Infrared image processing method and device, computer device and readable storage medium

Abstract

The application relates to an infrared image processing method and device, computer equipment and a readable storage medium, comprising: acquiring a to-be-corrected image; fusing identification results of multiple blind element screening criteria to generate a target blind element table; the blind element screening criteria comprise at least two of a response rate criterion, a time domain noise criterion, a spatial gray difference criterion, a linearity and two-point gain coefficient criterion; determining a to-be-corrected blind element according to the target blind element; the to-be-corrected blind element comprises a single blind element and a continuous blind element; adopting a preset dynamic gradient correction strategy, a preset edge adaptation strategy and a flow processing adaptation strategy, the to-be-corrected blind element is corrected and processed to obtain a target image. The application accurately identifies a continuous blind element through multi-criterion fusion screening, determines a correction parameter by using a dynamic gradient, improves the correction effect of an edge region by using an edge adaptation strategy, and ensures real-time performance by using flow processing compatibility, effectively solving the problems of poor correction effect and insufficient adaptability in the prior art.

Description

Technical Field

[0001] This application relates to the field of infrared image processing technology, and in particular to an infrared image processing method, apparatus, computer equipment, and readable storage medium. Background Technology

[0002] Infrared detectors inevitably contain blind pixels due to manufacturing defects. Consecutive blind pixels refer to multiple blind pixels in adjacent positions, such as 2-3 adjacent blind pixels. When the number of consecutive blind pixels is large, their selection and correction become very difficult. Current technologies primarily rely on a single criterion for blind pixel selection. However, blind pixel selection schemes fail to effectively distinguish between noise and true blind pixels, easily misclassifying normal pixels as blind pixels or missing detections of flash pixels.

[0003] In terms of blind pixel correction, existing algorithms design gradient calculation methods for 1, 2, and 3 blind pixels separately, which requires differentiation of the number of blind pixels, resulting in complex logic and high resource consumption. During data pipeline processing, the correction result of the nth row cannot be updated to the (n+1)th row in real time, leading to large deviations in the replacement values ​​of consecutive blind pixels at the bevel of the silicon wafer, resulting in jagged or blurry phenomena. Summary of the Invention

[0004] Based on this, it is necessary to provide an infrared image processing method, apparatus, computer equipment, and readable storage medium that can achieve efficient correction of 1-3 consecutive blind cells, especially optimize edge region processing, adapt to streaming data processing logic, and balance screening accuracy and correction adaptability, in order to address the above-mentioned technical problems.

[0005] In a first aspect, this application provides an infrared image processing method, the method comprising: Acquire an image to be corrected, wherein the image to be corrected is a series of consecutive frames of images acquired under multiple scene configuration parameters; the scene configuration parameters include gain parameters, integration time, and scene brightness; The identification results of multiple blind element screening criteria are integrated to generate a target blind element table; wherein, the target blind element table includes the location information of the blind elements; the blind element screening criteria include at least two criteria from the following: response rate criterion, temporal noise criterion, spatial gray-level difference criterion, linearity and two-point gain coefficient criterion; The blind cells to be corrected are determined based on the target blind cells; the blind cells to be corrected include single blind cells and consecutive blind cells. The target image is obtained by correcting the blind pixels to be corrected using a preset dynamic gradient correction strategy, a preset edge adaptation strategy, and a pipelined processing adaptation strategy.

[0006] In one embodiment, the gain parameter includes a first gain, a second gain, and a third gain; the integration time includes a first integration time, a second integration time, and a third integration time; and the scene brightness includes dark field brightness, mid-field brightness, and bright field brightness. Specifically, the first gain is less than the second gain, the second gain is less than the third gain, the first integration time is less than the second integration time, the second integration time is less than the third integration time, the dark field brightness is less than the mid-field brightness, and the mid-field brightness is less than the bright field brightness. The acquisition of the image to be corrected includes: Acquire dark field, mid field, and bright field images under different gains and integration times; for each scene configuration parameter, acquire 10 or more consecutive images.

[0007] In one embodiment, acquiring the image to be corrected includes: Acquire the bright-field image at the first gain and the first integration time; Obtain the midfield image under the second gain and the second integral time; Obtain the dark field image at the third gain and the third integration time.

[0008] In one embodiment, before generating the target blind element table by fusing the identification results of multiple blind element screening criteria, the method further includes: Blind elements in the image to be corrected are labeled according to different blind element screening criteria to obtain the recognition results corresponding to different blind element screening criteria; The response rate criterion is that if the difference between the brightness difference between the bright field image and the dark field image corresponding to a pixel and the average brightness of all pixels is greater than a preset difference threshold, the pixel is marked as a blind pixel. The temporal noise criterion is that if the temporal noise of a pixel is greater than twice the average noise of all pixels, the pixel is marked as a blind pixel. The spatial grayscale difference criterion is that if the difference between the grayscale response value of a pixel and the grayscale mean value in the preset neighborhood of the pixel is greater than a preset difference threshold, the pixel is marked as a blind pixel. The linearity and two-point gain coefficient criteria are as follows: if the linearity of a pixel does not fall within a preset threshold range, or the two-point correction gain coefficient of a pixel does not fall within a preset threshold range, the pixel is marked as a blind pixel.

[0009] In one embodiment, the preset edge adaptation strategy includes a row edge processing strategy and a column edge processing strategy; the row edge processing strategy uses a mirror expansion method to copy adjacent row data to expand the edge row data; the column edge processing strategy uses a column copy method to copy the correction results of blind columns to correct the edge column data; the pipelined processing adaptation strategy uses storing the correction results of the nth row in an independent buffer, and calculating the neighborhood of the (n+1)th row using the original data of the nth row. The step involves employing a preset dynamic gradient correction strategy, a preset edge adaptation strategy, and a pipelined processing adaptation strategy to correct the blind pixels to be corrected, thereby obtaining the target image. This includes: Calculate the basic gradient direction value of the blind cell to be corrected; wherein, if the blind cell to be corrected is a continuous blind cell, and there are more than or equal to 3 continuous blind cells, calculate the corresponding basic gradient direction value and supplementary gradient direction value. The pixel value of the blind cell to be corrected is replaced by the average pixel value in the direction corresponding to the minimum effective gradient direction value; wherein, the effective gradient direction value is the gradient direction value that is less than a preset gradient threshold, and the preset gradient threshold is the product of the maximum number of bits in the image and the dynamic coefficient.

[0010] In one embodiment, the method further includes: If the noise of the image to be corrected is greater than a first noise threshold, the dynamic coefficient is determined to be the first coefficient; If the noise of the image to be corrected is less than or equal to the first noise threshold and greater than or equal to the second noise threshold, the dynamic coefficient is determined to be the second coefficient; the second coefficient is less than the first coefficient. If the noise of the image to be corrected is less than the second noise threshold, the dynamic coefficient is determined to be the third coefficient; the third coefficient is less than the second coefficient.

[0011] Secondly, this application also provides an infrared image processing apparatus. The apparatus includes: An acquisition module is used to acquire an image to be corrected, wherein the image to be corrected is a series of consecutive frames of images acquired under multiple scene configuration parameters; the scene configuration parameters include gain parameters, integration time, and scene brightness. A filtering module is used to integrate the recognition results of multiple blind element filtering criteria to generate a target blind element table; wherein, the target blind element table includes the location information of the blind elements; the blind element filtering criteria include at least two criteria from the following: response rate criterion, temporal noise criterion, spatial gray-level difference criterion, linearity and two-point gain coefficient criterion; A determination module is used to determine blind cells to be corrected based on the target blind cells; the blind cells to be corrected include single blind cells and consecutive blind cells. The correction module is used to perform correction processing on the blind pixels to be corrected using a preset dynamic gradient correction strategy, a preset edge adaptation strategy, and a pipelined processing adaptation strategy to obtain the target image.

[0012] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the infrared image processing method described in the first aspect.

[0013] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the steps of the infrared image processing method described in the first aspect.

[0014] Fifthly, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, implements the steps of the infrared image processing method described in the first aspect.

[0015] In summary, this application provides an infrared image processing method, apparatus, computer device, and readable storage medium, comprising: acquiring an image to be corrected; fusing the recognition results of multiple blind pixel screening criteria to generate a target blind pixel table; the blind pixel screening criteria include at least two criteria selected from response rate criteria, temporal noise criteria, spatial gray-level difference criteria, linearity criteria, and two-point gain coefficient criteria; determining blind pixels to be corrected based on the target blind pixels; the blind pixels to be corrected include single blind pixels and continuous blind pixels; and performing correction processing on the blind pixels to be corrected using a preset dynamic gradient correction strategy, a preset edge adaptation strategy, and a pipelined processing adaptation strategy to obtain a target image. This application accurately identifies continuous blind pixels through multi-criteria fusion screening, uses dynamic gradient to determine correction parameters, improves the correction effect in edge regions through edge adaptation strategies, and ensures real-time performance through pipelined processing compatibility, effectively solving the problems of poor correction effect and insufficient adaptability in existing technologies.

[0016] In related technologies, local median filtering is used to correct continuous blind cells, but the gray-level difference in the edge region of three consecutive blind cells is >15DN, and the algorithm needs to be designed to distinguish the number of blind cells. This application reduces the gray-level difference to <5DN through a 5×5 neighborhood dynamic gradient, and does not require distinguishing the number of blind cells. Attached Figure Description

[0017] Figure 1 This is an application environment diagram of an infrared image processing method in one embodiment; Figure 2 This is a flowchart illustrating an infrared image processing method in one embodiment; Figure 3This is a flowchart illustrating the steps involved in correcting blind cells to be corrected in one embodiment. Figure 4 This is a schematic diagram showing the distribution of gradient direction values ​​corresponding to the blind cells to be corrected in one embodiment. Figure 5 This is a schematic diagram showing the distribution of gradient direction values ​​corresponding to the blind cells to be corrected in another embodiment; Figure 6 This is a schematic diagram of the pixel distribution in a 3*5 neighborhood corresponding to the blind cell to be corrected in one embodiment; Figure 7 This is a schematic diagram of the infrared image before correction in one embodiment; Figure 8 This is a schematic diagram of a corrected infrared image in one embodiment; Figure 9 This is a structural block diagram of an infrared image processing device in one embodiment; Figure 10 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0019] The infrared image processing method provided in this embodiment involves the detection, screening and correction of continuous blind elements in infrared cameras. It is applicable to short-wave, mid-wave and long-wave infrared array and linear array cameras, especially for the correction of three continuous blind elements.

[0020] The infrared image processing method provided in this application embodiment can be applied to, for example... Figure 1 The application environment is shown. Terminal 102 can be an infrared camera or a smart device with an infrared imaging module. Terminal 102 communicates with server 104 via a network. Terminal 102 acquires infrared images and sends them to server 104. Server 104 executes the infrared image processing method of this application, completes blind pixel screening and correction, and then returns the target image to terminal 102 or stores it in a data storage system. The data storage system can store the data that server 104 needs to process. The data storage system can be integrated into server 104 or can be a standalone cloud storage device. Terminal 102 can be, but is not limited to, an infrared camera, smartphone, tablet, IoT device, etc.; server 104 can be a standalone server or a server cluster composed of multiple servers.

[0021] In one embodiment, such as Figure 2 As shown, an infrared image processing method is provided, which can be applied to... Figure 1 Taking the application environment in [the document] as an example, the following steps are included: Step 202: Obtain the image to be corrected.

[0022] The images to be calibrated are multiple consecutive frames acquired under various scene configuration parameters. These scene configuration parameters include gain, integration time, and scene brightness.

[0023] Specifically, the image to be corrected refers to a series of consecutive infrared images acquired by an infrared array camera or an infrared linear array camera under multiple scene configuration parameters. In this embodiment, the actual type and specifications of the infrared camera can be configured according to the needs of the actual application scenario. A series of consecutive images refers to multiple frames of infrared images acquired during continuous shooting by the infrared camera. This embodiment, by acquiring a series of consecutive images, helps to obtain richer scene information, providing a sufficient data foundation for subsequent image correction.

[0024] Specifically, scene configuration parameters include adjustable parameters such as gain, integration time, and scene brightness. The gain parameter primarily adjusts the amplification of the signal by the infrared camera. By changing the gain parameter, the signal strength of the image can be enhanced or weakened, thus affecting the image's brightness and contrast. Appropriately setting the gain parameter under different scene conditions can optimize image quality, making image details more clearly visible. Integration time refers to the length of time the infrared camera sensor accumulates charge during signal acquisition. The length of the integration time directly affects the image exposure. A longer integration time allows the image to acquire more signal, thus increasing image brightness, but may also increase noise; a shorter integration time, conversely, results in lower image brightness but relatively less noise. Depending on the scene brightness, the integration time needs to be adjusted appropriately to achieve suitable exposure. Scene brightness refers to the light intensity of the scene being photographed. Different scenes have different brightness levels, such as bright outdoor scenes and dim indoor scenes. Changes in scene brightness significantly affect the quality of infrared images; therefore, other camera parameters, such as gain and integration time, need to be adjusted according to the actual scene brightness to ensure that the acquired image quality meets requirements.

[0025] In this embodiment, the gain parameters include three typical gains: a first gain G1, a second gain G2, and a third gain G3. The integration time includes three typical integration times: a first integration time T1, a second integration time T2, and a third integration time T3. The scene brightness includes dark field brightness, mid-field brightness, and bright field brightness. Specifically, the first gain G1 is less than the second gain G2, the second gain G2 is less than the third gain G3, the first integration time T1 is less than the second integration time T2, the second integration time T2 is less than the third integration time T3, the dark field brightness is less than the mid-field brightness, and the mid-field brightness is less than the bright field brightness. Specifically, the first gain G1 corresponds to low gain, the second gain G2 corresponds to medium gain, and the third gain G3 corresponds to high gain. The first integration time T1 corresponds to low integration time, the second integration time T2 corresponds to medium integration time, and the third integration time T3 corresponds to high integration time.

[0026] In this embodiment, a 14-bit infrared array camera (640×512 resolution) is used as an example, with a dynamic range of 0-16383 DN. The bright field brightness can be 10000 DN, the mid-field brightness can be 8000 DN, and the dark field brightness can be 4000 DN.

[0027] In one embodiment, the step of acquiring the image to be corrected may involve acquiring dark-field images, mid-field images, and bright-field images under different gains and integration times. Specifically, for each scene configuration parameter, 10 or more consecutive images are acquired.

[0028] In this embodiment, for example, a 14-bit infrared array camera (640×512 resolution) is selected, with a dynamic range of 0-16383DN. Three typical gains (low gain G1, medium gain G2, high gain G3) and three typical integration times (short integration time T1, medium integration time T2, long integration time T3) are configured. The scene is automatically matched according to the camera's dynamic range, and 10 or more consecutive images are acquired under each scene's configuration parameters. In a feasible embodiment, 10 images can be acquired in both dark and bright field environments, and 20 images can be acquired in a medium field environment to ensure the accuracy of subsequent blind pixel identification based on the image to be corrected.

[0029] It is worth noting that in the subsequent process of identifying blind pixels based on individual criteria, either multiple frames of raw data or multiple frames of average data can be used. Here, multiple frames of raw data refer to the original image data, while multiple frames of average data refer to the average of multiple image frames. For example, the temporal noise criterion uses multiple frames of raw data, while other criteria use multiple frames of average data.

[0030] In one embodiment, acquiring the image to be corrected includes: acquiring a bright field image at a first gain and a first integration time; acquiring a mid-field image at a second gain and a second integration time; and acquiring a dark field image at a third gain and a third integration time.

[0031] In this embodiment, the matching logic for gain and integration time is as follows: automatic matching is performed based on the dynamic range of the infrared camera; low gain and short integration time are adapted for bright fields, while high gain and long integration time are adapted for dark fields. Specifically, a bright field (10000DN) image can be acquired at the first gain G1 and the first integration time T1; a mid-field (8000DN) image can be acquired at the second gain G2 and the second integration time T2; and a dark field (4000DN) image can be acquired at the third gain G3 and the third integration time T3.

[0032] Based on the above scheme, the image quality of the collected images to be corrected can be effectively guaranteed, and the recognition accuracy of subsequent blind pixel screening criteria can be improved.

[0033] Step 204: Integrate the recognition results of multiple blind element screening criteria to generate a target blind element table.

[0034] The target blind cell table includes the location information of the blind cells. The blind cell selection criteria include at least two of the following criteria: response rate criterion, temporal noise criterion, spatial gray-level difference criterion, linearity criterion, and two-point gain coefficient criterion.

[0035] In this embodiment, a blind pixel refers to a pixel in an infrared image that responds abnormally and cannot accurately reflect scene information. Types of blind pixels include dead pixels, overheated pixels, and scintillation pixels.

[0036] In one embodiment, before generating the target blind pixel table by fusing the recognition results of multiple blind pixel selection criteria, it is necessary to label the blind pixels in the image to be corrected according to different blind pixel selection criteria, thereby obtaining the recognition results corresponding to different blind pixel selection criteria. The specific labeling rules for each criterion are as follows: The response rate criterion is as follows: for each pixel, calculate the brightness difference between its corresponding bright-field and dark-field images. If the difference between this difference and the average brightness of all pixels is greater than a preset difference threshold, then the pixel is marked as a blind pixel. In terms of distance, if the brightness difference between the bright-field and dark-field images is... All pixels The mean is ,like If the brightness difference deviates from the average brightness by ±30%, and the preset difference threshold is 30%, the pixel is marked as a blind pixel. It should be noted that the preset difference threshold can be configured according to the needs of the actual application scenario.

[0037] The temporal noise criterion is that for each pixel, if its temporal noise is greater than twice the average noise of all pixels, then that pixel is marked as a blind pixel. It is worth noting that blind pixels that meet the temporal noise criterion can be marked as flash pixels. For example, the temporal noise of each pixel is calculated... All pixels The mean is ,like If the noise level exceeds twice the average noise level, the pixel is marked as a blind pixel (flicker pixel). It should be noted that the preset noise multiple threshold can be configured according to the needs of the actual application scenario.

[0038] The spatial grayscale difference criterion is as follows: For each pixel, a preset neighborhood is determined. If the difference between the grayscale response value of that pixel and the average grayscale value within its neighborhood is greater than a preset difference threshold, then that pixel is marked as a blind pixel. For example, using a 3×3 neighborhood average... For reference, calculate the grayscale response value (mid-field average) of the pixel. If this response value... A pixel is marked as a blind pixel if it deviates from the neighborhood mean by ±30%, with a preset difference threshold of 30%. It should be noted that the preset difference threshold can be configured according to the needs of the actual application scenario.

[0039] The linearity and two-point gain coefficient criteria are as follows: if the linearity of a pixel does not fall within a preset threshold range, or if the two-point correction gain coefficient of a pixel does not fall within a preset threshold range, the pixel is marked as a blind pixel. Specifically, the linearity and two-point gain coefficient criteria are the criteria corresponding to the images to be corrected acquired by the infrared array camera. For example, if the linearity of a pixel deviates from the preset threshold range, i.e., R² < 0.95 of the fitted line of the response value at the same integration time, where the preset linearity threshold is 5%; or if the two-point correction gain coefficient G of a pixel does not fall within the preset threshold range, i.e. The preset gain coefficient threshold is 10%, and the pixel is marked as a blind pixel.

[0040] In this embodiment, linearity is used to measure the degree of deviation between the response value of a pixel and the fitted straight line at different integration times. Ideally, the response value of a pixel should have a strict linear relationship with the integration time, that is, the response value can fall well on a straight line, in which case the coefficient of determination R² of the fitted straight line is close to 1. The closer R² is to 1, the better the fit between the response value and the fitted straight line, and the higher the linearity; conversely, the smaller R² is, the worse the linearity. Given a threshold condition for linearity deviation of R² < 0.95, when the R² value obtained after calculating the fitted straight line for the response value of a pixel at the same integration time is less than 0.95, it indicates that the linearity of the pixel is poor, and its response value deviates significantly from the ideal linear relationship. In this case, the pixel is marked as a blind pixel.

[0041] Two-point calibration involves sampling two points of known brightness (typically a dark spot and a bright spot) to calculate a gain coefficient (G parameter) and a bias coefficient. These are used to correct the pixel's response value, making it more accurately reflect the brightness information of the actual scene. The gain coefficient G reflects the degree to which the pixel amplifies or reduces the input signal. Ideally, G should be close to 1, indicating that the pixel can accurately respond to the input signal. The threshold range for the given G parameter is... When the gain coefficient G of a pixel is calculated using the two-point correction method, if the value of G is not within the range of 0.9 to 1.1, it indicates that the pixel has a large deviation in its response to the input signal and cannot accurately amplify or reduce the input signal. In this case, the pixel is marked as a blind pixel.

[0042] Specifically, after obtaining the corresponding blind cell identification results based on the above criteria, the identification results of all the above criteria are merged, and the union of the four criteria is taken to mark the position information of continuous blind cells, including the coordinates of horizontal continuous blind cells. 、( ), ( ) or vertical continuous blind element coordinates ( ), ( ), ( ), generate a target blind element table containing blind element location information.

[0043] In one embodiment, step 204 further includes a blind pixel table dynamic update step. After acquiring 100 frames of images in step 202, the blind pixel screening process is re-executed based on multiple blind pixel screening criteria to dynamically update the position information of time-varying blind pixels such as flash pixels, ensuring the accuracy of the target blind pixel table.

[0044] Step 206: Determine the blind cells to be corrected based on the target blind cells.

[0045] In this embodiment, after generating the target blind cell table, it is necessary to further determine the blind cells to be corrected based on the target blind cells. The blind cells to be corrected are mainly divided into two categories: single blind cells and continuous blind cells.

[0046] A single blind cell refers to a single, independently existing anomalous cell. Its significant characteristic is that all adjacent cells of the blind cell are normal cells, meaning that there are no other blind cells in its surrounding neighborhood.

[0047] A consecutive blind cell refers to a group of blind cells that are immediately adjacent to each other. Based on the number of adjacent blind cells, it can be further subdivided into two-cell blind cells and three-cell blind cells. A two-cell blind cell consists of two adjacent blind cells. These two blind cells are spatially connected, for example, at coordinates [coordinate missing] in the horizontal direction. and( Two blind elements, or those with coordinates ( ) in the vertical direction. )and( Two blind cells.

[0048] A triplet blind cell consists of three consecutive adjacent blind cells. Taking the horizontal direction as an example, the coordinates are... 、( ), ( The three blind elements form a horizontal triple blind element. In the vertical direction, the coordinates ( ), ( ), ( The three blind cells form a vertical triple blind cell.

[0049] Step 208: Using a preset dynamic gradient correction strategy, a preset edge adaptation strategy, and a pipelined processing adaptation strategy, the blind pixels to be corrected are corrected to obtain the target image.

[0050] In this embodiment, the preset dynamic gradient correction strategy refers to using different correction methods for different types of blind cells to be corrected, so as to ensure the accuracy and effectiveness of the correction.

[0051] For single and double blind pixels, pixel values ​​along a common gradient direction are used for correction. The common gradient direction is determined based on the local features and gray-level variation trends of the image. By extracting normal pixel value information along this direction, the blind pixel is interpolated or estimated to recover its proper pixel value.

[0052] For triple blind pixels, the situation is more complex due to the large number of blind pixels involved. Therefore, a general gradient combined with pixel values ​​in the supplementary gradient direction is used for correction. The supplementary gradient direction is one or more additional directions selected based on the specific features of the image and the distribution of blind pixels, in addition to the general gradient direction. By integrating pixel value information from the general gradient direction and the supplementary gradient direction, the pixel values ​​of triple blind pixels can be recovered more comprehensively and accurately, improving the correction quality.

[0053] In this embodiment, the preset edge adaptation strategy includes a row edge processing strategy and a column edge processing strategy. The row edge processing strategy uses a mirror expansion method to copy adjacent row data to expand the edge row data. The column edge processing strategy uses a column copy method to copy the correction results of the blind cell column to correct the edge column data. Specifically, after determining the blind cell to be corrected, it is first determined whether the blind cell is located at the row edge or column edge of the image to be corrected based on its coordinate position. Specifically, if the blind cell is located in the area near the top two rows or the area near the bottom two rows of the image to be corrected, it is determined that the blind cell is located at the row edge. If the blind cell is located in the area of ​​the leftmost three columns or the rightmost three columns of the image to be corrected, it is determined that the blind cell is located at the column edge.

[0054] If the blind pixel is located at a row edge, the adjacent row data is copied using a mirror expansion method to generate a virtual row to supplement the 5×5 or 3×3 neighborhood. If the blind pixel is located at a column edge, the position of the nearest normal pixel is recorded to prepare for subsequent column copying. It should be noted that the actual execution methods of mirror expansion and column copying can be configured according to the needs of the specific application scenario.

[0055] In this embodiment, the pipelined processing adaptation strategy stores the correction result of the nth row in an independent buffer, and calculates the neighborhood of the (n+1)th row using the original data of the nth row. Specifically, the correction result of the current row is stored in an independent buffer and does not participate in the neighborhood calculation of the next row. The neighborhood calculation of the next row only uses the original data of the current row, ensuring the real-time performance of the pipelined processing.

[0056] Based on the above scheme, this embodiment provides an infrared image processing method. Supplementary gradient calculation is designed for triple blind pixels, and edge adaptation strategies such as mirror expansion and column replication are combined to solve the jaggedness and blurring problems in edge regions such as the bevel of the silicon wafer. After correction, the grayscale difference between the edge region of the triple blind pixel and surrounding pixels is effectively reduced. Images are acquired by configuring parameters for multiple scenes, and four types of blind pixel screening criteria are fused to effectively distinguish noise from real blind pixels. The blind pixel detection accuracy is high, reducing missed detections (such as flash pixels) and misjudgments (such as normal pixels being marked as blind pixels).

[0057] In a specific embodiment, step 204 requires preprocessing the acquired multi-frame images during actual execution. The temporal noise criterion directly uses the original data of multiple frames, while the response rate criterion, spatial grayscale difference criterion, linearity and two-point gain coefficient criterion use the average data of multiple frames. The dark field and bright field are taken as the average of 10 frames, and the mid-field is taken as the average of 20 frames to reduce random noise interference.

[0058] After labeling blind cells according to the judgment rules of response rate criterion, temporal noise criterion, spatial gray-level difference criterion, linearity criterion, and two-point gain coefficient criterion, and recording the coordinates of the blind cells, the union of the labeling results of the four criterions can be taken to remove blind cells with duplicate labels. All labeled blind cells are traversed to identify 1-3 blind cells that are horizontally continuous (coordinates differ by 1 column) or vertically continuous (coordinates differ by 1 row), and the number and complete coordinates of the continuous blind cells are determined. Finally, the location information (coordinates), the number of continuous blind cells, and the continuous direction (horizontal / vertical) of the blind cells are compiled into a target blind cell table and stored in a local or server database for use in subsequent correction steps.

[0059] refer to Figure 3 The target image is obtained by using a preset dynamic gradient correction strategy, a preset edge adaptation strategy, and a pipelined processing adaptation strategy to correct the blind pixels to be corrected, including the following steps: Step 301: Calculate the basic gradient direction value of the blind element to be corrected.

[0060] In this embodiment, as Figure 4 and Figure 5 As shown, the middle or left blind cell of the three-cell blind group is taken as the center ( The formula for calculating its basic gradient is as follows: Step 302: If the blind cells to be corrected are continuous blind cells and there are more than or equal to 3 continuous blind cells, calculate the corresponding basic gradient direction value and supplementary gradient direction value.

[0061] In this embodiment, the formula for calculating the supplementary gradient includes the following formula: If we take the middle blind of the three consecutive blinds as the center ( The formula for calculating the gradient can also include a vertical gradient: The formula for calculating the basic gradient direction value can be found in the specific implementation method in step 301 above, and will not be repeated here.

[0062] Step 303: Replace the pixel value of the blind cell to be corrected with the pixel mean value of the direction corresponding to the minimum effective gradient direction value.

[0063] The effective gradient direction value is the gradient direction value that is less than the preset gradient threshold, which is the product of the maximum number of bits in the image and the dynamic coefficient.

[0064] In this embodiment, a preset gradient threshold is used. ,in, These are dynamic coefficients, which can adaptively change within the range of 0.03-0.05. The maximum pixel value is the image bit depth, for example, the maximum pixel value for a 14-bit image. For 16383, 8-bit image It is 255.

[0065] Will be greater than or equal to the preset gradient threshold The gradient direction value is determined to be an invalid gradient direction value, which will be less than the preset gradient threshold. The gradient direction value is determined as the effective gradient direction value.

[0066] In this embodiment, the first gradient direction value The corresponding formulas for calculating the pixel mean include: in, To replace pixel values, coordinates The pixel value of the cell, coordinates The pixel value of the cell. First gradient direction value. The corresponding pixel mean is calculated using an equal-weighted average method.

[0067] Second gradient direction value The corresponding formulas for calculating the pixel mean include: in, coordinates The pixel value of the cell, coordinates The pixel value of the cell. Second gradient direction value. The corresponding pixel mean is calculated using a nearest-neighbor weighted average method.

[0068] Third gradient direction value The corresponding formulas for calculating the pixel mean include: in, coordinates The pixel value of the cell, coordinates The pixel value of the cell. Third gradient direction value. The corresponding pixel mean is calculated using a nearest-neighbor weighted average method.

[0069] Fourth gradient direction value The corresponding formulas for calculating the pixel mean include: in, coordinates The pixel value of the cell, coordinates The pixel value of the cell. Third gradient direction value. The corresponding pixel mean is calculated using a nearest-neighbor weighted average method.

[0070] Fifth gradient direction value The corresponding formulas for calculating the pixel mean include: in, coordinates The pixel value of the cell, coordinates Pixel value of the pixel. The third gradient direction value The corresponding pixel mean value is calculated by the nearest neighbor weighted average method.

[0071] The sixth gradient direction value The calculation formula of the corresponding pixel mean value includes: Where, is the coordinate pixel value of the pixel, is the coordinate pixel value of the pixel. The first gradient direction value The corresponding pixel mean value is calculated by the equal weight average method.

[0072] In practical applications, the pixel mean value in the direction corresponding to the minimum direction value of the effective gradient direction value is used to replace the pixel value of the to-be-corrected blind pixel. For example, for a single blind pixel and a double-connected blind pixel, the first gradient direction value , the second gradient direction value and the third gradient direction value are calculated. If the second gradient direction value is the minimum direction value, then the pixel replacement value of the blind pixel is calculated according to the pixel mean value calculation method corresponding to the second gradient direction value , and the blind pixel is corrected based on the pixel replacement value.

[0073] For a triple-connected blind pixel, the first gradient direction value , the second gradient direction value and the third gradient direction value are calculated. Then, the fourth gradient direction value and the fifth gradient direction value are further calculated, and only <Thr effective gradient direction values are retained. Assume the effective gradients are , , , and is the minimum direction value. Then, the pixel replacement value of the blind pixel is calculated according to the pixel mean value calculation method corresponding to the fourth gradient direction value , and the blind pixel is corrected based on the pixel replacement value.

[0074] It is worth noting that for blind cells in edge regions to be corrected, the neighborhood data is first expanded using either row or column edge processing strategies before gradient calculation and replacement value updates are performed. Row data is processed according to a pipelined adaptation strategy to ensure the real-time performance of the correction process. It should be noted that, through the pipelined adaptation strategy, row data latency is ≤1 row period (≤10μs at 14-bit 640×512 resolution), avoiding row data waiting delays, improving real-time performance, and meeting the high-speed requirements of infrared image processing.

[0075] In a more detailed embodiment, during the calculation of the gradient direction value, a reference is made. Figure 6 With blind cells as the center Using a sliding window of size 5*3, calculate the following four gradients: in, and The calculation formula corresponds to The calculation formula, The calculation formula corresponds to The calculation formula, The calculation formula corresponds to The calculation formula. It should be noted that the above combinations of letters and numbers, such as... All correspond to actual pixel coordinates, where, The gray area represents the blind pixels to be corrected, while the remaining gray areas correspond to the corrected pixels.

[0076] If the current blind cell is the left blind cell of a triple blind cell, then calculate the supplementary gradient. , ,otherwise , Assignment is executed. Specifically, , The calculation formula is: Pick ~ The minimum gradient value is used to calculate the gray-scale mean value in the direction of the minimum gradient as the correction value.

[0077] It is worth noting that the supplementary gradient direction value can be any optional gradient direction that differs from the general gradient direction. The actual calculation formula for the supplementary gradient direction value can be designed according to the needs of the specific application scenario.

[0078] Based on the above scheme, an additional supplementary gradient is designed for the triple blind pixel, which can solve the problem of substitution value deviation at the bevel edge of the silicon wafer, resulting in a smooth edge region without jagged edges or blurring. Specifically, infrared images that have not undergone blind pixel correction using the infrared image processing method provided in this embodiment are as follows: Figure 7As shown, the infrared image processed using the infrared image processing method provided in this embodiment for blind pixel correction is as follows: Figure 8 As shown.

[0079] Based on the above steps, without distinguishing the number of blind pixels, this embodiment adopts a combination of general gradient and supplementary gradient to adapt 1-3 consecutive blind pixels. It supports area / line array, short-wave / mid-wave / long-wave infrared cameras, and does not rely on blackbody calibration. It adapts to actual working scenarios with multiple gains and multiple integration times, greatly improving the application range of infrared image processing methods. Furthermore, supplementary gradient calculation is designed for three consecutive blind pixels. Combined with the edge adaptation strategy, it solves the jaggedness and blurring problems in edge areas such as the bevel of the silicon wafer. After correction, the grayscale difference between the edge area of ​​the three consecutive blind pixels and the surrounding pixels is effectively controlled.

[0080] In one embodiment, the infrared image processing method further includes: If the noise in the image to be corrected is greater than a first noise threshold, the dynamic coefficient is determined as the first coefficient. If the noise in the image to be corrected is less than or equal to the first noise threshold and greater than or equal to a second noise threshold, the dynamic coefficient is determined as the second coefficient; the second coefficient is less than the first coefficient. If the noise in the image to be corrected is less than the second noise threshold, the dynamic coefficient is determined as the third coefficient; the third coefficient is less than the second coefficient.

[0081] In this embodiment, the first noise threshold is preset to 10DN, the second noise threshold is preset to 5DN; the first coefficient is 0.05, the second coefficient is 0.04, and the third coefficient is 0.03.

[0082] If the noise σ of the image to be corrected is greater than 10DN, the dynamic coefficient is determined to be the first coefficient (0.05), and the preset gradient threshold Thr = 0.05 × MAX. If the noise 5DN ≤ σ ≤ 10DN of the image to be corrected, the dynamic coefficient is determined to be the second coefficient (0.04), and the preset gradient threshold Thr = 0.04 × MAX. If the noise σ of the image to be corrected is less than 5DN, the dynamic coefficient is determined to be the third coefficient (0.03), and the preset gradient threshold Thr = 0.03 × MAX.

[0083] Based on the above scheme, this embodiment adapts to different noise levels and time-varying blind cell scenarios by adaptive threshold adjustment and dynamic updating of the blind cell table, ensuring the stability of the correction effect.

[0084] In a more detailed embodiment, the specific execution steps of the infrared image processing method provided in this embodiment are as follows: Step 1: Obtain multiple consecutive images under multiple scene configuration parameters.

[0085] Step 2: Data preprocessing. The acquired multi-frame images are preprocessed. The temporal noise criterion directly uses the original multi-frame data, while the response rate criterion, spatial grayscale difference criterion, linearity criterion, and two-point gain coefficient criterion use the average data from multiple frames (10 frames for dark and bright fields, and 20 frames for the mid-field field) to reduce random noise interference.

[0086] Step 3: Perform single-criteria blind cell labeling. According to the judgment rules of response rate criterion, time domain noise criterion, spatial gray level difference criterion, linearity and two-point gain coefficient criterion, label the blind cells under each criterion and record the coordinates of the blind cells.

[0087] Step 4: Take the union of the four criteria labeling results and remove the blind cells with duplicate labels; traverse all labeled blind cells and identify 1-3 blind cells that are horizontally continuous (coordinates differ by 1 column) or vertically continuous (coordinates differ by 1 row) to determine the number of continuous blind cells and their complete coordinates.

[0088] Step 5: Organize the location information (coordinates), number of consecutive blind cells, and consecutive directions (horizontal / vertical) of the blind cells into a target blind cell table, and store it in a local or server database for use in subsequent calibration steps.

[0089] Step 6: Based on the coordinates of the blind cell to be corrected, determine whether it is located at the row edge (first two rows or last two rows) or column edge (first three columns or last three columns). If yes, proceed to step 7; otherwise, proceed to step 8.

[0090] Step 7: If it is a row edge, copy the adjacent row data using the mirror expansion method to generate a virtual row to supplement the 5×5 neighborhood; if it is a column edge, record the position of the nearest normal pixel to prepare for subsequent column copying.

[0091] Step 8: Based on the target blind cell table, determine whether the blind cell to be corrected is a single blind cell, two consecutive blind cells, or three consecutive blind cells; if it is a single or two consecutive blind cells, proceed to step 9; if it is three consecutive blind cells, proceed to step 10.

[0092] Step 9: Perform general gradient calculation and correction. Calculate three basic gradients within a 5×5 neighborhood, filter valid gradients and determine the direction of the minimum gradient, calculate replacement values, and update blind pixel values; if it is a column edge, for blind pixel columns that are not the nearest normal pixels, copy the results of the corrected blind pixel columns.

[0093] Step 10: Perform general gradient and supplementary gradient calculation and correction. First, calculate 3 basic gradients, then calculate 2 supplementary gradients (vertical supplementary gradient is added for vertically consecutive three-block blind pixels), filter all valid gradients and determine the direction of the minimum gradient, calculate the replacement value, and update the blind pixel value; if it is an edge region, complete the calculation by combining the extended data from Step 7.

[0094] Step 11: Store the correction result of the current row in an independent buffer. It will not participate in the neighborhood calculation of the next row. The neighborhood calculation of the next row will only use the original data of the current row to ensure the real-time performance of the pipelined processing.

[0095] Step 12: After completing the correction of all blind pixels to be corrected, output the target image, which can be stored in the data storage system or returned to the terminal device.

[0096] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0097] Based on the same inventive concept, this application also provides an infrared image processing apparatus for implementing the infrared image processing method described above. The solution provided by this apparatus is similar to the implementation described in the above method; therefore, the specific limitations in one or more infrared image processing apparatus embodiments provided below can be found in the limitations of the infrared image processing method described above, and will not be repeated here.

[0098] In one embodiment, such as Figure 9 As shown, an infrared image processing device 900 is provided, including: an acquisition module 910, a filtering module 920, a determination module 930, and a correction module 940, wherein: The acquisition module 910 is used to acquire the image to be corrected, wherein the image to be corrected is a series of consecutive frames of images acquired under multiple scene configuration parameters; the scene configuration parameters include gain parameters, integration time, and scene brightness. The filtering module 920 is used to integrate the recognition results of multiple blind element filtering criteria to generate a target blind element table; wherein, the target blind element table includes the location information of the blind elements; the blind element filtering criteria include at least two criteria from the following: response rate criterion, temporal noise criterion, spatial gray-level difference criterion, linearity and two-point gain coefficient criterion; The determination module 930 is used to determine blind cells to be corrected based on the target blind cells; the blind cells to be corrected include single blind cells and consecutive blind cells. The correction module 940 is used to perform correction processing on the blind pixels to be corrected by adopting a preset dynamic gradient correction strategy, a preset edge adaptation strategy and a pipelined processing adaptation strategy to obtain the target image.

[0099] In one embodiment, the acquisition module 910 is specifically used to acquire dark field images, mid-field images and bright field images under different gains and integration times; wherein, the images under each scene configuration parameter are acquired in 10 or more consecutive frames.

[0100] In one embodiment, the acquisition module 910 is further configured to acquire a bright field image under a first gain and a first integration time; acquire a mid-field image under a second gain and a second integration time; and acquire a dark field image under a third gain and a third integration time.

[0101] In one embodiment, the filtering module 920 is further configured to mark blind elements in the image to be corrected according to different blind element filtering criteria, so as to obtain the recognition results corresponding to different blind element filtering criteria; The response rate criterion is that if the difference between the brightness difference between the bright field image and the dark field image corresponding to a pixel and the average brightness of all pixels is greater than a preset difference threshold, the pixel is marked as a blind pixel. The temporal noise criterion is that if the temporal noise of a pixel is greater than twice the average noise of all pixels, the pixel is marked as a blind pixel. The spatial grayscale difference criterion is that if the difference between the grayscale response value of a pixel and the grayscale mean value in the preset neighborhood of the pixel is greater than a preset difference threshold, the pixel is marked as a blind pixel. The linearity and two-point gain coefficient criteria are as follows: if the linearity of a pixel does not fall within a preset threshold range, or the two-point correction gain coefficient of a pixel does not fall within a preset threshold range, the pixel is marked as a blind pixel.

[0102] In one embodiment, the correction module 940 is further configured to calculate the basic gradient direction value of the blind cell to be corrected; wherein, if the blind cell to be corrected is a continuous blind cell and there are more than or equal to 3 continuous blind cells, the corresponding basic gradient direction value and supplementary gradient direction value are calculated. The pixel value of the blind cell to be corrected is replaced by the average pixel value in the direction corresponding to the minimum effective gradient direction value; wherein, the effective gradient direction value is the gradient direction value that is less than a preset gradient threshold, and the preset gradient threshold is the product of the maximum number of bits in the image and the dynamic coefficient.

[0103] In one embodiment, the correction module 940 is further configured to determine the dynamic coefficient as a second coefficient if the noise of the image to be corrected is less than or equal to the first noise threshold and greater than or equal to the second noise threshold; the second coefficient is less than the first coefficient. If the noise of the image to be corrected is less than the second noise threshold, the dynamic coefficient is determined to be the third coefficient; the third coefficient is less than the second coefficient.

[0104] In summary, this embodiment provides an infrared image processing device that supplements gradient calculation for triple blind pixels and combines edge adaptation strategies such as mirror expansion and column replication to solve the jaggedness and blurring problems in edge regions such as the bevel of the silicon wafer. After correction, the grayscale difference between the edge region of the triple blind pixel and the surrounding pixels is effectively reduced. By acquiring images with multi-scene configuration parameters and fusing four types of blind pixel screening criteria, noise and true blind pixels are effectively distinguished, resulting in high blind pixel detection accuracy and reducing missed detections (such as flash pixels) and misjudgments (such as normal pixels being marked as blind pixels).

[0105] Each module in the aforementioned infrared image processing device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0106] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 10 As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When executed by the processor, the computer program implements an infrared image processing method. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0107] Those skilled in the art will understand that Figure 10The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0108] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps: Acquire an image to be corrected, wherein the image to be corrected is a series of consecutive frames of images acquired under multiple scene configuration parameters; the scene configuration parameters include gain parameters, integration time, and scene brightness; The identification results of multiple blind element screening criteria are integrated to generate a target blind element table; wherein, the target blind element table includes the location information of the blind elements; the blind element screening criteria include at least two criteria from the following: response rate criterion, temporal noise criterion, spatial gray-level difference criterion, linearity and two-point gain coefficient criterion; The blind cells to be corrected are determined based on the target blind cells; the blind cells to be corrected include single blind cells and consecutive blind cells. The target image is obtained by correcting the blind pixels to be corrected using a preset dynamic gradient correction strategy, a preset edge adaptation strategy, and a pipelined processing adaptation strategy.

[0109] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor: Acquire an image to be corrected, wherein the image to be corrected is a series of consecutive frames of images acquired under multiple scene configuration parameters; the scene configuration parameters include gain parameters, integration time, and scene brightness; The identification results of multiple blind element screening criteria are integrated to generate a target blind element table; wherein, the target blind element table includes the location information of the blind elements; the blind element screening criteria include at least two criteria from the following: response rate criterion, temporal noise criterion, spatial gray-level difference criterion, linearity and two-point gain coefficient criterion; The blind cells to be corrected are determined based on the target blind cells; the blind cells to be corrected include single blind cells and consecutive blind cells. The target image is obtained by correcting the blind pixels to be corrected using a preset dynamic gradient correction strategy, a preset edge adaptation strategy, and a pipelined processing adaptation strategy.

[0110] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps: Acquire an image to be corrected, wherein the image to be corrected is a series of consecutive frames of images acquired under multiple scene configuration parameters; the scene configuration parameters include gain parameters, integration time, and scene brightness; The identification results of multiple blind element screening criteria are integrated to generate a target blind element table; wherein, the target blind element table includes the location information of the blind elements; the blind element screening criteria include at least two criteria from the following: response rate criterion, temporal noise criterion, spatial gray-level difference criterion, linearity and two-point gain coefficient criterion; The blind cells to be corrected are determined based on the target blind cells; the blind cells to be corrected include single blind cells and consecutive blind cells. The target image is obtained by correcting the blind pixels to be corrected using a preset dynamic gradient correction strategy, a preset edge adaptation strategy, and a pipelined processing adaptation strategy.

[0111] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0112] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0113] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0114] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. An infrared image processing method, characterized in that, The method includes: Acquire an image to be corrected, wherein the image to be corrected is a series of consecutive frames of images acquired under multiple scene configuration parameters; the scene configuration parameters include gain parameters, integration time, and scene brightness; The identification results of multiple blind element screening criteria are integrated to generate a target blind element table; wherein, the target blind element table includes the location information of the blind elements; the blind element screening criteria include at least two criteria from the following: response rate criterion, temporal noise criterion, spatial gray-level difference criterion, linearity and two-point gain coefficient criterion; The blind cells to be corrected are determined based on the target blind cells; the blind cells to be corrected include single blind cells and consecutive blind cells. The target image is obtained by correcting the blind pixels to be corrected using a preset dynamic gradient correction strategy, a preset edge adaptation strategy, and a pipelined processing adaptation strategy.

2. The method according to claim 1, characterized in that, The gain parameters include a first gain, a second gain, and a third gain; the integration time includes a first integration time, a second integration time, and a third integration time; and the scene brightness includes dark field brightness, mid-field brightness, and bright field brightness. Specifically, the first gain is less than the second gain, the second gain is less than the third gain, the first integration time is less than the second integration time, the second integration time is less than the third integration time, the dark field brightness is less than the mid-field brightness, and the mid-field brightness is less than the bright field brightness. The acquisition of the image to be corrected includes: Acquire dark field, mid field, and bright field images under different gains and integration times; for each scene configuration parameter, acquire 10 or more consecutive images.

3. The method according to claim 2, characterized in that, The acquisition of the image to be corrected includes: Acquire the bright-field image at the first gain and the first integration time; Obtain the midfield image under the second gain and the second integral time; Obtain the dark field image at the third gain and the third integration time.

4. The method according to claim 1, characterized in that, Before generating the target blind element table by fusing the recognition results of multiple blind element screening criteria, the method further includes: Blind elements in the image to be corrected are labeled according to different blind element screening criteria to obtain the recognition results corresponding to different blind element screening criteria; The response rate criterion is as follows: if the difference between the brightness difference between the bright field image and the dark field image corresponding to a pixel and the average brightness of all pixels is greater than a preset difference threshold, the pixel is marked as a blind pixel. The temporal noise criterion is: if the temporal noise of a pixel is greater than twice the average noise of all pixels, the pixel is marked as a blind pixel; The spatial grayscale difference criterion is: if the difference between the grayscale response value of a pixel and the grayscale mean value in the preset neighborhood of the pixel is greater than the preset difference threshold, the pixel is marked as a blind pixel; The linearity and two-point gain coefficient criteria are as follows: if the linearity of a pixel does not fall within the preset threshold range, or the two-point correction gain coefficient of a pixel does not fall within the preset threshold range, the pixel is marked as a blind pixel.

5. The method according to claim 1, characterized in that, The preset edge adaptation strategy includes a row edge processing strategy and a column edge processing strategy. The row edge processing strategy uses a mirror expansion method to copy adjacent row data to expand the edge row data. The column edge processing strategy uses a column copy method to copy the correction results of blind columns to correct the edge column data. The pipelined processing adaptation strategy uses the correction results of the nth row to store in an independent buffer, and the neighborhood calculation of the (n+1)th row uses the original data of the nth row for calculation. The step involves employing a preset dynamic gradient correction strategy, a preset edge adaptation strategy, and a pipelined processing adaptation strategy to correct the blind pixels to be corrected, thereby obtaining the target image. This includes: Calculate the basic gradient direction value of the blind cell to be corrected; wherein, if the blind cell to be corrected is a continuous blind cell, and there are more than or equal to 3 continuous blind cells, calculate the corresponding basic gradient direction value and supplementary gradient direction value. The pixel value of the blind cell to be corrected is replaced by the average pixel value in the direction corresponding to the minimum effective gradient direction value; wherein, the effective gradient direction value is the gradient direction value that is less than a preset gradient threshold, and the preset gradient threshold is the product of the maximum number of bits in the image and the dynamic coefficient.

6. The method according to claim 5, characterized in that, The method further includes: If the noise of the image to be corrected is greater than a first noise threshold, the dynamic coefficient is determined to be the first coefficient; If the noise of the image to be corrected is less than or equal to the first noise threshold and greater than or equal to the second noise threshold, the dynamic coefficient is determined to be the second coefficient; the second coefficient is less than the first coefficient. If the noise of the image to be corrected is less than the second noise threshold, the dynamic coefficient is determined to be the third coefficient; the third coefficient is less than the second coefficient.

7. An infrared image processing device, characterized in that, The device includes: An acquisition module is used to acquire an image to be corrected, wherein the image to be corrected is a series of consecutive frames of images acquired under multiple scene configuration parameters; the scene configuration parameters include gain parameters, integration time, and scene brightness. A filtering module is used to integrate the recognition results of multiple blind element filtering criteria to generate a target blind element table; wherein, the target blind element table includes the location information of the blind elements; the blind element filtering criteria include at least two criteria from the following: response rate criterion, temporal noise criterion, spatial gray-level difference criterion, linearity and two-point gain coefficient criterion; A determination module is used to determine blind cells to be corrected based on the target blind cells; the blind cells to be corrected include single blind cells and consecutive blind cells. The correction module is used to perform correction processing on the blind pixels to be corrected using a preset dynamic gradient correction strategy, a preset edge adaptation strategy, and a pipelined processing adaptation strategy to obtain the target image.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the infrared image processing method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the infrared image processing method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the infrared image processing method according to any one of claims 1 to 6.

Images