CN116429272A - Blind pixel correction method suitable for infrared multispectral focal plane - Google Patents

Blind pixel correction method suitable for infrared multispectral focal plane Download PDF

Info

Publication number
CN116429272A
CN116429272A CN202310298787.3A CN202310298787A CN116429272A CN 116429272 A CN116429272 A CN 116429272A CN 202310298787 A CN202310298787 A CN 202310298787A CN 116429272 A CN116429272 A CN 116429272A
Authority
CN
China
Prior art keywords
blind
pixel
pixels
spectrum
response
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.)
Pending
Application number
CN202310298787.3A
Other languages
Chinese (zh)
Inventor
王憬诣
赵永强
王秉路
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.)
Northwestern Polytechnical University
Original Assignee
Northwestern Polytechnical University
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 Northwestern Polytechnical University filed Critical Northwestern Polytechnical University
Priority to CN202310298787.3A priority Critical patent/CN116429272A/en
Publication of CN116429272A publication Critical patent/CN116429272A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01J—MEASUREMENT OF INTENSITY, VELOCITY, SPECTRAL CONTENT, POLARISATION, PHASE OR PULSE CHARACTERISTICS OF INFRARED, VISIBLE OR ULTRAVIOLET LIGHT; COLORIMETRY; RADIATION PYROMETRY
    • G01J5/00—Radiation pyrometry, e.g. infrared or optical thermometry
    • G01J5/90—Testing, inspecting or checking operation of radiation pyrometers
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01J—MEASUREMENT OF INTENSITY, VELOCITY, SPECTRAL CONTENT, POLARISATION, PHASE OR PULSE CHARACTERISTICS OF INFRARED, VISIBLE OR ULTRAVIOLET LIGHT; COLORIMETRY; RADIATION PYROMETRY
    • G01J5/00—Radiation pyrometry, e.g. infrared or optical thermometry
    • G01J5/48—Thermography; Techniques using wholly visual means
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01J—MEASUREMENT OF INTENSITY, VELOCITY, SPECTRAL CONTENT, POLARISATION, PHASE OR PULSE CHARACTERISTICS OF INFRARED, VISIBLE OR ULTRAVIOLET LIGHT; COLORIMETRY; RADIATION PYROMETRY
    • G01J5/00—Radiation pyrometry, e.g. infrared or optical thermometry
    • G01J5/52—Radiation pyrometry, e.g. infrared or optical thermometry using comparison with reference sources, e.g. disappearing-filament pyrometer
    • G01J5/53—Reference sources, e.g. standard lamps; Black bodies
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Radiation Pyrometers (AREA)
  • Photometry And Measurement Of Optical Pulse Characteristics (AREA)
  • Transforming Light Signals Into Electric Signals (AREA)

Abstract

The invention discloses a blind pixel correction method suitable for an infrared multispectral focal plane, which comprises the steps of firstly separating each wave band contained in an original infrared multispectral image by using a mask, obtaining blind pixels by using a time domain, a space domain and a spectrum domain, realizing blind pixel detection, judging whether the blind pixels are in a single wave band or at a wave band junction by using a blind pixel positioning model, carrying out neighborhood compensation on the blind pixels in the single wave band, carrying out spatial spectrum joint compensation on the blind pixels at the wave band junction, and finally realizing blind pixel correction based on a focus-division plane infrared multispectral camera. The invention can effectively realize blind pixel detection and correction.

Description

Blind pixel correction method suitable for infrared multispectral focal plane
Technical Field
The invention belongs to the technical field of computer vision, and particularly relates to a blind pixel correction method suitable for an infrared multispectral focal plane.
Background
The pixels whose common infrared focal plane has lost detectability due to the limitations of the production materials and the defects of the process manufacturing, resulting in a response of a part of the pixels that deviates far from the response of the whole focal plane, are called blind pixels. GB/T17444-2013 defines pixels with response rates less than 50% of the average pixel response rate as dead blind pixels; the pixels with noise voltages greater than 2 times the average pixel noise voltage are overheat blind pixels.
The specific structure of the split focal plane type infrared multispectral camera (hereinafter referred to as an infrared multispectral camera) is that an optical filter array based on a grating structure is covered on the surface of an infrared detector, a basic mode of a single optical filter array comprises 9 wave bands, the single wave band covers 7×7 pixels on a sensor, and a mosaic image containing multispectral information can be obtained by a single snapshot. Besides the problem that the infrared multispectral camera has more serious blind pixels compared with the common infrared camera due to errors when the optical filter is processed and attached to the focal plane, the infrared multispectral focal plane has spectral characteristics, and the focal plane can generate additional blind pixels, namely spectral blind pixels. The existence of blind pixels not only affects the imaging quality, but also has serious influence on the spectrum calculation of the subsequent target object.
The blind pixel correction method comprises two parts of blind pixel detection and blind pixel compensation, is very mature for the infrared camera, is divided into a blind pixel detection method based on calibration and scenes, and is mainly used for detecting dead blind pixels and overheated blind pixels and cannot be suitable for detecting spectrum blind pixels of an infrared multispectral focal plane; the blind pixel compensation method for the infrared camera is usually carried out by replacing adjacent pixel elements, but because the infrared multispectral focal plane belongs to a heterogeneous plane and adjacent pixels of the infrared multispectral focal plane do not belong to the same wave band, the blind pixel compensation method for the infrared multispectral focal plane needs to be researched.
In summary, a blind pixel correction method is proposed for an infrared multispectral camera (focal plane), and the spectral characteristics of the infrared multispectral focal plane are combined, and corresponding blind pixel detection and blind pixel compensation methods are provided.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides a blind pixel correction method suitable for an infrared multispectral focal plane, which comprises the steps of firstly separating each wave band contained in an original infrared multispectral image by using a mask, obtaining blind pixels through a time domain, a space domain and a spectrum domain, realizing blind pixel detection, judging whether the blind pixels are in a single wave band or at a wave band junction by using a blind pixel positioning model, carrying out neighborhood compensation on the blind pixels in the single wave band, carrying out spatial spectrum joint compensation on the blind pixels at the wave band junction, and finally realizing blind pixel correction based on a focus-division plane infrared multispectral camera. The invention can effectively realize blind pixel detection and correction.
The technical scheme adopted by the invention for solving the technical problems comprises the following steps:
step 1: acquiring thermal radiation image sequences corresponding to black bodies with different temperatures by detecting the black bodies with different temperatures by using an infrared multispectral camera, separating each single wave band, and analyzing response characteristics of an infrared multispectral focal plane from time, space and spectrum dimensions according to the thermal radiation image sequences;
step 2: sequencing response values of single pixels in the thermal radiation image sequences with different wave bands and different temperatures, which are obtained in the step 1, along with time, removing extreme response values, and taking the rest average value as a standard response value at the temperature; the extreme response value is c pixels at the front part and the tail part of the image sequence;
step 3: performing blind pixel detection in a space dimension; acquiring pixel response values of different wave bands at each temperature by utilizing the step 2, and detecting and marking overheat blind pixels and dead blind pixels in different wave bands according to national standard definition;
step 4: performing blind pixel detection in spectrum dimension; using the median value of the pixel response values of different temperatures in different wave bands obtained in the step 2 as a standard pseudo-spectrum response curve of each wave band, and matching the pseudo-spectrum response curves of all pixels in each wave band with the standard pseudo-spectrum curve so as to detect and mark spectrum blind pixels;
step 5: establishing a positioning model aiming at the thermal blind pixels, the dead blind pixels and the spectrum blind pixels obtained in the step 3 and the step 4, judging whether the thermal blind pixels, the dead blind pixels and the spectrum blind pixels are in a single wave band or at the junction of wave bands according to the positioning model, and accumulating the space dimensions of the blind pixel tables of each wave band to obtain an infrared multispectral focal plane blind pixel table;
step 6: and (3) according to the blind pixel position obtained in the step (5), adopting an improved infrared blind pixel compensation method, and carrying out spatial spectrum joint compensation on the band junction by combining the spatial spectrum characteristics of the infrared multispectral focal plane, thereby completing the blind pixel correction of the whole infrared multispectral focal plane.
Further, the separating each single band in the step 1 specifically includes:
separating the bands of the original infrared multispectral phase with a mask, wherein I MSFA Is an infrared multispectral image, mask i The mask of the ith wave band, p is any pixel in the mask, and the masks of all wave bands are
Figure BDA0004144154770000021
The separation result corresponding to the ith wave band is I i ：
Figure BDA0004144154770000022
I i ＝I MSFA ⊙Mask i
Further, the step 2 specifically includes:
sorting pixel response values in a thermal radiation image sequence at the same temperature in each single wave band in a time domain, removing extreme values and taking an average;
Figure BDA0004144154770000031
wherein,,
Figure BDA0004144154770000032
for an unordered sequence of picture elements, the subscript represents the acquisition time from first to last,
Figure BDA0004144154770000033
for the ordered pixel sequence, the subscript represents the response value from small to large, c extreme values at two ends are removed, and the average value is taken as the response value at the temperature:
Figure BDA0004144154770000034
further, the step 3 specifically includes:
defining pixels with response rate less than 50% of the average pixel response rate as dead blind pixels according to national standard GB/T17444-2013; the pixels with the noise voltage being more than 2 times of the average pixel noise voltage are overheat blind pixels;
for the focal plane of I J, the pixel response for row I and column J is denoted as I (I, J), the average pixel response
Figure BDA0004144154770000035
The method comprises the following steps:
Figure BDA0004144154770000036
according to dead-blind pixel definition, a pixel conforming to the following inequality is a dead-blind pixel, and is marked as d:
Figure BDA0004144154770000037
after the dead blind pixels d are subtracted, the rest pixels participate in the operation to obtain the middle average noise voltage:
Figure BDA0004144154770000038
according to the definition of overheat blind pixels, the pixels conforming to the following inequality are overheat blind pixels, which are marked as h:
Figure BDA0004144154770000039
further, the specific method in the step 4 is as follows:
forming a pseudo spectrum curve by using standard response values of all pixels in an infrared multispectral focal plane at different temperatures, calculating gradients of all pseudo spectrum curves, and detecting and marking partial spectrum blind pixels by combining with the Planckian blackbody radiation law; and taking the median value of the response of each residual temperature pixel as a standard pseudo spectrum curve, matching the pseudo spectrum curves of other pixels with the standard pseudo spectrum curve by utilizing a spectrum gradient angle, setting a threshold value, and detecting and marking the residual spectrum blind pixels.
Definition of arbitrary pixelsp can form pseudo spectrum curve of the pixel at different temperature response values
Figure BDA00041441547700000310
Assuming that the blackbody radiation is uniform radiation, taking I i_Ts The median response of the standard temperature response is used as the standard response value of the ith wave band under the T temperature, so that the standard pseudo spectrum curve of the wave band can be formed by using the response values of the same wave band corresponding to different temperatures
Figure BDA0004144154770000041
C i_Ts ＝median(I i_Ts )
Calculating the spectral gradients of all pixels, and according to the Planckian blackbody radiation law, along with the rise of temperature, the pixel response value of the infrared multispectral focal plane also rises, so that when S GA (C i_T ) When the pixel is less than or equal to 0, the pixel is marked as a spectrum blind pixel; after removing the partial spectrum blind pixels, calculating a spectrum gradient angle of a pseudo spectrum curve and a standard pseudo spectrum curve of the pixel p; setting a threshold gamma, wherein the spectrum blind pixels are obtained after the threshold gamma is exceeded;
Figure BDA0004144154770000042
Figure BDA0004144154770000043
Figure BDA0004144154770000044
Figure BDA0004144154770000045
Figure BDA0004144154770000046
finally, an i-wave Duan Mang-element table is obtained through a blind element detection algorithm
Figure BDA0004144154770000047
Wherein->
Figure BDA0004144154770000048
Figure BDA0004144154770000049
Further, the positioning model in the step 5 specifically includes:
positioning the blind pixels through a blind pixel table, wherein the blind pixels in the wave band are
Figure BDA00041441547700000410
Blind pixel at band junction>
Figure BDA00041441547700000411
The positioning model is as follows:
Figure BDA00041441547700000412
Figure BDA00041441547700000413
further, the specific method in the step 6 is as follows:
firstly, carrying out neighborhood mean value compensation on the blind pixels in the single wave band detected in the step 5, and carrying out space spectrum joint compensation on the blind pixels at the joint of the wave bands; for blind pixels at the joint of the wave bands, firstly, spatial information compensation is carried out by using Euclidean distance as a spatial coefficient, then, spectral information compensation is carried out by using spectral difference as a spectral coefficient, and the two are combined to compensate the blind pixels.
Firstly, blind pixel compensation needs to add the space dimensions of the blind pixel table of each wave band to obtain an infrared-based multispectral mosaic imageObtaining a blind pixel table I m The method comprises the steps of carrying out a first treatment on the surface of the The blind pixels can be compensated by using adjacent pixel values of the blind pixels after the blind pixels are subjected to dot multiplication by using the blind pixel table and the real infrared multispectral image;
Figure BDA0004144154770000052
for blind pixels in wave bands
Figure BDA0004144154770000053
Infrared multispectral mosaic image processed by blind pixel table by combining Euclidean distance
Figure BDA0004144154770000054
And performing neighborhood compensation:
Figure BDA0004144154770000055
Figure BDA0004144154770000056
blind pixel I for band junction m_con And (3) performing spatial spectrum joint compensation:
Figure BDA0004144154770000057
Figure BDA0004144154770000058
the spectrum difference is used as a spectrum coefficient to act with a space coefficient together, so that the spectrum joint compensation of the blind pixels is realized.
Figure BDA0004144154770000059
△ n ＝|I m-con *H n |n∈{1,2,3}
Figure BDA00041441547700000510
Figure BDA00041441547700000511
Figure BDA0004144154770000061
Figure BDA0004144154770000062
Figure BDA0004144154770000063
The beneficial effects of the invention are as follows:
1. because the infrared multispectral focal plane has spectral characteristics, the traditional infrared blind pixel detection algorithm cannot be applied, the response characteristics of the infrared multispectral focal plane in time, space and spectrum are required to be analyzed, and the blind pixel detection method based on the infrared multispectral characteristics is obtained by combining the spectral characteristics, so that the spectral blind pixels contained in the infrared multispectral focal plane can be effectively detected.
2. Because the image acquired by the infrared multispectral focal plane is an infrared multispectral mosaic image, namely, a single image contains information of different wave bands. Therefore, the traditional blind pixel compensation method based on the infrared camera is not applicable any more. The method combines the spatial characteristics and the spectral characteristics of the infrared multispectral mosaic image, and provides a spatial spectrum joint compensation method aiming at blind pixels at the wave band junction.
Drawings
Fig. 1 is an imaging schematic diagram of an infrared multispectral camera.
Fig. 2 is a blackbody radiation diagram of an infrared multispectral camera.
Fig. 3 is a flow chart of the infrared multispectral focal plane blind pixel correction of the present invention.
FIG. 4 is a graph of the time domain response of an infrared multispectral focal plane pixel in accordance with an embodiment of the invention.
Fig. 5 is a graph of the response of an infrared multispectral focal plane based on planck's blackbody radiation law in an embodiment of the invention.
Fig. 6 is a view showing blind pixel positioning according to an embodiment of the present invention.
Detailed Description
The invention will be further described with reference to the drawings and examples.
The invention aims to provide a blind pixel correction method suitable for an infrared multispectral focal plane, so as to solve the problem of blind pixels of the infrared multispectral focal plane.
The invention adopts the following technical scheme: firstly, response characteristics of an infrared multispectral focal plane are analyzed from time, space and spectrum dimensions, and a blind pixel detection method based on the infrared multispectral characteristics is established. Meanwhile, the arrangement mode of the infrared multispectral focal plane is analyzed, and a spatial spectrum combined blind pixel compensation method based on the infrared multispectral mosaic image is established.
Firstly, an infrared multispectral camera acquires a corresponding thermal radiation image sequence by detecting blackbody with different temperatures, separates each single wave band, and analyzes response characteristics of an infrared multispectral focal plane from time, space and spectrum dimensions according to the thermal radiation image sequence;
step 2, sequencing response values of single pixels in the thermal radiation image sequences with different temperatures, which are obtained in the step 1, along with time, removing extreme response values, and taking an average value of the extreme response values as a standard response value at the temperature;
step 3: performing blind pixel detection in a space dimension; acquiring pixel response values of different wave bands at each temperature by utilizing the step 2, and detecting and marking overheat blind pixels and dead blind pixels in different wave bands according to national standard definition;
step 4: performing blind pixel detection in spectrum dimension; using the median value of the pixel response values of different temperatures in different wave bands obtained in the step 2 as a standard pseudo-spectrum response curve of each wave band, and matching the pseudo-spectrum response curves of all pixels in each wave band with the standard pseudo-spectrum curve so as to detect and mark spectrum blind pixels;
step 5, establishing a positioning model aiming at all types of blind pixels obtained in the step 3 and the step 4, and judging whether the blind pixels are in a single wave band or at a wave band junction according to the positioning model;
and 6, improving the blind pixel position obtained in the step 5 on the basis of the traditional infrared blind pixel compensation method, and performing spatial spectrum joint compensation on the band junction by combining the spatial spectrum characteristics of the infrared multispectral focal plane, so as to finish the blind pixel correction of the whole infrared multispectral focal plane.
The specific method of the step 4 is as follows: firstly, using response values corresponding to all pixels in an infrared multispectral focal plane in standard response diagrams of different temperatures to form a pseudo spectrum curve, calculating gradients of all the pseudo spectrum curves, and detecting and marking partial spectrum blind pixels by combining the Planckian blackbody radiation law. And taking the median value of the response of each residual temperature pixel as a standard pseudo spectrum curve, matching the pseudo spectrum curves of other pixels with the standard pseudo spectrum curve by utilizing a spectrum gradient angle, setting a threshold value, and detecting and marking the residual spectrum blind pixels.
The specific method of the step 6 is as follows: and (3) firstly, carrying out neighborhood mean value compensation on the blind pixels in the single wave band detected in the step (5), and carrying out space spectrum joint compensation on the blind pixels at the joint of the wave bands. For blind pixels at the joint of the wave bands, firstly, spatial information compensation is carried out by using Euclidean distance as a spatial coefficient, then, spectral information compensation is carried out by using spectral difference as a spectral coefficient, and the two are combined to compensate the blind pixels.
Specific examples:
the invention provides a blind pixel correction method suitable for an infrared multispectral focal plane, and the whole flow is shown in figure 3. Because the infrared multispectral focal plane has spectral characteristics, the traditional infrared blind pixel correction model is not accurate, the response characteristics of the infrared multispectral focal plane are required to be analyzed from time, space and spectral dimensions, and a blind pixel correction method based on the infrared multispectral focal plane is established by combining the traditional infrared blind pixel correction model.
The blind pixel correction algorithm of the infrared multispectral focal plane is divided into two parts, namely blind pixel detection and blind pixel compensation. The basic arrangement mode of the infrared multispectral focal plane is that each wave band contained in an original infrared multispectral image is separated by a mask, and then blind pixels are obtained through a time domain, a space domain and a spectrum domain, so that blind pixel detection is realized. Judging whether the blind pixels are in a single wave band or at the junction of the wave bands by using the blind pixel positioning model, carrying out neighborhood compensation on the blind pixels in the single wave band, and carrying out spatial spectrum joint compensation on the blind pixels at the junction of the wave bands, thereby finally realizing blind pixel correction based on the focal plane division infrared multispectral camera.
The method is implemented according to the following steps:
blind pixel detection:
step 1, firstly, obtaining a corresponding thermal radiation image sequence by using a blackbody radiation infrared multispectral focal plane, as shown in fig. 2. Since the infrared multispectral focal plane has the spectral characteristics of a plurality of wave bands, the thermal radiation image sequence needs to be subjected to wave band separation;
step 2, sequencing the pixel response values in the thermal radiation image sequence at the same temperature in each single wave band obtained in the step 1 in a time domain, removing extreme values, and taking the average value as a standard response value at the temperature;
step 3, obtaining standard response values at different temperatures by utilizing the step 2, and detecting and marking overheat blind pixels and dead blind pixels according to national standard definition;
step 4, as shown in fig. 4, the combination of response values of any pixel at different temperatures can be used as a pseudo spectrum curve of the pixel. And (3) taking the median value of the temperature response values of the different wave bands obtained in the step (3) as a standard pseudo spectrum response curve of each wave band. Then calculating the spectral gradient of other pixel pseudo-spectrum curves and the spectral gradient angle of the standard pseudo-spectrum curve, and setting the spectral blind pixel exceeding a threshold value;
the blind pixel detection is carried out by separating each single wave band by using a blackbody radiation infrared multispectral camera, and then detecting the blind pixels in three dimensions of time, space and spectrum respectively, and mainly aiming at the infrared multispectral characteristics.
Blind pixel compensation:
and step 1, positioning the detected blind pixels, and determining whether the blind pixels are in a single band or at a band junction according to a positioning model.
And 2, improving the blind pixel position obtained in the step 1 on the basis of the traditional infrared blind pixel compensation method, performing neighborhood mean value compensation on the blind pixels in a single band, and performing spatial spectrum joint compensation on the blind pixels at the joint of the bands. For blind pixels at the joint of the wave bands, firstly, the Euclidean distance is used as a space coefficient to carry out space information compensation, then, the spectrum difference is used as a spectrum coefficient to carry out spectrum information compensation, and the two are combined to carry out compensation on the blind pixels, so that the blind pixels of the whole infrared multispectral focal plane are corrected.
In the technical scheme of the invention, the blind pixel correction of the infrared multispectral focal plane is mainly divided into blind pixel detection and blind pixel compensation, and the whole flow is shown in figure 3. First, the mask is used to separate each band of original infrared multispectral phase, wherein I MSFA Is an infrared multispectral image, mask i The mask of the ith wave band, p is any pixel in the mask, and the masks of all wave bands are
Figure BDA0004144154770000091
The separation result corresponding to the ith wave band is I i 。
Figure BDA0004144154770000092
I i ＝I MSFA ⊙Mask i
And sequencing pixel response values in the thermal radiation image sequences at the same temperature in each single wave band in a time domain, removing extreme values and taking the average.
Figure BDA0004144154770000093
Wherein,,
Figure BDA0004144154770000094
for an unordered sequence of picture elements, the subscript represents the acquisition time from first to last,
Figure BDA0004144154770000095
for the ordered pixel sequence, the subscript represents the response value from small to large, c extreme values at two ends are removed, and the average value is taken as the standard response value at the temperature.
Figure BDA0004144154770000096
Defining pixels with response rate less than 50% of the average pixel response rate as dead blind pixels according to national standard GB/T17444-2013; the pixels with noise voltages greater than 2 times the average pixel noise voltage are overheat blind pixels. For the focal plane of I J, the pixel response for row I and column J is denoted as I (I, J), the average pixel response
Figure BDA0004144154770000097
The method comprises the following steps:
Figure BDA0004144154770000098
according to dead-blind pixel definition, a pixel conforming to the following inequality is a dead-blind pixel, and is marked as d:
Figure BDA0004144154770000099
after dead blind pixels d are deducted, the rest pixels participate in calculation to obtain intermediate average noise voltage
Figure BDA00041441547700000910
According to the definition of overheat blind pixels, the pixels conforming to the following inequality are overheat blind pixels, which are marked as h:
for the dead blind pixels and the overheat blind pixels detected in the single band
Figure BDA0004144154770000101
Defining pseudo-spectral curve of arbitrary picture element p which can be formed into said picture element at different temperature response values +.>
Figure BDA0004144154770000102
Assuming that blackbody radiation is uniform radiation, the response of the pixels is ideally uniform, but because the focal plane has certain non-uniformity due to the process structure, I is taken i_Ts The median response of the standard temperature response is taken as the standard response value of the ith band at T temperature. Then the response values corresponding to different temperatures in the same band can be used to form the standard pseudo spectrum curve of the band
Figure BDA0004144154770000103
C i_Ts ＝median(I i_Ts )
The spectral gradients of all the pixels are calculated, and as can be seen from fig. 5, the pixel response values of the infrared multispectral focal plane rise with the rise of temperature according to the planck blackbody radiation law. Thus when S GA (C i_T ) And when the pixel is less than or equal to 0, the pixel is marked as a spectrum blind pixel. And after removing the partial spectrum blind pixels, calculating the spectrum gradient angles of the pseudo spectrum curve and the standard pseudo spectrum curve of the pixel p. Setting a threshold gamma, and obtaining the spectrum blind pixels when the threshold gamma is exceeded.
Figure BDA0004144154770000104
Figure BDA0004144154770000105
Figure BDA0004144154770000106
Figure BDA0004144154770000107
Figure BDA0004144154770000108
Finally, an i-wave Duan Mang-element table is obtained through a blind element detection algorithm
Figure BDA0004144154770000109
Wherein->
Figure BDA00041441547700001010
Figure BDA00041441547700001011
Positioning the blind pixels through a blind pixel table, wherein the blind pixels in the wave band are
Figure BDA00041441547700001012
As shown in picture element a of fig. 6, blind pixels at band intersections +.>
Figure BDA00041441547700001013
As shown in fig. 6, pixel B. Wherein the positioning model is as follows:
Figure BDA00041441547700001014
Figure BDA00041441547700001015
firstly, blind pixel compensation needs to add the space dimensions of the blind pixel table of each wave band to obtain a blind pixel table I based on an infrared multispectral mosaic image m . And compensating the blind pixels by using adjacent pixel values of the blind pixels after the blind pixels are multiplied by the real infrared multispectral image points.
Figure BDA0004144154770000111
Figure BDA0004144154770000112
For blind pixels in wave bands
Figure BDA0004144154770000113
Infrared multispectral mosaic image processed by blind pixel table by combining Euclidean distance
Figure BDA0004144154770000114
And performing neighborhood compensation:
Figure BDA0004144154770000115
Figure BDA0004144154770000116
blind pixel I for band junction m_con And (3) performing spatial spectrum joint compensation:
Figure BDA0004144154770000117
Figure BDA0004144154770000118
the spectrum difference is used as a spectrum coefficient to act with a space coefficient together, so that the spectrum joint compensation of the blind pixels is realized.
Figure BDA0004144154770000119
△ n ＝|I m-con *H n |n∈{1,2,3}
Figure BDA00041441547700001110
Figure BDA00041441547700001111
Figure BDA00041441547700001112
Figure BDA0004144154770000121
Figure BDA0004144154770000122
And (3) blind pixel compensation of the infrared multispectral focal plane, and judging the position of the infrared multispectral focal plane through a blind pixel table. And performing spatial coefficient-based domain compensation for blind pixels in a single band, and performing spatial coefficient-based combined neighborhood compensation for blind pixels at the junction of the bands. By combining spectral characteristics, the blind pixel compensation problem based on the infrared multispectral focal plane is solved.
In a specific operation process, the invention firstly utilizes the infrared multispectral camera to detect blackbody with different temperatures to obtain corresponding radiation data, analyzes the response characteristics of the infrared multispectral focal plane from time, space and spectrum dimensions, and establishes a blind pixel detection model based on the infrared spectrum characteristics. And detecting blind pixels in the time domain, the space domain and the spectrum domain respectively, and generating a corresponding blind pixel table. And judging the position of the blind pixel according to the blind pixel table, and establishing a spatial spectrum combined blind pixel compensation method based on the infrared multispectral mosaic image. And completing blind pixel correction of the infrared multispectral camera through blind pixel detection and blind pixel compensation.
The infrared multispectral camera is adopted in the embodiment, and the imaging principle is shown in fig. 1. Firstly, the infrared multispectral camera is utilized to detect blackbody with different temperatures, so that radiation data of the blackbody with different temperatures are obtained, and the blackbody radiation principle is shown in fig. 2. Firstly, using blackbody radiation data with different temperatures to carry out band separation, and detecting blind pixels of an infrared focal plane from a time domain, a space domain and a spectrum domain. And generating a blind pixel table of each wave band according to the detection result, and judging the position of the blind pixel table according to a blind pixel positioning model, as shown in fig. 6. And accumulating blind pixels in different wave bands from the space dimension, and utilizing the effect of the blind pixel table and the infrared multispectral mosaic acquired by the real scene. And performing spatial coefficient-based domain compensation for blind pixels in a single band, and performing spatial coefficient-based combined neighborhood compensation for blind pixels at the junction of the bands. By utilizing the traditional blind pixel detection and blind pixel compensation method and combining spectral characteristics, the problem of blind pixel compensation of an infrared multispectral focal plane is solved.

Claims (7)

1. The blind pixel correction method suitable for the infrared multispectral focal plane is characterized by comprising the following steps of:
step 1: acquiring thermal radiation image sequences corresponding to black bodies with different temperatures by detecting the black bodies with different temperatures by using an infrared multispectral camera, separating each single wave band, and analyzing response characteristics of an infrared multispectral focal plane from time, space and spectrum dimensions according to the thermal radiation image sequences;
step 2: sequencing response values of single pixels in the thermal radiation image sequences with different wave bands and different temperatures, which are obtained in the step 1, along with time, removing extreme response values, and taking the rest average value as a standard response value at the temperature; the extreme response value is c pixels at the front part and the tail part of the image sequence;
step 3: performing blind pixel detection in a space dimension; acquiring pixel response values of different wave bands at each temperature by utilizing the step 2, and detecting and marking overheat blind pixels and dead blind pixels in different wave bands according to national standard definition;
step 4: performing blind pixel detection in spectrum dimension; using the median value of the pixel response values of different temperatures in different wave bands obtained in the step 2 as a standard pseudo-spectrum response curve of each wave band, and matching the pseudo-spectrum response curves of all pixels in each wave band with the standard pseudo-spectrum curve so as to detect and mark spectrum blind pixels;
step 5: establishing a positioning model aiming at the thermal blind pixels, the dead blind pixels and the spectrum blind pixels obtained in the step 3 and the step 4, judging whether the thermal blind pixels, the dead blind pixels and the spectrum blind pixels are in a single wave band or at the junction of wave bands according to the positioning model, and accumulating the space dimensions of the blind pixel tables of each wave band to obtain an infrared multispectral focal plane blind pixel table;
step 6: and (3) according to the blind pixel position obtained in the step (5), adopting an improved infrared blind pixel compensation method, and carrying out spatial spectrum joint compensation on the band junction by combining the spatial spectrum characteristics of the infrared multispectral focal plane, thereby completing the blind pixel correction of the whole infrared multispectral focal plane.
2. The blind pixel correction method for infrared multispectral focal planes according to claim 1, wherein the separating of each single band in step 1 is specifically:
separating the bands of the original infrared multispectral phase with a mask, wherein I MSFA Is an infrared multispectral image, mask i The mask of the ith wave band, p is any pixel in the mask, and the masks of all wave bands are
Figure FDA0004144154760000011
The separation result corresponding to the ith wave band is I i ：
Figure FDA0004144154760000012
I i ＝I MSFA ⊙Mask i 。
3. The blind pixel correction method for an infrared multispectral focal plane according to claim 2, wherein the step 2 specifically comprises:
sorting pixel response values in a thermal radiation image sequence at the same temperature in each single wave band in a time domain, removing extreme values and taking an average;
Figure FDA0004144154760000021
wherein,,
Figure FDA0004144154760000022
for an unordered sequence of picture elements, the subscript represents the acquisition time from first to last,
Figure FDA0004144154760000023
for the ordered pixel sequence, the subscript represents the response value from small to large, c extreme values at two ends are removed, and the average value is taken as the response value at the temperature:
Figure FDA0004144154760000024
4. a blind pixel correction method applicable to an infrared multispectral focal plane according to claim 3, wherein the step 3 specifically comprises:
defining pixels with response rate less than 50% of the average pixel response rate as dead blind pixels according to national standard GB/T17444-2013; the pixels with the noise voltage being more than 2 times of the average pixel noise voltage are overheat blind pixels;
for the focal plane of I J, the pixel response for row I and column J is denoted as I (I, J), the average pixel responseThe method comprises the following steps:
Figure FDA0004144154760000026
according to dead-blind pixel definition, a pixel conforming to the following inequality is a dead-blind pixel, and is marked as d:
Figure FDA0004144154760000027
after the dead blind pixels d are subtracted, the rest pixels participate in the operation to obtain the middle average noise voltage:
Figure FDA0004144154760000028
according to the definition of overheat blind pixels, the pixels conforming to the following inequality are overheat blind pixels, which are marked as h:
Figure FDA0004144154760000029
5. the blind pixel correction method applicable to an infrared multispectral focal plane according to claim 4, wherein the specific method of the step 4 is as follows:
forming a pseudo spectrum curve by using standard response values of all pixels in an infrared multispectral focal plane at different temperatures, calculating gradients of all pseudo spectrum curves, and detecting and marking partial spectrum blind pixels by combining with the Planckian blackbody radiation law; taking the median value of the response of each residual temperature pixel as a standard pseudo spectrum curve, matching the pseudo spectrum curves of other pixels with the standard pseudo spectrum curves by utilizing a spectrum gradient angle, setting a threshold value, and detecting and marking residual spectrum blind pixels;
defining pseudo-spectral curve of arbitrary pixel p which can be formed into the pixel at different temperature response values
Figure FDA0004144154760000031
Assuming that the blackbody radiation is uniform radiation, taking I i_Ts The median response of the standard temperature response is used as the standard response value of the ith wave band under the T temperature, so that the standard pseudo spectrum curve of the wave band can be formed by using the response values of the same wave band corresponding to different temperatures
Figure FDA0004144154760000032
C i_Ts ＝median(I i_Ts )
Calculating the spectral gradients of all pixels, and according to the Planckian blackbody radiation law, along with the rise of temperature, the pixel response value of the infrared multispectral focal plane also rises, so that when S GA (C i_T ) When the pixel is less than or equal to 0, the pixel is marked as a spectrum blind pixel; after removing the partial spectrum blind pixels, calculating a spectrum gradient angle of a pseudo spectrum curve and a standard pseudo spectrum curve of the pixel p; setting a threshold gamma, wherein the spectrum blind pixels are obtained after the threshold gamma is exceeded;
Figure FDA0004144154760000033
Figure FDA0004144154760000034
Figure FDA0004144154760000035
Figure FDA0004144154760000036
Figure FDA0004144154760000037
finally lead toObtaining an i-wave Duan Mang-element table through a blind element detection algorithm
Figure FDA0004144154760000038
Wherein->
Figure FDA0004144154760000039
Figure FDA00041441547600000310
6. The blind pixel correction method for an infrared multispectral focal plane according to claim 5, wherein the positioning model in step 5 specifically comprises:
positioning the blind pixels through a blind pixel table, wherein the blind pixels in the wave band are
Figure FDA00041441547600000311
Blind pixel at band junction>
Figure FDA00041441547600000312
The positioning model is as follows:
Figure FDA00041441547600000313
Figure FDA00041441547600000314
7. the blind pixel correction method for an infrared multispectral focal plane according to claim 6, wherein the specific method of step 6 is as follows:
firstly, carrying out neighborhood mean value compensation on the blind pixels in the single wave band detected in the step 5, and carrying out space spectrum joint compensation on the blind pixels at the joint of the wave bands; for blind pixels at the joint of the wave bands, firstly, performing spatial information compensation by using Euclidean distance as a spatial coefficient, and then performing spectral information compensation by using spectral difference as a spectral coefficient, wherein the two are combined to perform compensation on the blind pixels;
firstly, blind pixel compensation needs to add the space dimensions of the blind pixel table of each wave band to obtain a blind pixel table I based on an infrared multispectral mosaic image m The method comprises the steps of carrying out a first treatment on the surface of the The blind pixels can be compensated by using adjacent pixel values of the blind pixels after the blind pixels are subjected to dot multiplication by using the blind pixel table and the real infrared multispectral image;
Figure FDA0004144154760000041
Figure FDA0004144154760000042
for blind pixels in wave bands
Figure FDA0004144154760000043
Infrared multispectral mosaic image after being processed by a blind pixel table by combining Euclidean distance>
Figure FDA0004144154760000044
And performing neighborhood compensation:
Figure FDA0004144154760000045
Figure FDA0004144154760000046
blind pixel I for band junction m_con And (3) performing spatial spectrum joint compensation:
Figure FDA0004144154760000047
Figure FDA0004144154760000048
wherein, the spectrum difference is used as a spectrum coefficient to act with a space coefficient together, thereby realizing the spectrum joint compensation of the blind pixels;
Figure FDA0004144154760000049
△ n ＝|I m-con *H n |n∈{1,2,3}
Figure FDA0004144154760000051
Figure FDA0004144154760000052
Figure FDA0004144154760000053
Figure FDA0004144154760000054
Figure FDA0004144154760000055
CN202310298787.3A 2023-03-24 2023-03-24 Blind pixel correction method suitable for infrared multispectral focal plane Pending CN116429272A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310298787.3A CN116429272A (en) 2023-03-24 2023-03-24 Blind pixel correction method suitable for infrared multispectral focal plane

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310298787.3A CN116429272A (en) 2023-03-24 2023-03-24 Blind pixel correction method suitable for infrared multispectral focal plane

Publications (1)

Publication Number Publication Date
CN116429272A true CN116429272A (en) 2023-07-14

Family

ID=87090006

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310298787.3A Pending CN116429272A (en) 2023-03-24 2023-03-24 Blind pixel correction method suitable for infrared multispectral focal plane

Country Status (1)

Country Link
CN (1) CN116429272A (en)

Similar Documents

Publication Publication Date Title
AU2007345299B2 (en) System and method for adaptive non-uniformity compensation for a focal plane array
US9947086B2 (en) Image adjustment based on locally flat scenes
KR100315897B1 (en) Adaptive non-uniformity compensation algorithm
US9883178B2 (en) Method for measuring performance parameters and detecting bad pixels of an infrared focal plane array module
US10230912B2 (en) Fixed pattern noise mitigation for a thermal imaging system
US10467736B2 (en) Image adjustment based on locally flat scenes
CN110686783B (en) Blind pixel detection method and image processing method for InGaAs short-wave infrared imager
CN105890873A (en) Apparatus and method for blind pixel detection of thermal infrared hyperspectral imager
US7796153B1 (en) Equalization system and method for an imaging sensor
US20180315171A1 (en) Image adjustment based on locally flat scenes
JP2008092565A (en) Color matching method and image capturing device
US10867371B2 (en) Fixed pattern noise mitigation for a thermal imaging system
CN113379636A (en) Infrared image non-uniformity correction method, device, equipment and storage medium
CN116429272A (en) Blind pixel correction method suitable for infrared multispectral focal plane
Kumar Sensor non uniformity correction algorithms and its real time implementation for infrared focal plane array-based thermal imaging system
CN107633487A (en) A kind of system-level relative radiometric correction method of aviation sweep type multispectral scanner image
CN108470325B (en) Space-time three-dimensional noise identification and compensation method for area array staring infrared remote sensing image
CN109297604B (en) Method and system for acquiring two-point correction calibration parameters of thermal infrared imager
CN115002360A (en) Infrared video non-uniformity correction method based on robust estimation
Zhou et al. Local spatial correlation-based stripe non-uniformity correction algorithm for single infrared images
CN107664533A (en) A kind of three slit EO-1 hyperion moving target detection methods and device
JP7053434B2 (en) Image processing device and image processing method
Ratliff et al. Adaptive scene-based correction algorithm for removal of residual fixed pattern noise in microgrid image data
CN106780403B (en) A kind of push-broom type thermal infrared high-spectrum remote sensing asymmetric correction method
TWI450594B (en) Cross-color image processing systems and methods for sharpness enhancement

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