An infrared image processing system

Abstract

The present invention discloses an infrared image processing system, comprising the following steps: Step S1: Determine the calibration temperature points and the judgment threshold r for segmented correction based on the temperature-dependent curve of the detector's blackbody response at various integration times; Step S2: Calculate the gain coefficients k1 and k2 for two-point correction in the low-temperature and high-temperature segments, respectively, using an annular segmentation method; Step S3: Perform multi-point correction based on the judgment threshold r; Step S4: Compensate for blind pixels using the average of the compensation value of the blind pixel in the previous frame and the response value of the neighboring pixel in the current frame; Step S5: Perform scene-based blind pixel detection for the baffle scene; Step S6: Perform channel compensation; and Step S7: Enhance infrared image detail using a DDE algorithm. The present invention calculates the non-uniformity correction gain coefficients using an annular segmentation method, significantly improving the correction effect compared to traditional methods, especially for the central region of the detector at high temperatures.

Description

Technical Field

[0001] The present invention belongs to the technical field of infrared image processing, and in particular relates to an infrared image processing system. Background Art

[0002] Due to factors such as the uneven responsivity of individual pixel elements in infrared focal plane detectors and variations in readout circuit channels, non-uniformity is a common problem in infrared imaging systems. Currently, there are two main algorithms commonly used to correct for non-uniformity. One is a calibration-based correction algorithm. To improve the scene adaptability of multi-point correction, traditional calibration methods often require multi-point calibration with three or more stages, which consumes excessive flash storage space and increases hardware costs. The other is a scene-based correction algorithm, which is computationally intensive and has a low correction rate, making it unsuitable for hardware systems.

[0003] In addition to the non-uniformity problem, there are also a certain number of blind pixels in the infrared focal plane array, which will cause fixed black and white point noise in the imaging, so it needs to be compensated. The current common method is to replace it with the median filtering method, which is not suitable for processing cluster blind pixels.

[0004] Aside from heat sources like the sun, which can have high temperatures, the temperature differences in a scene are usually small. This results in a narrow histogram, low contrast, and unclear details in infrared images. Currently, common detail enhancement methods include linear mapping, histogram equalization, and DDE technology. Summary of the Invention

[0005] The object of the present invention is to provide an infrared image processing system to solve the problem raised in the background art that the two-stage multi-point calibration method used in the prior art to calculate correction parameters cannot effectively correct rich scenes.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0007] An infrared image processing system comprises the following steps:

[0008] Step S1, according to the curve of the detector's response to the black body as a function of temperature at each integration time, determine the calibration temperature point and the judgment threshold r for the segmented calibration;

[0009] Step S2, using a ring-shaped regional method to calculate the gain coefficients k1 and k2 of the two-point correction of the low temperature section and the high temperature section respectively; using a single-point temperature calibration method to obtain the bias coefficients b1 and b2, and matching the bias coefficients b1 and b2 with the gain coefficients k1 and k2 and saving them;

[0010] Step S3, performing multi-point calibration according to the judgment threshold r, performing blind pixel calibration on the slope of the blackbody temperature curve, the pixel response amplitude at a fixed integration time and fixed temperature, and the detector response standard deviation according to the pixel response at different integration times, and saving the calibration results;

[0011] Step S4, compensating the blind pixel with the average of the compensation value of the previous frame of the blind pixel and the response value of the neighboring pixel in the current frame, closing the camera baffle, performing single-point calibration, and correcting the bias coefficient b;

[0012] Step S5: Perform scene-based blind pixel detection on the baffle scene, update the blind pixel list, and compensate for new blind pixels;

[0013] Step S6, performing channel compensation;

[0014] Step S7: Use the DDE algorithm to enhance the details of the infrared image, increase the contrast, and optimize the visual effect.

[0015] According to the above technical solution, in step S1, the statistical analysis of the detector's response to black bodies of different temperatures under different integration time conditions is specifically implemented as follows:

[0016] Step S101, when plotting the curve of the detector's response to a blackbody versus temperature, a corner-cutting method is used to calculate the mean pixel response of the infrared focal plane as the ideal detector response to the blackbody at the current temperature: when calculating the mean detector response, only a circular area with a diameter of min(H, W) centered at the detector center is considered, where H and W represent the row height and column width of the focal plane array, respectively;

[0017] Step S102, selection of calibration temperature points: divide the response curve into low temperature section and high temperature section according to the gradient change of the response curve, make each section approximate to a linear function, select t1 and t2 as the two calibration points of the low temperature section; select t 3， t4 is used as the calibration temperature of the two points in the high temperature section, and the middle temperature of each section is recorded as t 11 ,t 12 As a single point calibration point, use the average of the two intermediate temperatures As the criterion for selecting the temperature interval Tt, the mean value of the detector response under Tt is used as the segmentation threshold r. When performing segmented correction, the correction coefficient is determined by comparing the mean value of the pixel response with the threshold.

[0018] According to the above technical solution, in step S1, the statistical analysis of the detector's response to black bodies of different temperatures under different integration time conditions is specifically implemented as follows:

[0019] Step S101, when plotting the curve of the detector's response to a blackbody versus temperature, a corner-cutting method is used to calculate the mean pixel response of the infrared focal plane as the ideal detector response to the blackbody at the current temperature: when calculating the mean detector response, only a circular area with a diameter of min(H, W) centered at the detector center is considered, where H and W represent the row height and column width of the focal plane array, respectively;

[0020] Step S102, selection of calibration temperature points: divide the response curve into low temperature section and high temperature section according to the gradient change of the response curve, make each section approximate to a linear function, select t1 and t2 as the two calibration points of the low temperature section; select t 3， t4 is used as the calibration temperature of the two points in the high temperature section, and the middle temperature of each section is recorded as t 11 ,t 12 As a single point calibration point, use the average of the two intermediate temperatures As the criterion for selecting the temperature interval Tt, the mean value of the detector response under Tt is used as the segmentation threshold r. When performing segmented correction, the correction coefficient is determined by comparing the mean value of the pixel response with the threshold.

[0021] According to the above technical solution, the ring partitioning method is used to calculate the gain coefficient in step S2, and the intermediate temperature t 11 , t 12 Perform a single-point calibration to calculate the bias coefficient:

[0022] Step S201, calculate the gain coefficient according to formula (1):

[0023]

[0024] In formula (1), Respectively represent t H , t L Radiative flux at temperature; represents the mean value of the original response output of the detector pixel in the time domain; Y H 、Y L Respectively represent t H , t L Ideal response value at temperature, k i,j Represents the gain coefficient; where, when obtaining Y H 、Y L When , the mean value of the annular area where the pixel is located is used for calculation according to the characteristic that the detector response gradually decays with distance from the center to the outside.

[0025] Step S202: Based on the gain coefficient obtained in step S201, perform gain correction in a segmented manner. 11 The bias coefficient is calculated according to formula (2) at temperature:

[0026]

[0027] In formula (2), Y represents the ideal pixel response output value; Represents the original response output value of the detector pixel; b i,j Represents the bias coefficient.

[0028] According to the above technical solution, in step S3, blind pixel calibration is performed based on the slope of the curve of the detector's response value to the black body versus temperature, the standard deviation of the pixel response at a fixed integration time and fixed temperature, and the response amplitude. The implementation process is as follows:

[0029] Step S301, by judging the conditions , calibrate blind pixels; among them, represents the mean value of the gain coefficient in the spatial domain, Represents the gain coefficient of each pixel;

[0030] Step S302: satisfy the condition The pixels of are marked as blind pixels, where Represents the gradient of the pixel response curve;

[0031] Step S303: If the grayscale value of the pixel response has a change value of more than 50 between the image of more than 50 frames within 100 frames and the image of the previous frame, the pixel is determined to be a blind pixel.

[0032] According to the above technical solution, in step S4, the blind pixel response value is replaced by the average of the blind pixel compensation value of the previous frame and the neighborhood pixel response value of the current frame to perform blind pixel compensation, and then single-point correction is performed based on the scene to correct the bias coefficient.

[0033] According to the above technical solution, in step S5, a variance-based blind pixel detection algorithm is used: the overall response is normalized after removing the background component, the degree of deviation between the pixel response and the average response is calculated, and the degree of deviation between the pixel response and the average response is compared with the overall mean square error to determine the blind pixel.

[0034] According to the above technical solution, in step S6, channel compensation is performed by correcting the channel mean to further remove the stripe noise: the channel compensation value is obtained according to formula (3): , according to the channel where the pixel is located, the corresponding Add to the original response value to perform channel compensation;

[0035]

[0036] In formula (3), P i represents the mean value in each channel, P avg Represents the overall detector response mean.

[0037] According to the above technical solution, the DDE algorithm is used to enhance the image in step S7. The algorithm is implemented as follows:

[0038] Step S701, determining the grayscale threshold by formula (4) to divide the infrared image into dark part and highlight part;

[0039]

[0040] In formula (4), pixMed is the grayscale median, pixAvg is the grayscale mean, and MM is the threshold shift coefficient;

[0041] Step S702 : performing histogram equalization on the dark part sub-image and the bright part sub-image respectively and then splicing them together to obtain an enhanced image.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] The present invention calculates the non-uniform correction gain coefficient by annularly dividing the regions. Compared with the traditional method, the correction effect is significantly improved, especially for the central region of the detector at high temperature.

[0044] The present invention performs blind pixel calibration based on the slope of the temperature change curve of the detector response and the response standard deviation and amplitude at a fixed integration time and fixed temperature. Combined with a scene-based blind pixel detection method, it can calibrate blind pixels more comprehensively without over-detection.

[0045] The present invention performs detail enhancement through the DDE algorithm, which can better improve the contrast of the image than the traditional single histogram equalization. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 This is a flow chart of the system of the present invention. DETAILED DESCRIPTION

[0047] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0048] It should be understood that the essence of the multi-point temperature calibration method is the segmented two-point calibration method. The multi-point temperature calibration used in the present invention is actually a combination of two segments of two-point calibration. For the convenience of description, the terms "two-point calibration" or "two-point correction" that may appear in the present invention are only for the convenience of describing the implementation of multi-point calibration and do not represent the method used in this example.

[0049] Example 1

[0050] like Figure 1 As shown, an infrared image processing system includes the following steps:

[0051] Step S1, according to the curve of the detector's response to the black body as a function of temperature at each integration time, determine the calibration temperature point and the judgment threshold r for the segmented calibration;

[0052] Step S2, using a ring-shaped regional method to calculate the gain coefficients k1 and k2 of the two-point correction of the low temperature section and the high temperature section respectively; using a single-point temperature calibration method to obtain the bias coefficients b1 and b2, and matching the bias coefficients b1 and b2 with the gain coefficients k1 and k2 and saving them;

[0053] Step S3, performing multi-point calibration according to the judgment threshold r, performing blind pixel calibration on the slope of the blackbody temperature curve, the pixel response amplitude at a fixed integration time and fixed temperature, and the detector response standard deviation according to the pixel response at different integration times, and saving the calibration results;

[0054] Step S4, compensating the blind pixel with the average of the compensation value of the previous frame of the blind pixel and the response value of the neighboring pixel in the current frame, closing the camera baffle, performing single-point calibration, and correcting the bias coefficient b;

[0055] Step S5: Perform scene-based blind pixel detection on the baffle scene, update the blind pixel list, and compensate for new blind pixels;

[0056] Step S6, performing channel compensation;

[0057] Step S7: Use the DDE algorithm to enhance the details of the infrared image, increase the contrast, and optimize the visual effect.

[0058] The present invention calculates the non-uniform correction gain coefficient by annularly dividing the regions. Compared with the traditional method, the correction effect is significantly improved, especially for the central region of the detector at high temperature.

[0059] The present invention performs blind pixel calibration based on the slope of the temperature change curve of the detector response and the response standard deviation and amplitude at a fixed integration time and fixed temperature. Combined with a scene-based blind pixel detection method, it can calibrate blind pixels more comprehensively without over-detection.

[0060] The present invention performs detail enhancement through the DDE algorithm, which can better improve the contrast of the image than the traditional single histogram equalization.

[0061] Example 2

[0062] This embodiment is a further refinement of the first embodiment. In step S1, the statistical analysis of the detector's response to black bodies of different temperatures under different integration time conditions is specifically implemented as follows:

[0063] Step 101: When plotting a curve of the detector's response to a blackbody versus temperature, a corner-cutting method is used to calculate the average pixel response of the infrared focal plane as the ideal detector response to the blackbody at the current temperature. The calculation of the average detector response only considers a circular area centered at the detector center and having a diameter of min(H, W), where H and W represent the row height and column width of the focal plane array, respectively.

[0064] Step 102, selection of calibration temperature points: divide the response curve into low temperature section and high temperature section according to its gradient change, make each section approximate to a linear function, select t1 and t2 as the two calibration points of the low temperature section; select t 3， t4 is used as the calibration temperature of the two points in the high temperature section, and the middle temperature of each section is recorded as t 11 ,t 12 As a single point calibration point, use the average of the two intermediate temperatures As the criterion for selecting the temperature interval Tt, the mean value of the detector response under Tt is used as the segmentation threshold r. When performing segmented correction, the correction coefficient is determined by comparing the mean value of the pixel response with the threshold.

[0065] In step S2, the bias coefficient is calculated by using a ring partition method and the intermediate temperature t 11 , t 12 Perform a single-point calibration to calculate the bias coefficients. Taking the low temperature section as an example, the operation process is as follows:

[0066] Step S201, calculate the gain coefficient according to formula (1):

[0067]

[0068] In formula (1), Respectively represent t H , t L Radiative flux at temperature; represents the mean value of the original response output of the detector pixel in the time domain; Y H 、Y L Respectively represent t H , t L Ideal response value at temperature, k i,j Represents the gain coefficient; when calculating the gain coefficient, compared with the traditional two-point calibration method that uses the response of the entire infrared detector array to replace the ideal response value Y, the present invention innovatively proposes to use the mean value of the annular area where the pixel is located as the ideal pixel response value Y for calculation

[0069] Step S202: Based on the gain coefficient obtained in step 201, perform gain correction in a segmented manner. 11 According to formula (2), the bias coefficient is calculated at temperature:

[0070]

[0071] In formula (2), Y represents the ideal pixel response output value; Represents the original response output value of the detector pixel; b i,j Represents the bias coefficient.

[0072] In order to correct the entire image to the same response value, the calculation of Y in the formula is performed by calculating the average response value of the entire image. i,j Combined with the gain coefficient obtained in step 201, a complete two-point correction parameter is obtained.

[0073] In step S3, blind pixel calibration is performed based on the slope of the curve of the detector's response to the blackbody versus temperature, the standard deviation of the pixel response at a fixed integration time and fixed temperature, and the response amplitude. The implementation process is as follows:

[0074] Step S301: According to the characteristic that the response values ​​of dead pixels and overheated pixels do not increase with the increase of temperature, the following conditions are used to determine the response values ​​of dead pixels and overheated pixels: , calibrate blind pixels; among them, represents the mean value of the gain coefficient in the spatial domain, Represents the gain coefficient of each pixel;

[0075] Step S302: According to the characteristic that the gradient of the blind pixel response curve with temperature change seriously deviates from the average gradient of the overall pixel response curve, the condition The pixels of are marked as blind pixels, where Represents the gradient of the pixel response curve;

[0076] Step S303: If the grayscale value of the pixel response has a change value of more than 50 between the image of more than 50 frames within 100 frames and the image of the previous frame, the pixel is determined to be a blind pixel.

[0077] In step S4, the blind pixel response value is replaced by the average of the blind pixel compensation value of the previous frame and the neighborhood pixel response value of the current frame to perform blind pixel compensation, and then single-point correction is performed based on the scene to correct the bias coefficient.

[0078] In step S5, to address the problem of new blind pixels generated after long-term use after calibration, a variance-based blind pixel detection algorithm is adopted: according to the characteristic that the blind pixel response deviates seriously from the average response of the detector, the overall response is normalized by removing the background component, and the degree of deviation between the pixel response and the average response is calculated, and compared with the overall mean square error to determine the blind pixel.

[0079] In step S6, channel compensation is performed by correcting the channel mean to further remove the stripe noise. The actual sampling of the detector is realized through multiple channels. Due to the differences in A / D devices, the collected responses show periodic differences. Statistics show that the pixel response mean of each channel of the infrared focal plane is consistent, so the channel compensation value can be obtained according to formula (3) , according to the channel where the pixel is located, the corresponding Add to the original response value to perform channel compensation;

[0080]

[0081] In formula (3), P i represents the mean value in each channel, P avg Represents the overall detector response mean.

[0082] In step S7, the DDE algorithm is used to enhance the image. The algorithm is implemented as follows:

[0083] Step S701, determining the grayscale threshold by formula (4) to divide the infrared image into dark part and highlight part;

[0084]

[0085] In formula (4), pixMed is the grayscale median, pixAvg is the grayscale mean, and MM is the threshold shift coefficient;

[0086] Step S702 : performing histogram equalization on the dark part sub-image and the bright part sub-image respectively and then splicing them together to obtain an enhanced image.

[0087] Example 3

[0088] The technical concept of the present invention is:

[0089] Step 1. Based on the detector's response to blackbodies at different temperatures at different integration times, plot the detector response versus temperature. In this example, the available integration times are 1.5ms, 2.0ms, 2.5ms, 3.0ms, 3.5ms, 4.0ms, 4.5ms, and 5.0ms, the blackbody temperature range is [-50°C, 50°C], and the sampling interval is 5°C. Because the detector pixel response gradually decays with distance from the center, to more accurately reflect the detector's response to the current temperature, this example only considers a circular region centered at the detector center and with a diameter of min(H, W) (where H and W represent the row height and column width of the focal plane array, respectively).

[0090] According to the gradient of the detector response curve to temperature, the response curve is divided into low temperature section and high temperature section, so that each section is approximately a linear function. The middle temperature of each section (respectively denoted as t 11 ,t 12 ) to calculate the bias correction coefficient in this temperature range, using the average of the two intermediate temperatures As the criterion for temperature interval selection (Tt), the mean response value under Tt is used as the segmentation threshold r when performing segmentation correction.

[0091] Step 2: Calculate the gain correction coefficient within the segment using the two-point calibration method in an annular area. The traditional two-point temperature calibration algorithm assumes that all pixels in the detector respond exactly the same at the same temperature. Therefore, according to the formula When calculating the gain coefficient, the average response value of all detector pixels is used for each pixel, replacing the ideal response value for the current pixel at the current temperature. However, analysis of actual detector responses shows that the central region of the detector has a stronger response, and the response intensity gradually decreases with the distance from the pixel to the center. Therefore, at high temperatures, the central region will reach saturation earlier, and the gradient of the temperature response curve between the central region and the boundary regions will deviate significantly. Using traditional methods to calculate two-point parameters will not effectively remove streak noise from the entire image.

[0092] Based on the above detector response characteristics, the present invention innovatively proposes a method for calculating correction parameters in annular regions: when calculating the current pixel gain coefficient, the average pixel response of the annular region where the pixel is located is used as the ideal response value. The definition of the annular region is based on the following: with the image center as the center of the circle, the radius gradually increases from 1, and all pixels are reordered in a concentric circle. The annular region where the current pixel is located is the radius of The set of all concentric circles (where r represents the radius of the concentric circle where the current pixel is located, Used to adjust the ring width, in this example 17 is used).

[0093] Step 3: Perform gain correction in sections using the hysteresis method. ( is the mean pixel response calculated by the corner removal method, Represents the hysteresis coefficient) to determine the correction parameters used by the current pixel to perform correction and obtain the grayscale value y1 after two-point correction. Can be avoided Frequent switching of correction parameters near r causes system abnormality. The value is 20.

[0094] Step 4: Calculate the bias correction parameters. (Y represents the ideal pixel response output value, which is replaced by the overall response mean of the detector; X I,j Represents the original response output value of the detector pixel) calculate the bias coefficient b i,j Match the calculated bias coefficient with the corresponding gain coefficient and save them.

[0095] Step 5: Pre-calibration of blind pixels. According to the characteristics that the blind pixel response (flickering pixels are not considered here) does not change with temperature, the judgment condition (in Represents the mean value of the gain coefficient in the spatial domain, k i,j Represents the gain coefficient of each pixel) to calibrate the blind pixel.

[0096] According to the characteristic that the gradient of the blind pixel response curve changes with temperature seriously deviates from the average gradient of the overall pixel response curve, the condition will be met. (in Pixels with the pixel response curve gradient are calibrated as blind pixels.

[0097] Based on the characteristic that the response of flickering pixels changes frequently and significantly between frames, if the grayscale value of the pixel response changes by more than 50 from the previous frame in more than 50 frames within 100 frames, the pixel is judged to be a blind pixel.

[0098] Step 6: Blind pixel compensation: The blind pixel response value is replaced by the average of the blind pixel compensation value of the previous frame and the neighboring pixel response value of the current frame to perform blind pixel compensation.

[0099] Step 7: Correct the bias coefficient. Close the camera shutter and perform single-point calibration to correct the bias coefficient.

[0100] Step 8: Blind pixel detection. Based on the characteristic that blind pixel responses deviate significantly from the detector's average response, the overall response is normalized by removing background components. The degree of deviation between the pixel response and the average response is calculated and compared with the overall mean square error to determine the blind pixel. The blind pixel list is then updated.

[0101] Step 9, channel compensation. In this embodiment, the actual sampling of the detector is realized through 4 channels. Due to the differences of A / D devices, the infrared image will show periodic stripe noise, so this fixed additive noise is corrected by adding bias to the pixels in the channel. Statistics show that the mean value of the pixel response of each channel of the infrared focal plane is consistent, so (P i Represents the mean value within the four channels, P avg Represents the overall response mean of the detector) to obtain the correction compensation value of the 4 channels .

[0102] Step 10: In actual application scenarios, the temperature difference of the scene is usually small, resulting in a relatively concentrated grayscale distribution in the image, lack of prominent details, and poor visual effects. This embodiment uses the DDE algorithm to enhance the image details.

[0103] according to (pixMed is the grayscale median, pixAvg is the grayscale mean, and MM is the threshold offset coefficient. Given the relatively low grayscale values ​​in normal temperature scenes, MM is set to 0.49 in this example.) Calculate the segmentation threshold PixThreshold, divide the infrared image into dark and bright parts, perform histogram equalization on the dark and bright sub-images, and then stitch them together to obtain the enhanced image.

[0104] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0105] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art will be able to modify the technical solutions described in the aforementioned embodiments or substitute equivalents for some of the technical features. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. An infrared image processing system, characterized in that: The following steps are involved: Step S1, according to the curve of the detector's response to the black body as a function of temperature at each integration time, determine the calibration temperature point and the judgment threshold r for the segmented calibration; The statistical analysis of the detector's response to blackbodies of different temperatures under different integration time conditions is implemented as follows: Step S101, when plotting the curve of the detector's response to a blackbody versus temperature, a corner-cutting method is used to calculate the mean pixel response of the infrared focal plane as the ideal detector response to the blackbody at the current temperature: when calculating the mean detector response, only a circular area with a diameter of min(H, W) centered at the detector center is considered, where H and W represent the row height and column width of the focal plane array, respectively; Step S102, selection of calibration temperature points: divide the response curve into low temperature section and high temperature section according to the gradient change of the response curve, make each section approximate to a linear function, select t1 and t2 as the two calibration points of the low temperature section; select t 3， t4 is used as the calibration temperature of the two points in the high temperature section, and the middle temperature of each section is recorded as t 11 ,t 12 As a single point calibration point, use the average of the two intermediate temperatures As the criterion for selecting the temperature interval Tt, the mean value of the detector response under Tt is used as the segmentation threshold r. When performing segmented correction, the correction coefficient is determined by comparing the mean value of the pixel response with the threshold. Step S2, using a ring-shaped regional method to calculate the gain coefficients k1 and k2 of the two-point correction of the low temperature section and the high temperature section respectively; using a single-point temperature calibration method to obtain the bias coefficients b1 and b2, and matching the bias coefficients b1 and b2 with the gain coefficients k1 and k2 and saving them; When calculating the gain coefficient, a ring partition method was used, and the intermediate temperature t 11 , t 12 Perform a single-point calibration to calculate the bias coefficient: Step S201, calculate the gain coefficient according to formula (1): In formula (1), Respectively represent t H , t L Radiative flux at temperature; represents the mean value of the original response output of the detector pixel in the time domain; Y H 、Y L Respectively represent t H , t L Ideal response value at temperature, k i,j Represents the gain coefficient; where, when obtaining Y H 、Y L When , according to the characteristic that the detector response gradually decays with distance from the center to the outside, the mean value of the annular area where the pixel is located is used for calculation; Step S202: Based on the gain coefficient obtained in step S201, perform gain correction in a segmented manner. 11 The bias coefficient is calculated according to formula (2) at temperature: In formula (2), Y represents the ideal pixel response output value; b represents the original response output value of the detector pixel; i,j represents the bias coefficient; Step S3, performing multi-point calibration according to the judgment threshold r, performing blind pixel calibration on the slope of the blackbody temperature curve, the pixel response amplitude at a fixed integration time and fixed temperature, and the detector response standard deviation according to the pixel response at different integration times, and saving the calibration results; Step S4, compensating the blind pixel with the average of the compensation value of the previous frame of the blind pixel and the response value of the neighboring pixel in the current frame, closing the camera baffle, performing single-point calibration, and correcting the bias coefficient b; Step S5: Perform scene-based blind pixel detection on the baffle scene, update the blind pixel list, and compensate for new blind pixels; Step S6, performing channel compensation; Step S7: Use the DDE algorithm to enhance the details of the infrared image, increase the contrast, and optimize the visual effect.

2. The infrared image processing system according to claim 1, characterized in that: In step S3, blind pixel calibration is performed based on the slope of the curve of the detector's response to the blackbody versus temperature, the standard deviation of the pixel response at a fixed integration time and fixed temperature, and the response amplitude. The implementation process is as follows: Step S301, by judging the conditions , calibrate blind pixels; among them, represents the mean value of the gain coefficient in the spatial domain, Represents the gain coefficient of each pixel; Step S302: satisfy the condition The pixels of are marked as blind pixels, where Represents the gradient of the pixel response curve; Step S303: If the pixel response gray value If the change value between the image of more than 50 frames and the previous frame exceeds 50 within 100 frames, the pixel is determined to be a blind pixel.

3. The infrared image processing system according to claim 1, wherein: In step S4, the blind pixel response value is replaced by the average of the blind pixel compensation value of the previous frame and the neighborhood pixel response value of the current frame to perform blind pixel compensation, and then single-point correction is performed based on the scene to correct the bias coefficient.

4. The infrared image processing system according to claim 1, wherein: In step S5, a variance-based blind pixel detection algorithm is used: the overall response is normalized after removing the background component, the deviation degree of the pixel response from the average response is calculated, and the deviation degree of the pixel response from the average response is compared with the overall mean square error to determine the blind pixel.

5. The infrared image processing system according to claim 1, wherein: In step S6, channel compensation is performed by correcting the channel mean to further remove the stripe noise: the channel compensation value is obtained according to formula (3): , according to the channel where the pixel is located, the corresponding Add to the original response value to perform channel compensation; In formula (3), P i represents the mean value in each channel, P avg Represents the overall detector response mean.

6. The infrared image processing system according to claim 1, characterized in that: In step S7, the DDE algorithm is used to enhance the image. The algorithm is implemented as follows: Step S701, determining the grayscale threshold by formula (4) to divide the infrared image into dark part and highlight part; In formula (4), pixMed is the grayscale median, pixAvg is the grayscale mean, and MM is the threshold shift coefficient; Step S702 : performing histogram equalization on the dark part sub-image and the bright part sub-image respectively and then splicing them together to obtain an enhanced image.

Images