CN111047521B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G06T5/90—
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/50—Image enhancement or restoration by the use of more than one image, e.g. averaging, subtraction
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10048—Infrared image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20212—Image combination
    • G06T2207/20216—Image averaging
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20212—Image combination
    • G06T2207/20224—Image subtraction

Abstract

The invention discloses an infrared non-uniformity parameterization correction optimization method based on image entropy. The specific implementation steps are as follows: 1) calculating the entropies of the odd pixel image and the even pixel image of the original image, and selecting the image with the larger entropy value as a reference image; 2) carrying out non-uniformity parametric correction optimization processing on the reference image based on the image entropy to obtain a corrected reference image, 3) carrying out non-uniformity correction based on linear approximation on the corrected reference image, and finally obtaining a result image after the original image is corrected. The invention makes up the defects of the traditional infrared image non-uniformity correction method and provides a more effective non-uniformity correction method based on the image entropy quality evaluation standard. The invention can effectively remove the non-uniform strip noise in the infrared image, furthest maintain the original information of the image and obtain the high-quality infrared correction image.

Description

Infrared image nonuniformity parameterization correction optimization method based on image entropy
Technical Field
The invention relates to the field of infrared image processing, in particular to an infrared image non-uniformity parameter correction optimization method based on image entropy.
Background
Due to the influence of manufacturing processes and materials, the response transfer function of each pixel of the infrared detector is different, so that stripe noise, called non-uniformity noise, exists in the obtained infrared image. The noise seriously reduces the image quality, greatly hinders the wide application of infrared imaging in the fields of medical treatment, monitoring, agriculture and forestry and the like, and therefore when the infrared remote sensing is carried out, the non-uniformity correction is required to improve the quality of the infrared image.
The non-uniformity correction method for infrared images mainly comprises two types: calibration-based methods and scene-based methods. The calibration-based method requires a standard radiation source, such as a black body, and is very limited in practical applications. The scene-based method extracts parameters according to the features of images in the scene to realize non-uniformity correction, has better applicability, and is widely used. Generally, a scene-based correction method adopts the roughness of a corrected image as an evaluation index, which easily causes the corrected image to be excessively smooth and loses original information contained in the image.
The image entropy expresses the information content contained in the gray level distribution aggregation characteristics in the image, reflects the definition of the image, and is a common quantization standard for evaluating the image.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: the method overcomes the defects of the prior art and provides the infrared image heterogeneity parameterization correction optimal selection method based on the image entropy, which has large image information amount and high definition.
The technical solution of the invention is as follows: an infrared image nonuniformity parameterization correction optimization method based on image entropy comprises the following steps:
(1) calculating the image entropy of an odd pixel image and an even pixel image of an input image X, selecting an image with a large entropy as a reference image B, wherein the dimension of the reference image B is m rows n1 columns, recording the entropy e of the reference image B, recording another image as an image C, and the dimension of the image C is m rows n2 columns; wherein m and n1, n2 are positive integers;
(2) the method comprises the following steps of performing non-uniformity parameterization correction optimization processing based on image entropy on a reference image B to obtain a corrected reference image, and specifically comprises the following steps:
(21) initializing k to 1;
(22) assigning the reference image B to the process image P, and initializing i to 1;
(23) calculating the gray value difference of corresponding pixels of the (i + 1) th column and the ith column in the process image P to obtain an m-dimensional column difference vector V, and sequencing the absolute values of all elements of the column difference vector V from small to large to obtain a sequenced absolute value vector U;
(24) calculating an index array Q of all elements with absolute values smaller than U (k) in the column difference vector V, wherein the index array Q is a Q-element array, and calculating an update parameter according to the following formula:
Figure BDA0002257988930000021
Figure BDA0002257988930000022
Figure BDA0002257988930000023
Figure BDA0002257988930000024
P(j,i+1)＝aP(j,i+1)+b，j＝1,2,3,...,m；
u (k) is the kth element in the absolute value vector U;
r1、r2、r3、r4respectively intermediate amounts;
p (x, y) is the value of the x row and the y column in the process image P;
a. b are correction coefficients respectively; q (j) is the jth element in the index array Q;
(25) if i < ═ n-1, updating i < ═ i +1, returning to the step (23), otherwise, entering the step (26);
(26) calculating an entropy value t of the current process image P, if the entropy value t is larger than the entropy value e, updating the entropy value e to the entropy value t, and updating the current result image R to be the process image P;
(27) if it is not
Figure BDA0002257988930000025
Updating k to k +1, and returning to the step (22); otherwise, entering a step (28);
(28) updating the reference image B into a result image R;
(3) and the other image C performs non-uniformity correction based on linear approximation on the reference image B to finally obtain a corrected result, and the specific steps are as follows:
(31) and calculating an image D to be approximated from the reference image B:
if the image C is an odd pixel image of the input image X, assigning the 1 st column of the image D as the 1 st column of the image B, and sequentially assigning the last n-1 columns of the image D as average images of two adjacent columns at corresponding positions of the image B; otherwise, if the image C is an even image element image of the input image X, assigning the first n-1 columns of the image D as average images of two adjacent columns at the corresponding position of the image B in sequence, and assigning the last column of the image D as the last column of the image B;
(32) traversing g from 1 to n, calculating a left-end matrix F and a right-end vector H:
Figure BDA0002257988930000031
F(2g,2g)＝2m，
Figure BDA0002257988930000032
Figure BDA0002257988930000033
g＝1,2,3…n；j＝1,2,3…m；
f (x, y) is the value of the x row and the y column in the left matrix F;
c (x, y) is the value of the x row and the y column in the image C;
h (x) is the x-th element in the right-end vector H;
d (x, y) is the value of the x row and the y column in the image D;
(33) calculating a linear approximation parameter T ═ inv (F) H; inv (F) is the inversion of the matrix F;
(34) if the image C is an odd pixel image of the input image X, the even pixel gray scale of the image Y is set as the corresponding pixel gray scale of the reference image B, and the odd pixel of the image Y is updated from 1 to n by traversing g as follows:
y (h,2g-1) ═ T (2g-1) C (h, g) + T (2g), where h is 1,2,3,.., m;
t (2g-1) and T (2g) are elements of the 2g-1 st and the 2g nd in the matrix T respectively; y (x, Y) is the value of the x row and the Y column in the image Y;
otherwise, if the image C is an odd pixel image of the input image X, the odd pixel gray scale of the image Y is set to the pixel gray scale corresponding to the reference image B, and the even pixel of the image Y is updated from 1 to n by traversing g as follows:
y (h,2g) ═ T (2g-1) C (h, g) + T (2g), where h is 1,2,3,. said, m;
image Y is the final corrected image.
An infrared image nonuniformity parameterization correction optimization system based on image entropy comprises:
the image entropy calculation method comprises a first module, a second module and a third module, wherein the first module is used for calculating the image entropy of an odd pixel image and an even pixel image of an input image X, selecting an image with a large entropy as a reference image B, the dimension of the reference image B is m rows and n1 columns, recording the entropy e of the reference image B, marking the other image as an image C, and the dimension of the image C is m rows and n2 columns; wherein m and n1, n2 are positive integers;
the second module is used for carrying out non-uniformity parametric correction optimization processing on the reference image B based on the image entropy to obtain a corrected reference image;
and the third module is used for carrying out non-uniformity correction based on linear approximation on the other image C to the reference image B, and finally obtaining a corrected result.
In the second module, a specific method for obtaining the corrected reference image is as follows:
(21) initializing k to 1;
(22) assigning the reference image B to the process image P, and initializing i to 1;
(23) calculating the gray value difference of corresponding pixels of the (i + 1) th column and the ith column in the process image P to obtain an m-dimensional column difference vector V, and sequencing the absolute values of all elements of the column difference vector V from small to large to obtain a sequenced absolute value vector U;
(24) calculating an index array Q of all elements with absolute values smaller than U (k) in the column difference vector V, wherein the index array Q is a Q-element array, and calculating an update parameter according to the following formula:
Figure BDA0002257988930000041
Figure BDA0002257988930000042
Figure BDA0002257988930000043
Figure BDA0002257988930000044
P(j,i+1)＝aP(i,i+1)+b，j＝1,2,3,...,m；
u (k) is the kth element in the absolute value vector U;
r1、r2、r3、r4respectively intermediate amounts;
p (x, y) is the value of the x row and the y column in the process image P;
a. b are correction coefficients respectively; q (j) is the jth element in the index array Q;
(25) if i < ═ n-1, updating i < ═ i +1, returning to the step (23), otherwise, entering the step (26);
(26) calculating an entropy value t of the current process image P, if the entropy value t is larger than the entropy value e, updating the entropy value e to the entropy value t, and updating the current result image R to be the process image P;
(27) if it is not
Figure BDA0002257988930000051
Updating k to k +1, and returning to the step (22); otherwise, entering a step (28);
(28) the reference image B is updated to the result map R.
In the third module, the non-uniformity correction based on linear approximation is performed on the other image C to the reference image B, and the specific method for obtaining the correction result is as follows:
(31) and calculating an image D to be approximated from the reference image B:
if the image C is an odd pixel image of the input image X, assigning the 1 st column of the image D as the 1 st column of the image B, and sequentially assigning the last n-1 columns of the image D as average images of two adjacent columns at corresponding positions of the image B; otherwise, if the image C is an even image element image of the input image X, assigning the first n-1 columns of the image D as average images of two adjacent columns at the corresponding position of the image B in sequence, and assigning the last column of the image D as the last column of the image B;
(32) traversing g from 1 to n, calculating a left-end matrix F and a right-end vector H:
Figure BDA0002257988930000052
F(2g,2g)＝2m，
Figure BDA0002257988930000053
Figure BDA0002257988930000054
g＝1,2,3…n；j＝1,2,3…m；
f (x, y) is the value of the x row and the y column in the left matrix F;
c (x, y) is the value of the x row and the y column in the image C;
h (x) is the x-th element in the right-end vector H;
d (x, y) is the value of the x row and the y column in the image D;
(33) calculating a linear approximation parameter T ═ inv (F) H; inv (F) is the inversion of the matrix F;
(34) if the image C is an odd pixel image of the input image X, the even pixel gray scale of the image Y is set as the corresponding pixel gray scale of the reference image B, and the odd pixel of the image Y is updated from 1 to n by traversing g as follows:
y (h,2g-1) ═ T (2g-1) C (h, g) + T (2g), where h is 1,2,3,.., m;
t (2g-1) and T (2g) are elements of the 2g-1 st and the 2g nd in the matrix T respectively; y (x, Y) is the value of the x row and the Y column in the image Y;
otherwise, if the image C is an odd pixel image of the input image X, the odd pixel gray scale of the image Y is set to the pixel gray scale corresponding to the reference image B, and the even pixel of the image Y is updated from 1 to n by traversing g as follows:
y (h,2g) ═ T (2g-1) C (h, g) + T (2g), where h is 1,2,3,. said, m;
image Y is the final corrected image.
Compared with the prior art, the invention has the advantages that:
(1) according to the method, the infrared image quality after infrared non-uniformity correction is evaluated from the angle of image entropy, so that the problem that the quality evaluation of the traditional infrared image non-uniformity correction method depends on roughness indexes and cannot represent the limitation of image information quantity is avoided;
(2) the method provides a selection mechanism of the infrared image non-uniformity parametric correction method, so that the existing infrared image non-uniformity parametric correction method is very easy to expand, and the infrared image non-uniformity parametric correction method with better performance is convenient to generate;
(3) the method effectively combines the information of the odd pixel image and the even pixel image of the original infrared image to carry out the non-uniformity correction method, and has better infrared image correction effect which is obviously influenced by odd-even non-uniformity compared with the traditional infrared image non-uniformity correction method;
(4) the method ensures the maximization of the image information quantity while correcting the non-uniformity of the infrared image, and the corrected infrared image has higher definition and target recognition.
Drawings
FIG. 1 is a flow chart of the method of the present invention;
FIG. 2 is a 269X 384 infrared image to be nonuniformity corrected;
FIG. 3 is a graph of entropy values for process images corresponding to different K values;
FIG. 4 is a graph of the results of an example infrared image after non-uniformity correction.
Detailed Description
The present invention will be described in further detail with reference to examples and drawings, but the present invention is not limited thereto.
Taking 269 × 384 infrared image to be corrected for non-uniformity as an example, the flow of the infrared image non-uniformity parameterization correction preferred method based on image entropy is shown in fig. 1, and the specific process is as follows:
(1) the image entropy of the input image X odd pixel image and even pixel image is calculated, an image with a large entropy is selected as a reference image B (m rows n1 columns), the entropy e is recorded, and the other image is marked as C (m rows n2 columns). The image entropy can be calculated by selecting one-dimensional entropy or two-dimensional entropy, wherein m and n1 and n2 are positive integers.
In this embodiment, as shown in fig. 2, the input infrared image X to be subjected to the non-uniformity correction is a 220 × 320 matrix, the odd pixel images are X odd with m 269 rows and n1 192 columns, the corresponding pixel columns are the 1 st, 3 rd, 5 th,. 383 columns of X, the even images are m 269 rows and n2 192 columns, and the corresponding pixel columns are the 2 nd, 4 th, 6 th,. 384 th columns of X. The image entropy is calculated by using a one-dimensional entropy calculation formula, wherein the image entropy of X odd is 5.8094, and the image entropy of X even is 5.9658, so that the reference image B is an odd image, the other image C is an even image, and the image entropy of the original image X is 6.1223.
(2) The method comprises the following steps of carrying out nonuniformity parameterization correction based on image entropy on a reference image B to obtain a corrected reference image, and specifically comprises the following steps:
(21) traversal parameter k from 1 to
Figure BDA0002257988930000071
Carrying out the following steps, wherein symbols representTaking the maximum integer not greater than b;
(22) assigning the reference image B to a process image P, traversing the process image P according to columns, and sequentially initializing i to 1;
(23) calculating gray value differences of corresponding pixels of an i +1 th column and an i th column in the P to obtain a m-dimensional column difference vector V, and sequencing absolute values of elements of the column difference vector V from small to large to obtain a sequenced absolute value vector U;
(24) calculating an index array Q (Q-element array) of all elements with absolute values smaller than U (k) in V, and calculating an updating parameter according to the following formula:
Figure BDA0002257988930000081
Figure BDA0002257988930000082
Figure BDA0002257988930000083
Figure BDA0002257988930000084
P(j,i+1)＝aP(j,i+1)+b,j＝1,2,3,...,m；
u (k) is the kth element in the absolute value vector U;
r1、r2、r3、r4respectively intermediate amounts;
p (x, y) is the value of the x row and the y column in the process image P;
a. b are correction coefficients respectively; q (j) is the jth element in the index array Q;
(25) if i < ═ n-1, updating i < ═ i +1, returning to step (23), otherwise, entering step (26)
(26) Calculating an entropy value t of the current process image P, if t is larger than e, updating e to t, and updating a current result image R to be P;
(27) if it is not
Figure BDA0002257988930000085
Updating k to k +1, returning to step (22), otherwise, entering step (28)
Fig. 3 is a graph of entropy for process images for different values of K, where the maximum entropy is 6.5362 and occurs at the location where K is 82.
(28) Updating the reference image B into a result image R;
(3) and the other image C performs non-uniformity correction based on linear approximation on the reference image B to finally obtain a corrected result, and the specific steps are as follows:
(31) calculating to-be-approximated image D from reference image B
If the image C is an odd pixel image of the original image, assigning the 1 st column of the image D as the 1 st column of the image B, sequentially assigning the last n-1 columns of the image D as average images of two adjacent columns at the corresponding position of the image B, and otherwise, sequentially assigning the first n-1 columns of the image D as average images of two adjacent columns at the corresponding position of the image B, and the last column of the image D as the last column of the image B.
(32) Calculating a left-end matrix F and a right-end vector H by traversing g from 1 to n by:
Figure BDA0002257988930000091
F(2g,2g)＝2m，
Figure BDA0002257988930000092
Figure BDA0002257988930000093
f (x, y) is the value of the x row and the y column in the left matrix F;
c (x, y) is the value of the x row and the y column in the image C;
h (x) is the x-th element in the right-end vector H;
d (x, y) is the value of the x row and the y column in the image D;
(33) calculating a linear approximation parameter T ═ inv (F) H,
(34) if the image C is an odd pixel image of the original image, the even pixel gray scale of the image Y is set as the pixel gray scale corresponding to the reference image B, and the odd pixel of the image Y is updated by traversing g from 1 to n according to the following steps:
y (h,2g-1) ═ T (2g-1) C (h, g) + T (2g), where h is 1,2,3,.., m;
t (2g-1) and T (2g) are elements of the 2g-1 st and the 2g nd in the matrix T respectively; y (x, Y) is the value of the x row and the Y column in the image Y;
otherwise, the odd pixel gray scale of the image Y is set as the corresponding pixel gray scale of the reference image B, and the even pixel of the image Y is updated by traversing g from 1 to n according to the following steps
Y (h,2g) ═ T (2g-1) C (h, g) + T (2g), where h is 1,2,3,. said, m;
image Y is the final corrected image.
The image Y calculated in this embodiment is shown in fig. 4, and the image entropy of the result graph is 6.6785 and the roughness index 0.0862.
An infrared image nonuniformity parameterization correction optimization system based on image entropy comprises:
the image entropy calculation method comprises a first module, a second module and a third module, wherein the first module is used for calculating the image entropy of an odd pixel image and an even pixel image of an input image X, selecting an image with a large entropy as a reference image B, the dimension of the reference image B is m rows and n1 columns, recording the entropy e of the reference image B, marking the other image as an image C, and the dimension of the image C is m rows and n2 columns; wherein m and n1, n2 are positive integers;
the second module is used for carrying out non-uniformity parametric correction optimization processing on the reference image B based on the image entropy to obtain a corrected reference image;
and the third module is used for carrying out non-uniformity correction based on linear approximation on the other image C to the reference image B, and finally obtaining a corrected result.
In the second module, a specific method for obtaining the corrected reference image is as follows:
(21) initializing k to 1;
(22) assigning the reference image B to the process image P, and initializing i to 1;
(23) calculating the gray value difference of corresponding pixels of the (i + 1) th column and the ith column in the process image P to obtain an m-dimensional column difference vector V, and sequencing the absolute values of all elements of the column difference vector V from small to large to obtain a sequenced absolute value vector U;
(24) calculating an index array Q of all elements with absolute values smaller than U (k) in the column difference vector V, wherein the index array Q is a Q-element array, and calculating an update parameter according to the following formula:
Figure BDA0002257988930000101
Figure BDA0002257988930000102
Figure BDA0002257988930000103
Figure BDA0002257988930000104
P(j,i+1)＝aP(j,i+1)+b，j＝1,2,3,...,m；
u (k) is the kth element in the absolute value vector U;
r1、r2、r3、r4respectively intermediate amounts;
p (x, y) is the value of the x row and the y column in the process image P;
a. b are correction coefficients respectively; q (j) is the jth element in the index array Q;
(25) if i < ═ n-1, updating i < ═ i +1, returning to the step (23), otherwise, entering the step (26);
(26) calculating an entropy value t of the current process image P, if the entropy value t is larger than the entropy value e, updating the entropy value e to the entropy value t, and updating the current result image R to be the process image P;
(27) if it is not
Figure BDA0002257988930000111
Updating k to k +1, and returning to the step (22); otherwise, entering a step (28);
(28) the reference image B is updated to the result map R.
In the third module, the non-uniformity correction based on linear approximation is performed on the other image C to the reference image B, and the specific method for obtaining the correction result is as follows:
(31) and calculating an image D to be approximated from the reference image B:
if the image C is an odd pixel image of the input image X, assigning the 1 st column of the image D as the 1 st column of the image B, and sequentially assigning the last n-1 columns of the image D as average images of two adjacent columns at corresponding positions of the image B; otherwise, if the image C is an even image element image of the input image X, assigning the first n-1 columns of the image D as average images of two adjacent columns at the corresponding position of the image B in sequence, and assigning the last column of the image D as the last column of the image B;
(32) traversing g from 1 to n, calculating a left-end matrix F and a right-end vector H:
Figure BDA0002257988930000112
F(2g，2g)＝2m，
Figure BDA0002257988930000113
Figure BDA0002257988930000114
g＝1,2,3…n；j＝1,2,3…m；
f (x, y) is the value of the x row and the y column in the left matrix F;
c (x, y) is the value of the x row and the y column in the image C;
h (x) is the x-th element in the right-end vector H;
d (x, y) is the value of the x row and the y column in the image D;
(33) calculating a linear approximation parameter T ═ inv (F) H; inv (F) is the inversion of the matrix F;
(34) if the image C is an odd pixel image of the input image X, the even pixel gray scale of the image Y is set as the corresponding pixel gray scale of the reference image B, and the odd pixel of the image Y is updated from 1 to n by traversing g as follows:
y (h,2g-1) ═ T (2g-1) C (h, g) + T (2g), where h is 1,2,3,.., m;
t (2g-1) and T (2g) are elements of the 2g-1 st and the 2g nd in the matrix T respectively; y (x, Y) is the value of the x row and the Y column in the image Y;
otherwise, if the image C is an odd pixel image of the input image X, the odd pixel gray scale of the image Y is set to the pixel gray scale corresponding to the reference image B, and the even pixel of the image Y is updated from 1 to n by traversing g as follows:
y (h,2g) ═ T (2g-1) C (h, g) + T (2g), where h is 1,2,3,. said, m;
image Y is the final corrected image.
The present invention has not been described in detail, partly as is known to the person skilled in the art.

Claims (6)

1. An infrared image nonuniformity parameterization correction optimization method based on image entropy is characterized by comprising the following steps:
(1) calculating the image entropy of an odd pixel image and an even pixel image of an input image X, selecting an image with a large entropy as a reference image B, wherein the dimension of the reference image B is m rows n1 columns, recording the entropy e of the reference image B, recording another image as an image C, and the dimension of the image C is m rows n2 columns; wherein m and n1, n2 are positive integers;
(2) carrying out non-uniformity parametric correction optimization processing on the reference image B based on the image entropy to obtain a corrected reference image;
(3) and performing non-uniformity correction based on linear approximation on the other image C to the reference image B to obtain a corrected result.
2. The method for preferably correcting the infrared image nonuniformity based on the image entropy as claimed in claim 1, wherein the specific steps of the step (2) are as follows:
(21) initializing k to 1;
(22) assigning the reference image B to the process image P, and initializing i to 1;
(23) calculating the gray value difference of corresponding pixels of the (i + 1) th column and the ith column in the process image P to obtain an m-dimensional column difference vector V, and sequencing the absolute values of all elements of the column difference vector V from small to large to obtain a sequenced absolute value vector U;
(24) calculating an index array Q of all elements with absolute values smaller than U (k) in the column difference vector V, wherein the index array Q is a Q-element array, and calculating an update parameter according to the following formula:
Figure RE-FDA0002337120010000011
Figure RE-FDA0002337120010000012
Figure RE-FDA0002337120010000013
Figure RE-FDA0002337120010000014
P(j,i+1)＝aP(j,i+1)+b，j＝1,2,3,...,m；
u (k) is the kth element in the absolute value vector U;
r1、r2、r3、r4respectively intermediate amounts;
p (x, y) is the value of the x row and the y column in the process image P;
a. b are correction coefficients respectively; q (j) is the jth element in the index array Q;
(25) if i < ═ n-1, updating i < ═ i +1, returning to the step (23), otherwise, entering the step (26);
(26) calculating an entropy value t of the current process image P, if the entropy value t is larger than the entropy value e, updating the entropy value e to the entropy value t, and updating the current result image R to be the process image P;
(27) if it is not
Figure RE-FDA0002337120010000021
Updating k to k +1, and returning to the step (22); otherwise, entering a step (28);
(28) the reference image B is updated to the result map R.
3. The method for preferably correcting infrared image nonuniformity based on image entropy as claimed in claim 1, wherein the specific steps of the step (3) are as follows:
(31) and calculating an image D to be approximated from the reference image B:
if the image C is an odd pixel image of the input image X, assigning the 1 st column of the image D as the 1 st column of the image B, and sequentially assigning the last n-1 columns of the image D as average images of two adjacent columns at corresponding positions of the image B; otherwise, if the image C is an even image element image of the input image X, assigning the first n-1 columns of the image D as average images of two adjacent columns at the corresponding position of the image B in sequence, and assigning the last column of the image D as the last column of the image B;
(32) traversing g from 1 to n, calculating a left-end matrix F and a right-end vector H:
Figure RE-FDA0002337120010000022
F(2g,2g)＝2m，
Figure RE-FDA0002337120010000023
Figure RE-FDA0002337120010000024
g＝1,2,3…n；j＝1,2,3…m；
f (x, y) is the value of the x row and the y column in the left matrix F;
c (x, y) is the value of the x row and the y column in the image C;
h (x) is the x-th element in the right-end vector H;
d (x, y) is the value of the x row and the y column in the image D;
(33) calculating a linear approximation parameter T ═ inv (F) H; inv (F) is the inversion of the matrix F;
(34) if the image C is an odd pixel image of the input image X, the even pixel gray scale of the image Y is set as the corresponding pixel gray scale of the reference image B, and the odd pixel of the image Y is updated from 1 to n by traversing g as follows:
y (h,2g-1) ═ T (2g-1) C (h, g) + T (2g), where h is 1,2,3,.., m;
t (2g-1) and T (2g) are elements of the 2g-1 st and the 2g nd in the matrix T respectively; y (x, Y) is the value of the x row and the Y column in the image Y;
otherwise, if the image C is an odd pixel image of the input image X, the odd pixel gray scale of the image Y is set to the pixel gray scale corresponding to the reference image B, and the even pixel of the image Y is updated from 1 to n by traversing g as follows:
y (h,2g) ═ T (2g-1) C (h, g) + T (2g), where h is 1,2,3,. said, m;
image Y is the final corrected image.
4. An infrared image nonuniformity parameterization correction optimization system based on image entropy is characterized by comprising:
the image entropy calculation method comprises a first module, a second module and a third module, wherein the first module is used for calculating the image entropy of an odd pixel image and an even pixel image of an input image X, selecting an image with a large entropy as a reference image B, the dimension of the reference image B is m rows and n1 columns, recording the entropy e of the reference image B, marking the other image as an image C, and the dimension of the image C is m rows and n2 columns; wherein m and n1, n2 are positive integers;
the second module is used for carrying out non-uniformity parametric correction optimization processing on the reference image B based on the image entropy to obtain a corrected reference image;
and the third module is used for carrying out non-uniformity correction based on linear approximation on the other image C to the reference image B, and finally obtaining a corrected result.
5. The infrared image nonuniformity parameterization correction and optimization system based on image entropy as claimed in claim 4, wherein in the second module, the specific method for obtaining the corrected reference image is as follows:
(21) initializing k to 1;
(22) assigning the reference image B to the process image P, and initializing i to 1;
(23) calculating the gray value difference of corresponding pixels of the (i + 1) th column and the ith column in the process image P to obtain an m-dimensional column difference vector V, and sequencing the absolute values of all elements of the column difference vector V from small to large to obtain a sequenced absolute value vector U;
(24) calculating an index array Q of all elements with absolute values smaller than U (k) in the column difference vector V, wherein the index array Q is a Q-element array, and calculating an update parameter according to the following formula:
Figure RE-FDA0002337120010000041
Figure RE-FDA0002337120010000042
Figure RE-FDA0002337120010000043
Figure RE-FDA0002337120010000044
P(j,i+1)＝aP(j,i+1)+b，j＝1,2,3,...,m；
u (k) is the kth element in the absolute value vector U;
r1、r2、r3、r4respectively intermediate amounts;
p (x, y) is the value of the x row and the y column in the process image P;
a. b are correction coefficients respectively; q (j) is the jth element in the index array Q;
(25) if i < ═ n-1, updating i < ═ i +1, returning to the step (23), otherwise, entering the step (26);
(26) calculating an entropy value t of the current process image P, if the entropy value t is larger than the entropy value e, updating the entropy value e to the entropy value t, and updating the current result image R to be the process image P;
(27) if it is not
Figure RE-FDA0002337120010000045
Updating k to k +1, and returning to the step (22); otherwise, entering a step (28);
(28) the reference image B is updated to the result map R.
6. The system for correcting and optimizing the infrared image nonuniformity based on image entropy as claimed in claim 5, wherein in the third module, the nonuniformity correction based on linear approximation is performed on the other image C to the reference image B, and the specific method for obtaining the correction result is as follows:
(31) and calculating an image D to be approximated from the reference image B:
if the image C is an odd pixel image of the input image X, assigning the 1 st column of the image D as the 1 st column of the image B, and sequentially assigning the last n-1 columns of the image D as average images of two adjacent columns at corresponding positions of the image B; otherwise, if the image C is an even image element image of the input image X, assigning the first n-1 columns of the image D as average images of two adjacent columns at the corresponding position of the image B in sequence, and assigning the last column of the image D as the last column of the image B;
(32) traversing g from 1 to n, calculating a left-end matrix F and a right-end vector H:
Figure RE-FDA0002337120010000051
F(2g,2g)＝2m，
Figure RE-FDA0002337120010000052
Figure RE-FDA0002337120010000053
g＝1,2,3…n；j＝1,2,3…m；
f (x, y) is the value of the x row and the y column in the left matrix F;
c (x, y) is the value of the x row and the y column in the image C;
h (x) is the x-th element in the right-end vector H;
d (x, y) is the value of the x row and the y column in the image D;
(33) calculating a linear approximation parameter T ═ inv (F) H; inv (F) is the inversion of the matrix F;
(34) if the image C is an odd pixel image of the input image X, the even pixel gray scale of the image Y is set as the corresponding pixel gray scale of the reference image B, and the odd pixel of the image Y is updated from 1 to n by traversing g as follows:
y (h,2g-1) ═ T (2g-1) C (h, g) + T (2g), where h is 1,2,3,.., m;
t (2g-1) and T (2g) are elements of the 2g-1 st and the 2g nd in the matrix T respectively; y (x, Y) is the value of the x row and the Y column in the image Y; otherwise, if the image C is an odd pixel image of the input image X, the odd pixel gray scale of the image Y is set to the pixel gray scale corresponding to the reference image B, and the even pixel of the image Y is updated from 1 to n by traversing g as follows:
y (h,2g) ═ T (2g-1) C (h, g) + T (2g), where h is 1,2,3,. said, m;
image Y is the final corrected image.
CN201911065298.3A 2019-11-01 2019-11-01 Infrared image non-uniformity parameterization correction optimization method based on image entropy Active CN111047521B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911065298.3A CN111047521B (en) 2019-11-01 2019-11-01 Infrared image non-uniformity parameterization correction optimization method based on image entropy

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911065298.3A CN111047521B (en) 2019-11-01 2019-11-01 Infrared image non-uniformity parameterization correction optimization method based on image entropy

Publications (2)

Publication Number Publication Date
CN111047521A true CN111047521A (en) 2020-04-21
CN111047521B CN111047521B (en) 2023-08-11

Family

ID=70232719

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911065298.3A Active CN111047521B (en) 2019-11-01 2019-11-01 Infrared image non-uniformity parameterization correction optimization method based on image entropy

Country Status (1)

Country Link
CN (1) CN111047521B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112150377A (en) * 2020-08-27 2020-12-29 北京空间机电研究所 Infrared image non-uniformity correction coefficient analytical solution alternative iteration optimization method
CN113781368A (en) * 2021-11-12 2021-12-10 国科天成科技股份有限公司 Infrared imaging device based on local information entropy

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060110071A1 (en) * 2004-10-13 2006-05-25 Ong Sim H Method and system of entropy-based image registration
US20090303111A1 (en) * 2008-06-09 2009-12-10 Cho Kwang M Autofocus for minimum entry through multi-dimensional optimization
CN102081219A (en) * 2010-12-10 2011-06-01 中国空间技术研究院 In-orbit automatic focusing method for space CCD (charge coupled device) optical remote sensor
DE102011085927A1 (en) * 2011-11-08 2013-05-08 Siemens Aktiengesellschaft Method for correction of movement effects during reconstruction of image data set from two dimensional projection image of moving target area recorded with x-ray device, involves determining entropy of image data set by correction parameter
CN105957034A (en) * 2016-04-28 2016-09-21 武汉大学 Scanning infrared imaging system scene non-uniformity correction based on registration
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

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060110071A1 (en) * 2004-10-13 2006-05-25 Ong Sim H Method and system of entropy-based image registration
US20090303111A1 (en) * 2008-06-09 2009-12-10 Cho Kwang M Autofocus for minimum entry through multi-dimensional optimization
CN102081219A (en) * 2010-12-10 2011-06-01 中国空间技术研究院 In-orbit automatic focusing method for space CCD (charge coupled device) optical remote sensor
DE102011085927A1 (en) * 2011-11-08 2013-05-08 Siemens Aktiengesellschaft Method for correction of movement effects during reconstruction of image data set from two dimensional projection image of moving target area recorded with x-ray device, involves determining entropy of image data set by correction parameter
CN105957034A (en) * 2016-04-28 2016-09-21 武汉大学 Scanning infrared imaging system scene non-uniformity correction based on registration
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

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112150377A (en) * 2020-08-27 2020-12-29 北京空间机电研究所 Infrared image non-uniformity correction coefficient analytical solution alternative iteration optimization method
CN112150377B (en) * 2020-08-27 2024-03-29 北京空间机电研究所 Infrared image non-uniformity correction coefficient analysis solution alternation iteration optimization method
CN113781368A (en) * 2021-11-12 2021-12-10 国科天成科技股份有限公司 Infrared imaging device based on local information entropy
CN113781368B (en) * 2021-11-12 2022-01-21 国科天成科技股份有限公司 Infrared imaging device based on local information entropy

Also Published As

Publication number Publication date
CN111047521B (en) 2023-08-11

Similar Documents

Publication Publication Date Title
Abdelhamed et al. Noise flow: Noise modeling with conditional normalizing flows
CN100563303C (en) Image processing equipment, image capture device and image processing method
CN107590779B (en) Image denoising and deblurring method based on image block clustering dictionary training
JP2011512086A (en) Reduction of noise and / or flicker in video sequences using spatial and temporal processing
CN108230249B (en) Anisotropic-based L1 norm total variation regularization non-uniformity correction method
CN106197690B (en) Image calibrating method and system under the conditions of a kind of wide temperature range
CN110415198B (en) Medical image fusion method based on Laplacian pyramid and parameter self-adaptive pulse coupling neural network
CN111047521B (en) Infrared image non-uniformity parameterization correction optimization method based on image entropy
CN110956581A (en) Image modality conversion method based on dual-channel generation-fusion network
CN105915757B (en) A kind of color calibration method based on RGB color channel error function
CN104406699A (en) Infrared thermal imager based on adaptive infrared image correction algorithm
CN111080561A (en) Time domain high-pass filtering method
CN111652815B (en) Mask plate camera image restoration method based on deep learning
Lu Stripe non-uniformity correction of infrared images using parameter estimation
CN105279742B (en) A kind of image de-noising method quickly based on piecemeal estimation of noise energy
US8089534B2 (en) Multi illuminant shading correction using singular value decomposition
CN114463196A (en) Image correction method based on deep learning
CN103868601A (en) Bilateral total variation regularization correction method for non-uniform response of IRFPA detector
CN111798381A (en) Image conversion method, image conversion device, computer equipment and storage medium
US11145033B2 (en) Method and device for image correction
CN112348754B (en) Low-illumination color image enhancement method and device
CN112508807B (en) Image denoising method based on multi-direction total variation
CN111028159B (en) Image stripe noise suppression method and system
CN112150481B (en) Powdery mildew image segmentation method
CN113670440A (en) Compressed spectrum imaging method based on adaptive dictionary

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