Self-adaptive infrared imaging system peripheral light quantity correction method based on guided filtering

By calibrating the "four-angle bright" and "four-angle dark" template images and using guide filters to process it, the problem of non-uniformity of the peripheral light quantity at large field of view is solved, and efficient light quantity correction and imaging quality improvement is achieved.

CN119071600APending Publication Date: 2024-12-03TIANJIN JINHANG INST OF TECH PHYSICS
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411268439.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-11
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

The "four-angle brightness" or "four-angle darkness" problems caused by the non-uniformity of peripheral light volume at large field of view are complex calculations and are prone to over-correction, which affects the imaging quality.

Method used

By calibrating the "four-corner bright" and "four-corner dark" template images in advance as the guide image, the template image is adaptively selected based on the mean difference of the low-frequency blurred image of the real-time image, and the real-time image is filtered using a guide filter to filter the local linear relationship. Make corrections.

Benefits of technology

Effectively eliminate the non-uniformity of peripheral light quantity in infrared imaging systems, maintain edge details, improve imaging quality, and avoid over-correction problems.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119071600A_ABST
    Figure CN119071600A_ABST
Patent Text Reader

Abstract

The invention relates to a self-adaptive infrared imaging system peripheral light quantity correction method based on guided filtering, and belongs to the technical field of infrared imaging preprocessing. According to the invention, for an infrared imaging system with a large field angle, template images of'bright at four corners' and'dark at four corners' of an optical system are taken as guide images, the template images can be adaptively selected through real-time calculation, and coefficients of residual non-uniformity in the real-time images are obtained through calculation of a guide filter; and non-uniformity of peripheral light quantity introduced by an optical system is eliminated.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of infrared imaging preprocessing, and relates to an adaptive peripheral light correction method for an infrared imaging system, and in particular to a correction method based on calibration guided filtering. Background Art

[0002] Infrared imaging non-uniformity is an inherent property of focal plane infrared imaging systems. It is initially caused by device non-uniformity due to defects in detector materials and processes. In the case of uniform infrared incident radiation, it manifests as inconsistent output response values ​​of each pixel of the infrared detector. In engineering, a multi-stage and multi-segment two-point correction algorithm based on blackbody calibration is often used. This method can solve device non-uniformity well in application scenarios working at room temperature. However, for optical systems with large field of view, when they work in high and low temperature environments, the difference in peripheral light of the infrared imaging system will produce residual non-uniformity. The larger the field of view, the worse the residual non-uniformity. The main source is the difference between the incident radiation temperature and the ambient temperature, which is specifically manifested as "bright at the four corners" or "dark at the four corners", resulting in a "pot lid" effect.

[0003] The non-uniformity introduced by the optical system needs to be corrected and the peripheral light needs to be eliminated to ensure the normal use of the imaging system. At present, in order to solve the problem of peripheral light non-uniformity, most infrared imaging systems use the method of extracting optical low-frequency components from real-time images, establishing N-order surface equations, fitting the image with the template to obtain its polynomial fitting coefficients, obtaining the optical low-frequency components in the image, and subtracting the background from the real-time image. This method has high computational complexity and will cause over-correction in scenes with normal peripheral light. Summary of the invention

[0004] 1. Technical issues to be resolved

[0005] The technical problem to be solved by the present invention is to design an adaptive peripheral light correction method for an infrared imaging system.

[0006] (II) Technical solution

[0007] In order to solve the above technical problems, the present invention provides a method for correcting peripheral light of an adaptive infrared imaging system based on guided filtering, comprising the following steps:

[0008] Step 1: by stimulating the environment change, two template images of "four corners bright" and "four corners dark" are calibrated in advance as candidate images for the guide image and stored;

[0009] Step 2: obtain a low-frequency blurred image of the real-time image to be corrected by processing, and adaptively select one of the two template images as a guide image according to the difference between the peripheral mean and the central mean of the low-frequency blurred image;

[0010] Step 3: Based on the idea that there is a local linear relationship between the guide image and the real-time image to be corrected, the calibrated template image is used as the guide image, and the guide filter is used to filter the real-time image;

[0011] Step 4: Correct the real-time image based on the filtering result obtained in step 3.

[0012] Preferably, step 1 is specifically as follows: operate the infrared imaging system at high temperature, collect images facing the low temperature uniform radiation surface, and obtain the "four corners bright" template image Guide_I light (x, y); Make the infrared imaging system work at low temperature, collect images facing the high temperature uniform radiation surface, and obtain the "four corners dark" template image Guide_I dark (x, y), and store these two sets of pre-calibrated template images.

[0013] Preferably, in step 2, a low-frequency blurred image of the real-time image to be corrected is obtained by filtering, and the peripheral mean of the low-frequency blurred image is calculated. If the peripheral mean is larger than the central area mean, the template image Guide_I is selected. light (x, y) is used as the guide image. If the calculated peripheral mean is less than or equal to the central area mean, the template image Guide_I is used. dark (x,y) is used as the guide image.

[0014] Preferably, in step 3, the feature parameter a is extracted from the real-time image by performing correlation operation between the real-time image and the guide image. k 、b k As the correction factor, it is as follows:

[0015] Assume that the real-time image to be corrected is P, the guide image is Guide_I, and Guide_I is Guide_I light (x,y) or Guide_I dark (x, y), the output image after the real-time image is processed by the guided filter is Q, then there is the following relationship:

[0016]

[0017] Among them, q i represents the gray value of the i-th pixel of the output image Q obtained after the guided filter operation; W i is the gray value of the i-th pixel of the processing kernel function of the P-guided filter, P i is the gray value of the i-th pixel of the real-time image P to be corrected;

[0018] Based on the idea that there is a local linear relationship between the guide image Guide_I and the output image Q, we have:

[0019] q i =a k Guide_I i +b k (2)

[0020] Among them, a k , b k is the correlation coefficient between the real-time image and the template image, k is the kth pixel point in the window of the guide filter, and the correlation coefficient a between the real-time image and the template image is calculated according to formulas (3) and (4): k , b k , formula (3) and (4) are based on the traditional guided filtering formula for a k Corrections were made:

[0021]

[0022] In the formula, is the window w that guides the filter k (If the window is a 3*3 window, the value range of k is 1 to 9) The grayscale average value of the guide image, var(Guide_I k )、var(q k ) represent the window w k The variance of the internal guide image and the variance of the real-time image to be corrected. The square root calculation represents the standard deviation of the product of the variance of the guide image and the variance of the real-time image to be corrected, cov(Guide_I k ,q k ) represents window w k The covariance of the two, ε is the regularization parameter; is the window w k The grayscale average value of the image to be corrected;

[0023] Based on a k , b k The new guided filtering result is obtained as shown in formula (5):

[0024]

[0025] Among them, w k The value is the radius of the window, and |w| is the window w k The number of inner pixels; They are window w k All a k , b k The mean of q i 'Get the output image Q.

[0026] Preferably, step 4 specifically includes: subtracting the image Q obtained by guided filtering from the real-time image P to obtain a corrected real-time image I: I=PQ.

[0027] The present invention also provides a system for implementing the method.

[0028] The invention also provides an infrared imaging non-uniformity correction method implemented based on the method.

[0029] The invention also provides an infrared imaging non-uniformity correction system implemented based on the method.

[0030] The invention also provides an infrared imaging preprocessing method implemented based on the method.

[0031] The invention also provides an infrared imaging preprocessing system implemented based on the method.

[0032] (III) Beneficial effects

[0033] The effect of the present invention is that for an infrared imaging system with a large field of view, by using the "four bright corners" and "four dark corners" template images of the optical system as guide images, the template images can be adaptively selected through real-time calculation, and the coefficient of residual non-uniformity in the real-time image can be obtained through calculation of the guide filter, thereby eliminating the peripheral light non-uniformity introduced by the optical system. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 is a flow chart of the method of the present invention;

[0035] Figure 2 The guiding filter correction principle and three-dimensional grayscale image of the present invention are shown, wherein a) is a guiding filter correction principle diagram, b) is a three-dimensional grayscale distribution diagram of the "four-corner bright" template I, c) is a three-dimensional grayscale image of the image p to be corrected, and d) is a three-dimensional grayscale image of the image corrected by the present method;

[0036] Figure 3 a) is the infrared scene 1 with “bright corners”, and b) is the image after adaptive correction of the present invention;

[0037] Figure 4 a) is the infrared scene 2 with "dark corners", and b) is the infrared image after adaptive correction using this method;

[0038] Figure 5 In the figure, a) is the infrared scene 3 with “normal four corners”, and b) is the infrared image after adaptive correction using the method. DETAILED DESCRIPTION

[0039] In order to make the purpose, content and advantages of the present invention more clear, the specific implementation methods of the present invention are further described in detail below in conjunction with the drawings and examples.

[0040] The object of the present invention is to provide a peripheral light correction method for an adaptive infrared imaging system based on guided filtering.

[0041] The technical solution for implementing the present invention is as follows:

[0042] (1) Using the change of the environment as an incentive, two template images (calibration templates) of "four-corner bright" and "four-corner dark" are calibrated in advance as candidate images for the guidance image and stored;

[0043] (2) obtaining a low-frequency blurred image of the real-time image to be corrected by filtering or other methods, and adaptively selecting one of the two template images as a guide image according to the difference between the mean value of the peripheral area and the mean value of the central area of ​​the low-frequency blurred image;

[0044] (3) Based on the idea that there is a local linear relationship between the guide image and the real-time image to be corrected, the calibrated template image is used as the guide image, and the real-time image is filtered using a guide filter;

[0045] (4) Correcting the real-time image based on the filtering result obtained in step (3) to reduce the non-uniformity of the peripheral light while maintaining the edge details as much as possible.

[0046] Reference below Figure 1 , the infrared detector fixed pattern noise correction method of the present invention is further described:

[0047] 1) Step 1, calibration:

[0048] The infrared imaging system works at high temperature and collects images facing the low-temperature uniform radiation surface to obtain the "four-corner bright" template image Guide_I light (x, y); The infrared imaging system works at low temperature and collects images facing the high temperature uniform radiation surface to obtain the "four corners dark" template image Guide_I dark (x, y), and store these two sets of pre-calibrated template images;

[0049] 2) Step 2, adaptively select the guide image:

[0050] Use filtering or other methods to obtain the low-frequency blurred image of the real-time image to be corrected, calculate the peripheral mean of the low-frequency blurred image, and if the peripheral mean is larger than the central area mean, select the template image Guide_I light (x, y) is used as the guide image. If the calculated peripheral mean is less than or equal to the central area mean, the template image Guide_I is used. dark (x,y) is used as the guide image.

[0051] 3) Step 3, guide filter calculation:

[0052] Through the correlation operation between the real-time image and the guided image, the characteristic parameter a is extracted from the real-time image. k , b k As the correction factor, it is as follows:

[0053] Assume that the real-time image to be corrected is P, and the guide image is Guide_I (Guide_I light (x,y) or Guide_I dark (x,y)), the output image after the real-time image is processed by the guided filter is Q, then there is the relationship shown in the following formula:

[0054]

[0055] Among them, q i represents the gray value of the i-th pixel of the output image Q obtained after the guided filter operation; W i is the gray value of the i-th pixel of the processing kernel function of the P-guided filter, P i is the gray value of the i-th pixel of the real-time image P to be corrected.

[0056] Based on the idea that there is a local linear relationship between the guide image Guide_I and the output image Q, we have the following formula (2):

[0057] q i =a k Guide_I i +b k (2)

[0058] Among them, a k , b k is the correlation coefficient between the real-time image and the template image (i.e., the linear transformation coefficient), k is the kth pixel point in the window of the guide filter, and the correlation coefficient a between the real-time image and the template image is calculated according to the following formulas (3) and (4): k , b k , formula (3) and (4) are based on the traditional guided filtering formula for a k Corrections were made:

[0059]

[0060] In the formula, is the window w that guides the filter k (If the window is a 3*3 window, the value range of k is 1 to 9) The grayscale average value of the guide image, var(Guide_I k )、var(qk ) represent the window w k The variance of the internal guide image and the variance of the real-time image to be corrected. The square root calculation represents the standard deviation of the product of the variance of the guide image and the variance of the real-time image to be corrected, cov(Guide_I k ,q k ) represents window w k The covariance of the two, ε is the regularization parameter; is the window w k The grayscale average value of the image to be corrected.

[0061] Based on a k 、b k The new guided filtering result is shown in the following equation (5):

[0062]

[0063] Among them, q i ' is the i-th pixel in the output image, w k For a window, w k The value is the radius of the window, and |w| is the window w k The number of pixels in the window, i is the pixel currently being processed, and k is the window w k The kth pixel in ; They are window w k All a k , b k The mean of i 'Get the output image Q;

[0064] 4) Step 4, calibration:

[0065] The image Q obtained by guided filtering is subtracted from the real-time image P to obtain the corrected real-time image I: I=PQ.

[0066] Guided filtering principle diagram and three-dimensional grayscale distribution diagram display:

[0067] A guided filter is used to correct the peripheral light of the real-time image. The correction principle is as follows: Figure 2 a). The three-dimensional grayscale distribution of the "four-corner bright" template is shown in Figure 2 b), the three-dimensional grayscale distribution diagram of the real-time image to be corrected is as follows Figure 2 c), the three-dimensional grayscale distribution diagram of the image corrected by the method of the present invention is as follows Figure 2 d).

[0068] Infrared image:

[0069] Figure 3a) is an infrared scene 1 with "bright corners", and b) is an image after adaptive correction of the present invention, and it can be seen that the situation of bright corners is improved; Figure 4 a) Infrared scene 2 with "dark corners". Figure b) is the infrared image after adaptive correction using this method. It can be seen that the dark corners have been improved. Figure 5 a) shows the infrared scene 3 with “normal four corners”, and b) shows the infrared image after adaptive correction using this method. It can be seen that this method does not over-correct the infrared image.

[0070] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for correcting peripheral light of an adaptive infrared imaging system based on guided filtering, characterized in that: The following steps are involved: Step 1: By stimulating the change of the environment, two template images of "four corners bright" and "four corners dark" are calibrated in advance as candidate images for the guide image and stored; Step 2: obtain a low-frequency blurred image of the real-time image to be corrected by processing, and adaptively select one of the two template images as a guide image according to the difference between the peripheral mean and the central mean of the low-frequency blurred image; Step 3: Based on the idea that there is a local linear relationship between the guide image and the real-time image to be corrected, the calibrated template image is used as the guide image, and the guide filter is used to filter the real-time image; Step 4: Correct the real-time image based on the filtering result obtained in step 3.

2. The method according to claim 1, characterized in that Step 1 is as follows: operate the infrared imaging system at high temperature, collect images facing the low temperature uniform radiation surface, and obtain the "four corners bright" template image Guide_I light (x, y); Make the infrared imaging system work at low temperature, collect images facing the high temperature uniform radiation surface, and obtain the "four corners dark" template image Guide_I dark (x, y), and store these two sets of pre-calibrated template images.

3. The method according to claim 2, characterized in that In step 2, a low-frequency blurred image of the real-time image to be corrected is obtained by filtering, and the peripheral mean of the low-frequency blurred image is calculated. If the peripheral mean is larger than the central area mean, the template image Guide_I is selected. light (x, y) is used as the guide image. If the calculated peripheral mean is less than or equal to the central area mean, the template image Guide_I is used. dark (x,y) is used as the guide image.

4. The method according to claim 3, characterized in that In step 3, the feature parameter a is extracted from the real-time image through correlation operation between the real-time image and the guide image. k 、b k As the correction factor, it is as follows: Assume that the real-time image to be corrected is P, the guide image is Guide_I, and Guide_I is Guide_I light (x,y) or Guide_I dark (x, y), the output image after the real-time image is processed by the guided filter is Q, then there is the following relationship: Among them, q i represents the gray value of the i-th pixel of the output image Q obtained after the guided filter operation; W i is the gray value of the i-th pixel of the processing kernel function of the P-guided filter, P i is the gray value of the i-th pixel of the real-time image P to be corrected; Based on the idea that there is a local linear relationship between the guide image Guide_I and the output image Q, we have: q i =a k Guide_I i +b k (2) Among them, a k , b k is the correlation coefficient between the real-time image and the template image, k is the kth pixel point in the window of the guide filter, and the correlation coefficient a between the real-time image and the template image is calculated according to formulas (3) and (4): k , b k , formula (3) and (4) are based on the traditional guided filtering formula for a k Corrections were made: In the formula, is the window w that guides the filter k (If the window is a 3*3 window, the value range of k is 1 to 9) The grayscale average value of the guide image, var(Guide_I k )、var(q k ) represent the window w k The variance of the internal guide image and the variance of the real-time image to be corrected. The square root calculation represents the standard deviation of the product of the variance of the guide image and the variance of the real-time image to be corrected, cov(Guide_I k ,q k ) represents window w k The covariance of the two, ε is the regularization parameter; is the window w k The grayscale average value of the image to be corrected; Based on a k , b k The new guided filtering result is obtained as shown in formula (5): Among them, w k The value is the radius of the window, and |w| is the window w k The number of pixels inside; a k , b k They are window w k All a k , b k The mean of i 'Get the output image Q.

5. The method according to claim 4, characterized in that Step 4 is specifically as follows: subtract the image Q obtained by guided filtering from the real-time image P to obtain a corrected real-time image I: I=PQ.

6. A system for implementing the method according to any one of claims 1 to 5.

7. A method for correcting infrared imaging non-uniformity based on the method according to any one of claims 1 to 5.

8. An infrared imaging non-uniformity correction system implemented based on the method as claimed in claim 7.

9. An infrared imaging preprocessing method implemented based on the method as claimed in any one of claims 1 to 5.

10. An infrared imaging preprocessing system implemented based on the method according to claim 9.