Image processing method and device for infrared imaging and storage medium

Abstract

The invention discloses an image processing method and device for infrared imaging and a storage medium, and belongs to the technical field of image processing. The method comprises the steps that image feature separation is carried out on an input image through filtering processing of an edge preserving guiding filter, a base layer image and a detail residual image are obtained, the detail residual image comprises detail content and stripe content of the input image, one-dimensional row guiding filtering is carried out on the detail residual image, and the one-dimensional row guiding filtering is carried out on the detail residual image. The method comprises the following steps of: performing one-dimensional row guide filtering on a base layer image, obtaining a detail residual error, obtaining a one-dimensional row guide filtering output result and the detail residual error to obtain a row filtering high-frequency image, performing one-dimensional column guide filtering by taking the row filtering high-frequency image as an input image and taking the base layer image as a guide image to obtain a fringe layer image, and obtaining a fringe-removed image of the input image by obtaining a difference between the fringe layer image and the base layer image. According to the method, random noise, stripe noise and scene details are subjected to hierarchical processing and layer-by-layer suppression, so that the texture integrity of the infrared image is kept while overall noise reduction is performed, and the imaging effect of the infrared image is improved.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and in particular to an image processing method, device and storage medium for infrared imaging. Background Technology

[0002] To reduce image noise caused by detector inhomogeneity during infrared imaging, infrared imaging typically employs a process of non-uniformity correction, secondary noise reduction, and image enhancement to remove image noise and generate an infrared image for display.

[0003] However, in related technologies, infrared image processing typically performs two-point or multi-point non-uniformity correction on the raw infrared data, and then directly uses mean filtering, bilateral filtering, or anisotropic diffusion algorithms for overall smoothing to suppress random noise and fixed pattern noise, outputting an enhanced infrared image. While suppressing noise, this process blurs scene details and stripe noise, resulting in the loss of texture information in the infrared image.

[0004] The above content is only used to help understand the technical solution of this application and does not represent an admission that the above content is prior art. Summary of the Invention

[0005] The main objective of this application is to provide an image processing method, device, and storage medium for infrared imaging, aiming to solve the technical problem of lost texture information in infrared images.

[0006] To achieve the above objectives, this application provides an image processing method for infrared imaging, the method comprising the following steps: The input image is processed by an edge-preserving guided filter to separate image features, resulting in a base layer image and a detail residual image. The detail residual image contains the detail content and stripe content of the input image. A one-dimensional row guided filter is performed on the detail residual image, and the difference between the output of the one-dimensional row guided filter and the detail residual is calculated to obtain the row filtered high-frequency image. Using the row-filtered high-frequency image as the input image and the base layer image as the guide image, a one-dimensional column-guided filter is performed to obtain the stripe layer image; The difference between the stripe layer image and the base layer image is calculated to obtain the stripe-free image of the input image.

[0007] In one embodiment, the step of performing image feature separation on the input image through filtering to obtain the base layer image and the detail residual image includes: Perform Gaussian filtering on the input image to obtain a smoothed image; The smoothed image is subjected to edge-preserving guided filtering to generate the base layer image; The difference between the smoothed image and the base layer is calculated to obtain the detail residual image; In one embodiment, after the step of subtracting the stripe layer image from the base layer image to obtain the destriped image of the input image, the method further includes: The low-frequency and high-frequency components of the destriped image are separated by a low-pass filter to generate a background image and a detail image. By adjusting the dynamic range, noise reduction and enhancement are performed on the background image to obtain the target background image; And / or, perform image enhancement on the detail image to obtain a detail layer image; The detail layer image and the target background image are combined to generate a display image.

[0008] In one embodiment, before the step of separating image features from the input image through filtering to obtain the base layer image and the detail residual image, the method further includes: Raw infrared data is collected using an infrared detector; From the non-uniformity correction table pre-stored in the infrared detector, read the gain coefficient and bias value corresponding to each pixel in the original infrared data; The original pixel value of the pixel is multiplied by the corresponding gain coefficient to obtain the intermediate pixel value; The intermediate pixel value is added to the corresponding bias value to obtain the corrected pixel value; The corrected pixel values ​​are arranged based on the original pixel positions to form the input image.

[0009] In one embodiment, the step of subtracting the stripe layer image from the base layer image to obtain the destriped image of the input image includes: Align the pixel coordinates of the stripe layer image with the base layer image to determine the target pixel point of the pixel point in the stripe layer image that has the same pixel coordinates as the pixel point in the base layer image. Calculate the pixel difference between the stated pixel and the target pixel; Based on the pixel coordinates, the pixel differences are used to reconstruct the image, resulting in the striped image.

[0010] In one embodiment, after the step of subtracting the stripe image from the base layer image to obtain the destriped image of the input image, the method further includes: Edge detection is performed on the destriped image to obtain an edge map; The high-frequency region of the destriped image is determined based on the edge map, and super-resolution reconstruction is performed in the high-frequency region to obtain a super-resolution image; Output the super-resolution image.

[0011] In addition, to achieve the above objectives, this application also provides an infrared imaging image processing device, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the infrared imaging image processing method as described above.

[0012] In addition, to achieve the above objectives, this application also provides a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the infrared imaging image processing method as described above.

[0013] One or more technical solutions proposed in this application have at least the following technical effects: This application obtains an input image by performing non-uniformity correction on infrared images acquired by an infrared detector, thereby reducing the inherent non-uniformity of the detector. A guided filter is then applied to the input image to generate a base layer image, filtering out random noise while preserving scene edges and texture. Subsequently, based on this base layer image, stripe separation is performed on the input image to obtain a detail residual image, allowing stripe noise, originally mixed with details, to be extracted independently. A statistical stripe algorithm is then used to denoise the detail residual image, obtaining a residual layer image. This denoising only targets the stripe component without affecting scene details. Finally, the residual layer image is merged into the base layer image to obtain a de-striped image. Without expanding the smoothing range or sacrificing texture information, random noise, stripe noise, and scene details are processed and suppressed layer by layer, thus maintaining the texture integrity of the infrared image while achieving overall noise reduction, overcoming the detail loss defect caused by overall smoothing in existing technologies. Attached Figure Description

[0014] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating the first embodiment of the infrared imaging image processing method of this application; Figure 2 This is a schematic diagram of the infrared image stripe removal process involved in the first embodiment of this application; Figure 3 This is a schematic diagram of the input image involved in the first embodiment of this application; Figure 4 This is a schematic diagram of a striped image related to the first embodiment of this application; Figure 5 This is a flowchart illustrating the second embodiment of the infrared imaging image processing method of this application; Figure 6 This is a flowchart illustrating the third embodiment of the infrared imaging image processing method of this application; Figure 7 This is a flowchart illustrating the fourth embodiment of the infrared imaging image processing method of this application; Figure 8 This is a flowchart illustrating the fifth embodiment of the infrared imaging image processing method of this application; Figure 9 This is a schematic diagram of the structure of an infrared imaging image processing device in the hardware operating environment involved in the embodiments of this application.

[0017] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0018] It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.

[0019] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0020] The main solution of this application embodiment is as follows: through filtering, image feature separation is performed on the input image to obtain a base layer image and a detail residual image. The detail residual image contains the detail content and stripe content of the input image. One-dimensional row guided filtering is performed on the detail residual image, and the output result of the one-dimensional row guided filtering is subtracted from the detail residual to obtain a row filtered high-frequency image. Using the row filtered high-frequency image as the input image and the base layer image as the guided image, one-dimensional column guided filtering is performed to obtain a stripe layer image. The difference between the stripe layer image and the base layer image is calculated to obtain the destriped image of the input image.

[0021] Existing technologies remove image noise through a process of non-uniformity correction, secondary noise reduction, and image enhancement to generate infrared images for display. However, after performing two-point or multi-point non-uniformity correction on the raw infrared data, infrared image processing typically uses mean filtering, bilateral filtering, or anisotropic diffusion algorithms for overall smoothing to suppress random noise and fixed pattern noise, outputting an enhanced infrared image. While suppressing noise, this process blurs scene details and stripe noise, resulting in the loss of texture information in the infrared image.

[0022] This application employs filtering to separate image features from the input image, obtaining a base layer image and a detail residual image containing details and stripe content of the input image. A one-dimensional row-guided filtering is then performed on the detail residual image, and the difference between the output of the one-dimensional row-guided filtering and the detail residual is calculated to obtain a row-filtered high-frequency image. Furthermore, using the row-filtered high-frequency image as the input image and the base layer image as the guide image, a one-dimensional column-guided filtering is performed to obtain a stripe layer image. The difference between the stripe layer image and the base layer image is calculated to obtain a destriped image of the input image. This process, without expanding the smoothing range or sacrificing texture information, layers random noise, stripe noise, and scene details, and suppresses them layer by layer. This achieves overall noise reduction while maintaining the texture integrity of the infrared image, overcoming the detail loss defects caused by overall smoothing in existing technologies.

[0023] To better understand the above technical solutions, exemplary embodiments of this application will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of this application are shown in the drawings, it should be understood that this application can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of this application and to fully convey the scope of this application to those skilled in the art.

[0024] It should be noted that the executing entity in this embodiment can be an infrared imaging system, or a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device or infrared imaging image processing device capable of the above functions. This embodiment does not specifically limit the specific implementation. The following uses an infrared imaging system as an example to describe this embodiment and the following embodiments.

[0025] Based on this, embodiments of this application provide an image processing method for infrared imaging, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the infrared imaging image processing method of this application.

[0026] In this embodiment, the infrared imaging image processing method includes steps S10 to S40: Step S10: The input image is separated by filtering with an edge-preserving guided filter to obtain the base layer image and the detail residual image. In this embodiment, image feature separation utilizes digital filtering technology to decouple the image signal into different components based on spatial frequency differences. The original input image is decomposed into a low-frequency base layer containing the main scene structure and a residual layer containing all high-frequency oscillation information. The detail residual image is obtained by removing the base layer from the original image or its intermediate representation, concentrating all sharp variations in the image, including object edges, subtle surface textures, random noise, and structural stripe noise distributed in a stripe pattern. The base layer image mainly contains large smooth areas and gradually varying background radiation information; its texture is simple, and its grayscale changes are gradual.

[0027] Specifically, the infrared imaging system achieves feature separation through a two-stage filtering differential strategy. The first stage uses Gaussian filtering to preprocess the input image. Gaussian filtering is a linear smoothing filter whose convolution kernel weights follow a two-dimensional Gaussian distribution. By performing a weighted average calculation on the neighborhood of each pixel in the input image, it suppresses high-frequency random noise and fine textures in the image, outputting an overall smooth intermediate image with slightly blurred edges. The second stage uses this intermediate image as the processing object, applying edge-preserving guided filtering. Guided filtering is a non-linear filter that, under the constraint of a specified guide image, smooths the input image while preserving significant edge structures in the guide image.

[0028] Optionally, the infrared imaging system uses an intermediate image generated by Gaussian filtering as a guide image. This allows the filtering process to smooth uniform regions within the image while simultaneously enhancing and restoring the main object contours weakened by Gaussian filtering. The output after the guide filtering is the desired base layer image, which has uniform internal grayscale and sharp edges. The system performs pixel-by-pixel algebraic subtraction between the first-stage Gaussian-smooth intermediate image and the second-stage base layer image. Since the base layer image is smoother than the Gaussian-smooth intermediate image and has sharper edges, the difference between the two accurately captures all high-frequency components progressively removed by the two-stage smoothing process, forming a detailed residual image.

[0029] As an alternative implementation, the feature separation process can employ parameterized adaptive filtering. In the Gaussian filtering stage, the kernel size and standard deviation are adaptively selected based on the global noise estimate of the input image. For images with high noise levels, a larger standard deviation and kernel size are used to achieve sufficient smoothing; for images with low noise, smaller parameters are used to avoid excessive blurring. In the edge-preserving guided filtering stage, the filtering radius and regularization parameters are not fixed but dynamically adjusted based on the local gradient information of the Gaussian-smoothed intermediate image. Larger radii and stronger regularization are used in regions with gentle gradients to obtain an extremely smooth base layer; in regions with significant gradients, the radius is automatically reduced and the regularization intensity is decreased to accurately preserve edge shapes. The difference calculation uses high-precision floating-point operations and introduces pixel-level weight mapping. For known important detail regions in the image, the weights are appropriately reduced when calculating the difference to lessen the intensity of detail transfer to the residual layer, thereby retaining more key information in the base layer.

[0030] Specifically, such as Figure 2 As shown, Figure 2 This is a schematic diagram of the infrared image stripe removal process involved in this embodiment. The infrared imaging system first performs Gaussian filtering on the input image to obtain a smooth image, and then performs edge-preserving guided filtering on the smooth image to generate a base layer image. Further, the difference between the smooth image and the base layer is calculated to obtain the detail residual image, which contains a mixture of image details and feature information such as stripe noise.

[0031] It's important to note that guided filtering is an image filtering technique in computer vision based on guided maps. It achieves consistency between texture details and the guided map through a local linear model, preserving the features of the input image while inheriting the texture structure of the guided map. Its output is controlled by the content of the guided map, supports both single-channel and multi-channel image processing, and features a fast algorithm independent of the filtering radius. Specifically, when the guided map and input image are identical, this technique can achieve edge-preserving smoothing effects, similar to bilateral filtering but with superior edge preservation. When applied to different images, it can be used for image matting, detail enhancement, and other scenarios.

[0032] In this embodiment, the infrared imaging system guides the image to undergo local linear filtering by using the edge information of the image as a constraint, thereby obtaining a base layer image. This allows the base layer image to completely preserve scene texture and edge details while removing random noise.

[0033] Specifically, the infrared imaging system acquires a guide image based on the input image and then synchronously slides the filter window across both the guide and input images using a preset filter window. Within the filter window, the local mean of the input image and the target local mean and target local variance of the guide image are calculated. Based on the target local mean, target local variance, and preset regularization parameters, the linear coefficients corresponding to the filter window are calculated. Based on these linear coefficients, a linear transformation is performed on the local mean of the input image to obtain the smoothed result of the center pixel of the filter window. After completing the traversal, the smoothed results are integrated based on the position of the center pixel relative to the pixel in the input image to form the base layer image.

[0034] As an alternative implementation, the infrared imaging system can directly use the input image as a guide image simultaneously. First, a copy of the input image is made to serve as the guide image. Then, a fixed-size filtering window is synchronously slid across both the guide image and the input image. Inside the window, the processor calculates the target local mean and target local variance of the guide image, and also calculates the local mean of the input image. These target local mean, target local variance, and local mean are used to construct a local linear model. Subsequently, the infrared imaging system solves for linear coefficients based on the local variance and preset regularization parameters. These linear coefficients determine the weight distribution of the center pixels within the window. The processor then uses these linear coefficients to perform a weighted linear combination of the local means of the input image to obtain a smoothed result for the center pixels. After window traversal is complete, all center pixels are integrated according to their original coordinates to form the base layer image. Alternatively, the infrared imaging system can first generate a structure map based on the input image, for example, an image after performing a small-scale Gaussian or median filter, and then use this structure map as the guide image to further suppress texture misjudgments.

[0035] In this embodiment, the infrared imaging system aligns the coordinates of pixels in the base layer image with those in the input image to determine the target pixel in the input image that shares the same coordinates with the target pixel in the base layer image. By calculating the pixel difference between the target pixel and the input pixel, and based on the pixel coordinates, image reconstruction is performed on the pixel difference to obtain a detail residual image.

[0036] Optionally, the infrared imaging system establishes a spatial mapping table with the same resolution as the input image to record pixel coordinates. It then performs a first action row-by-row and column-by-column: locating the current pixel coordinates in the mapping table and reading the pixel value at the same coordinates from the base layer image. Next, it performs a second action: subtracting the pixel value of the base layer image from the pixel value at the corresponding coordinates in the input image to obtain the difference. All these differences are then written to a new frame buffer according to their original coordinates, forming a detail residual image.

[0037] Step S20: Perform one-dimensional row guided filtering on the detail residual image, and calculate the difference between the output of the one-dimensional row guided filtering and the detail residual to obtain the row filtered high-frequency image; In this embodiment, the infrared imaging system initially separates and enhances components with vertical continuity from the mixed detail residuals. It should be noted that one-dimensional row-guided filtering is a special form of guided filtering. Its filtering window has a non-unit length only in the horizontal direction and a length of 1 in the vertical direction. That is, the filtering operation is performed independently on each row of image data, utilizing only the local statistical relationships of pixels within the row. The row-filtered high-frequency image refers to the image obtained by subtracting the filtering result from the original detail residual after horizontal smoothing filtering. This image mainly retains components with gentle or regular gray-level changes in the horizontal direction but potentially abrupt changes in the vertical direction, which are typical characteristics of vertical stripes and steep vertical edges.

[0038] Specifically, the infrared imaging system performs one-dimensional row-guided filtering on the detail residual image. The detail residual image serves as both the input and guide image for the filter. The filter slides a window with only a horizontal width along each row of the image. Within the window, the algorithm builds a local linear model based on the guide image, estimating model parameters by minimizing the difference between the input and output. Since the window has no vertical extension, data from different rows are completely independent of each other's calculations. This operation effectively smooths irregular gray-level jumps and subtle fluctuations within each row, such as isolated noise points or short horizontal texture fragments, but its smoothing effect is limited for vertical features that exhibit continuous or periodic variations within a row. The second operation is image differencing. The original detail residual image is subtracted pixel-by-pixel from the image processed by the one-dimensional row-guided filtering. The differencing operation removes image components that can be predicted and fitted by the horizontal linear model, i.e., those features that can be smoothed within the row. Therefore, the difference results significantly retain those components that appear as residuals or unpredictable in the row direction filtering model, corresponding to the vertical structures that are highly consistent or gradually changing in the horizontal direction and are not smoothed out by the filter, forming the final row filter high-frequency map.

[0039] Step S30: Using the row-filtered high-frequency image as the input image and the base layer image as the guide image, perform one-dimensional column-guided filtering to obtain the stripe layer image; Step S40: Calculate the difference between the stripe layer image and the base layer image to obtain the destriped image of the input image.

[0040] In this embodiment, column-guided filtering is similar in principle to row-guided filtering but orthogonal in direction. Its filtering window has a non-unit length only in the vertical direction, enabling independent processing of each column of data. The stripe layer image is the additional noise field estimated by the algorithm, and ideally should only contain the stripe pattern. The destriped image is the output image obtained by subtracting the estimated stripe layer from the base layer image representing the main scene, thus removing the stripe interference.

[0041] It should be noted that the infrared imaging system achieves precise directional separation and extraction of stripe noise through two orthogonal and complementary filtering operations. Row filtering, smoothing along the horizontal direction, aims to remove random noise and horizontal textures from the detail residual image. The difference between its output and the original image is equivalent to performing vertical feature enhancement, concentrating the subsequent processed signals primarily into vertical stripes and edges. The infrared imaging system uses this image rich in vertical features as input and introduces a stripe-free base layer as guidance for column filtering. This vertical filtering operation, under the constraint of the guide image, distinguishes the vertically continuous components in the input into uniform stripes independent of scene structure and true vertical features related to scene edges, thereby accurately estimating the pure stripe noise layer. This ensures maximum simplification of the problem background and highlights target features, while achieving high-precision separation of noise and content based on cleaner input and more discriminative guidance.

[0042] Specifically, the guided filtering operation uses the row-filtered high-frequency image as the input image and the base layer image as the guide image, performing one-dimensional column-guided filtering. The base layer image contains spatial structural information of the scene but is largely free of high-frequency noise and stripes. In this configuration, the guided filtering algorithm operates along each column of the image. In regions with uniform grayscale in the base layer image, the algorithm applies strong vertical smoothing constraints, forcing the vertical features of the corresponding column in the input row-filtered high-frequency image to be smoothed into a uniform grayscale value, which is then extracted into the output. In regions with vertical edges in the base layer image, the edge-preserving property of the guided filtering is activated, and the algorithm limits the smoothing intensity near that location, thus helping to prevent parts of the input image that coincide with the vertical edges of real objects from being incorrectly smoothed into stripes. The output of this operation is the estimated stripe layer image. The image synthesis operation involves algebraically subtracting the base layer image from the stripe layer image. The base layer image is the best low-frequency approximation of the scene content but may contain low-frequency gradient components of stripes. Subtracting the estimated stripe layer image is equivalent to removing stripe components identified as noise from the scene content. Because the base layer itself has good edge preservation properties, this subtraction operation can effectively suppress stripes while preserving the true structure and details of the image to the greatest extent.

[0043] For example, refer to Figure 2The infrared imaging system first performs Gaussian filtering on the input image. This linear smoothing process effectively suppresses high-frequency random noise in the image through Gaussian kernel convolution, generating an overall smooth intermediate image. Using this intermediate image as input, and itself as a guide image, edge-preserving guided filtering is performed. This non-linear filtering process smooths the internal regions of the image while preserving and enhancing the main edge structures inherent in the guide image. The output is a base layer image carrying the outline of the main scene and low-frequency information. Pixel-level differences are calculated between the initial Gaussian-filtered intermediate image and this base layer image, resulting in the detail and stripe layers containing all high-frequency oscillation information. One-dimensional row guided filtering is then performed on this detail and stripe layer. This filter's operation window extends only horizontally, processing each row of pixel data independently. It effectively smooths intra-row gray-level abrupt changes and suppresses horizontal features, producing a row-filtered output image. The difference between the detail and stripe layer and the row-filtered output image is calculated. This difference operation removes components that can be fitted by a horizontal linear model, resulting in a row-filtered high-frequency image that primarily highlights continuous vertical features.

[0044] Furthermore, the infrared imaging system uses the base layer image generated by edge-preserving guided filtering as the guide image and the row-filtered high-frequency image as the input image to be processed, performing one-dimensional column-guided filtering. The computational window of this filter extends only in the vertical direction and is constrained by the structure of the guide image. It applies strong column-direction smoothing to the input image in the uniform gray-level regions of the base layer, thereby extracting and homogenizing the continuous vertical stripes in the row-filtered high-frequency image. The output result is the column-filtered output image, i.e., the estimated stripe noise layer. The difference between the base layer image and this column-filtered output image is calculated to subtract the estimated stripe components from the scene content, generating a destriped image. Here, the reference... Figure 3 and Figure 4 , Figure 3 This is a schematic diagram of the input image involved in this embodiment. It represents the image to be processed by the infrared imaging system before stripe removal. Figure 4 This is a schematic diagram of the de-striped image involved in this embodiment. Figure 3 The image shows the imaging effect after the stripe removal process.

[0045] This embodiment obtains an input image by performing non-uniformity correction on the infrared image acquired by the infrared detector, thereby reducing the inherent non-uniformity of the detector. A guided filter is then applied to the input image to generate a base layer image, filtering out random noise while preserving scene edges and texture. Subsequently, based on this base layer image, stripe separation is performed on the input image to obtain a detail residual image, allowing stripe noise, originally mixed with details, to be extracted independently. A statistical stripe algorithm is then used to denoise the detail residual image, obtaining a residual layer image. This suppresses only the stripe component without affecting scene details. Finally, the residual layer image is merged into the base layer image to obtain a de-striped image. Without expanding the smoothing range or sacrificing texture information, random noise, stripe noise, and scene details are processed and suppressed layer by layer, thus maintaining the texture integrity of the infrared image while achieving overall noise reduction. This overcomes the detail loss defect caused by overall smoothing in existing technologies.

[0046] Based on the same inventive concept, this application also provides a second embodiment, referring to... Figure 5 , Figure 5 This is a flowchart illustrating the second embodiment of the infrared imaging image processing method of this application.

[0047] In this embodiment, the infrared imaging image processing method further includes steps S10 to S60: Step S10: Through filtering, the input image is subjected to image feature separation to obtain the base layer image and the detail residual image; Step S50: Perform denoising on the detail residual image using the statistical stripe algorithm to obtain the residual layer image; In this embodiment, the infrared imaging system, after separating the stripe information in the input image to obtain a detailed residual image, uses a statistical stripe algorithm to statistically model the pixel values ​​along the stripe direction and estimate the noise amplitude, thereby obtaining the stripe-removed residual data, i.e., the residual layer image. Specifically, the statistical stripe algorithm constructs an estimation model that matches the fixed pattern noise of the infrared detector by calculating statistics in the row or column direction, thus separating the estimated values ​​from the original data and achieving precise suppression of stripe noise.

[0048] Specifically, the infrared imaging system determines the image traversal direction based on the fringe information of the detail residual image, where the fringe information represents potential fringe noise in the detail residual image. The infrared imaging system traverses the fringe information of the detail residual image based on the image traversal direction to obtain the grayscale distribution characteristics of the fringe information. Combining this with a pre-defined noise estimation model, and based on these grayscale distribution characteristics, and / or information such as average pixel value and normal distribution, the system calculates the noise estimate value for the corresponding fringe pixel. Subtracting the fringe pixel value from the corresponding noise estimate value yields the residual layer image.

[0049] In one example, the infrared imaging system analyzes the spatial texture orientation of the detail residual image to determine whether the stripes are mainly distributed along the horizontal row direction or the vertical column direction, and sets the traversal direction to row-by-row or column-by-column mode accordingly. Specifically, when the stripes are along the horizontal direction, the traversal direction is row-by-row; when the stripes are distributed along the vertical direction, the traversal direction is column-by-column.

[0050] As an alternative implementation, the noise estimation model can be constructed using a neural network based on historical denoised data from the infrared detector. Several sets of labeled infrared detail residual images and their corresponding noise-free residual layer images of the infrared detector or similar models are collected as training samples. End-to-end training is performed using a convolutional neural network or Transformer architecture, with the loss function being the minimum mean square error between predicted and actual noise. This yields lightweight noise estimation network weights that can be deployed in infrared imaging systems.

[0051] As an alternative implementation, the infrared imaging system can also calculate the mean of each row or column of the image to obtain column mean data or column average data. Taking column mean data as an example, the average of all column mean data is calculated, and the average value is subtracted from the column mean data to obtain the corrected data. The calculation process for the corrected data of the row mean data is similar. The infrared imaging system subtracts the corrected data point by point from each row of pixels in the image to obtain a stripe-reduced image.

[0052] Step S60: Merge the residual layer image into the base layer image to obtain a striped image.

[0053] In this embodiment, merging refers to adding two spatially aligned images pixel by pixel, fusing the information they carry into a single output frame. The infrared imaging system aligns the base layer image and the residual layer image based on the same coordinate system, and adds the pixel values ​​of the residual layer image pixels to the pixel values ​​of the target pixels at the same coordinates in the base layer image one by one to obtain a striped image.

[0054] Specifically, the infrared imaging system establishes a unified coordinate system and performs coordinate alignment verification between the base layer image and the residual layer image. Then, the processor performs pixel-by-pixel addition, adding the pixel value at a certain coordinate in the residual layer image to the pixel value at the same coordinate in the base layer image to obtain the fused pixel value. For example, if the base layer pixel value is 148 and the residual layer value is -0.8, then the pixel value at that coordinate in the de-striated image will be 147.2. The system writes all fused pixel values ​​to the output frame buffer according to the original coordinates, forming the de-striated image. In this process, the scene details and textures of the infrared data acquired by the infrared detector are fully preserved, and stripe noise is completely eliminated, resulting in an infrared image with both high signal-to-noise ratio and high definition.

[0055] This application embodiment obtains an input image by performing non-uniformity correction on the infrared image acquired by the infrared detector, thereby reducing the inherent non-uniformity of the detector. A guided filter is then applied to the input image to generate a base layer image, filtering out random noise while preserving scene edges and texture. Subsequently, based on this base layer image, stripe separation is performed on the input image to obtain a detail residual image, allowing stripe noise, originally mixed with details, to be extracted independently. A statistical stripe algorithm is then used to denoise the detail residual image, obtaining a residual layer image. This suppresses only the stripe component without affecting scene details. Finally, the residual layer image is merged into the base layer image to obtain a de-striped image. Without expanding the smoothing range or sacrificing texture information, random noise, stripe noise, and scene details are processed and suppressed layer by layer, thus maintaining the texture integrity of the infrared image while achieving overall noise reduction, overcoming the detail loss defect caused by overall smoothing in existing technologies.

[0056] Based on the same inventive concept, this application also provides a third embodiment, referring to... Figure 6 , Figure 6 This is a flowchart illustrating the third embodiment of the infrared imaging image processing method of this application.

[0057] In this embodiment, the infrared imaging image processing method further includes steps S01 to S05: Step S01: Collect raw infrared data using an infrared detector; Step S02: Read the gain coefficient and bias value corresponding to each pixel in the original infrared data from the non-uniformity correction table pre-stored in the infrared detector; Step S03: Multiply the original pixel value of the pixel by the corresponding gain coefficient to obtain the intermediate pixel value; Step S04: Add the intermediate pixel value to the corresponding bias value to obtain the corrected pixel value; Step S05: Arrange the corrected pixel values ​​based on the original pixel positions to form the input image.

[0058] In this embodiment, non-uniformity correction refers to a digital processing technique that compensates for the response differences of each pixel through gain coefficients and bias values. This technique is based on the fixed-mode noise that causes inconsistent outputs from each pixel under uniform radiation sources. After receiving the raw infrared data acquired by the infrared detector, the infrared imaging system performs preliminary noise reduction on the raw infrared data through non-uniformity correction to reduce the noise impact caused by detector non-uniformity, generating an input image for subsequent further noise reduction and image enhancement of the infrared data.

[0059] As an optional implementation, the infrared imaging system acquires raw infrared data using an infrared detector. From a pre-stored non-uniformity correction table, it reads the gain coefficient and bias value corresponding to each pixel of the raw infrared data. The original pixel value of each pixel is multiplied by its corresponding gain coefficient to obtain an intermediate pixel value. This intermediate pixel value is then added to its corresponding bias value to obtain the corrected pixel value. The corrected pixel values ​​are arranged based on the original pixel positions to form the input image.

[0060] Specifically, the infrared imaging system reads pre-calibrated gain and bias tables from non-volatile memory. The gain table records the scaling factor for each pixel, and the bias table records the zero-position offset for each pixel. The infrared imaging system performs a first action pixel by pixel: multiplying the pixel's value in the original infrared data by the corresponding scaling factor in the gain table to obtain an intermediate pixel value. This multiplication weakens high-response pixels and amplifies low-response pixels. Immediately following, a second action is performed: adding the intermediate pixel value to the corresponding zero-position offset in the bias table. This addition eliminates the fixed DC offset. After all pixels have completed the multiplication and addition operations, the processor rearranges the corrected values ​​according to the original pixel coordinates to generate the input image.

[0061] Optionally, images are pre-acquired at at least two different temperature points using a uniform blackbody or uniform radiation source. For each pixel, a linear or polynomial curve is fitted using the response values ​​from these two or more points. The slope obtained from the fitting is used as the gain coefficient for that pixel, and the intercept or zero-point offset is used as the bias value. All gain coefficients and bias values ​​are stored in the detector's read-only memory or system memory. During actual imaging, the system is powered on or read from the memory at specified intervals and applied to the real-time image to complete pixel-by-pixel gain multiplication and bias addition operations.

[0062] For example, before the infrared detector leaves the factory or during on-site calibration, a uniform blackbody is used to take pictures of the detector at two temperature points. The host computer software defines a linear function relationship for each pixel, where the slope of this linear function is the gain coefficient and the intercept is the bias value. By writing the gain table and bias table into the camera's electrically erasable programmable read-only memory (EEPROM), or by packaging it with the firmware, the tables can be directly read from storage after the infrared detector is powered on, allowing for real-time non-uniformity correction of each frame of image using "pixel value × gain + bias".

[0063] Based on the same inventive concept, this application also provides a fourth embodiment, referring to... Figure 7 , Figure 7 This is a flowchart illustrating the fourth embodiment of the infrared imaging image processing method of this application.

[0064] In this embodiment, the infrared imaging image processing method further includes steps S60-S90: Step S70: Separate the low-frequency and high-frequency components of the destriped image using a low-pass filter to generate a background image and a detail image; Step S80: Perform noise reduction and enhancement on the background image through dynamic range adjustment to obtain the target background image; Step S90: and / or, perform detail enhancement on the detail image to obtain a detail layer image; Step S100: Combine the detail layer image and the target background image to generate the display image.

[0065] In this embodiment, the infrared imaging system uses a low-pass filter to separate the image into a smooth background and a detailed target. The low-pass filter is a digital filter that preserves low frequencies and suppresses high frequencies. The infrared imaging system needs to further reduce noise in the low-frequency image and enhance the high-frequency image to form the final output image.

[0066] It should be noted that while secondary denoising reduces stripes and random noise, it does not address the inherent limitations of the image, such as low contrast, low resolution, and low dynamic range. Furthermore, due to the narrow dynamic range of infrared detectors, the grayscale levels remain concentrated in a narrow range after secondary denoising, resulting in a dark image when directly displayed, making it difficult for the human eye to discern temperature differences. Additionally, physical resolution is limited by pixel size, leaving target edges blurry even after denoising. Therefore, infrared imaging systems require further enhancement of the de-striped image to form the final display image.

[0067] Specifically, the infrared imaging system performs noise reduction and enhancement on the background image and the detail image separately. Contrast adjustment can be used to dynamically transform the background image, allowing for separate noise reduction. Simultaneously, the infrared imaging system can enhance the detail image using image super-resolution techniques or interpolation algorithms. Finally, the enhanced detail image and the noise-reduced background image are merged to obtain the final display image.

[0068] For example, an infrared imaging system can first smooth the detail image again with the same Gaussian kernel to obtain a blurred version, then subtract the blurred version from the original detail image to obtain a detail layer, and then multiply the detail layer by a gain coefficient and superimpose it back onto the original detail image to complete edge enhancement.

[0069] This application embodiment first separates the background and the target by low-pass filtering, then reduces noise in the background and enhances the target respectively. The final synthesized image has no stripe residue and retains clear details, which significantly improves the human eye's observation effect and the performance of the back-end algorithm.

[0070] Since the system described in Embodiment 4 of this application is a system used to implement the method of Embodiment 1 of this application, those skilled in the art can understand the specific structure and variations of the system based on the method described in Embodiment 1 of this application, and therefore will not be described again here. All systems used in the method of Embodiment 1 of this application fall within the scope of protection of this application.

[0071] Based on the same inventive concept, this application also provides a fifth embodiment, referring to... Figure 8 , Figure 8 This is a flowchart illustrating the fifth embodiment of the infrared imaging image processing method of this application.

[0072] In this embodiment, the infrared imaging image processing method further includes steps S51-S53: Step S51: Perform edge detection on the destriped image to obtain an edge map; Step S52: Determine the high-frequency region of the destriped image based on the edge map, and perform super-resolution reconstruction in the high-frequency region to obtain the super-resolution image; In this embodiment, the infrared imaging system can also employ an image super-resolution algorithm to improve the resolution of the infrared image. By calculating the gradient magnitude and direction pixel by pixel on the destriped image, and after non-maximum suppression and double thresholding, a binary edge map is obtained, in which the edge map marks the target contour and texture boundary.

[0073] Furthermore, the infrared imaging system uses the edge map as a mask to divide the destriped image into high-frequency and low-frequency regions. In the high-frequency region, the processor runs a lightweight SR-UNet network with the original image cropped as input and the detail blocks magnified twice as output. The network weights are trained offline using a blackbody-scene hybrid dataset collected by the same type of detector. After reconstruction, the high-frequency region is replaced with the super-resolution result, while the low-frequency region remains unchanged, forming a super-resolution image.

[0074] Step S53: Output super-resolution image.

[0075] Optionally, the infrared imaging system can output the super-resolution image to the image enhancement module for further image enhancement, or it can display the super-resolution image as the final display image.

[0076] This application embodiment, based on a high signal-to-noise ratio image with stripes removed, uses edge-guided super-resolution reconstruction to amplify only the high-frequency region, avoiding mis-amplification of the smooth background. This improves the target detail resolution, suppresses artifacts, and enhances the accuracy of subsequent recognition and ranging.

[0077] Since the system described in Embodiment 5 of this application is a system used to implement the method of Embodiment 1 of this application, those skilled in the art can understand the specific structure and variations of the system based on the method described in Embodiment 1 of this application, and therefore will not be described again here. All systems used in the method of Embodiment 1 of this application fall within the scope of protection of this application.

[0078] This application provides an infrared imaging image processing device, the device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the infrared imaging image processing method of the above embodiment 1.

[0079] The following is for reference. Figure 9 This document illustrates a schematic diagram of an image processing device suitable for implementing infrared imaging in the embodiments of this application. The image processing device for infrared imaging in the embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 9 The infrared imaging image processing device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0080] like Figure 9As shown, the infrared imaging image processing device may include a processing unit 1001 (e.g., a core processor, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the infrared imaging image processing device. The processing unit 1001, the read-only memory 1002, and the RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the infrared imaging image processing device to communicate wirelessly or wiredly with other devices to exchange data. Although the figure shows an infrared imaging image processing device with various systems, it should be understood that it is not required to implement or possess all the systems shown. More or fewer systems can be implemented alternatively.

[0081] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from read-only memory 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0082] The infrared imaging image processing device provided in this application, employing the infrared imaging image processing method in the above embodiments, can solve the technical problem of lost texture information in infrared images. Compared with the prior art, the beneficial effects of the infrared imaging image processing device provided in this application are the same as those of the infrared imaging image processing method provided in the above embodiments, and other technical features in this infrared imaging image processing device are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0083] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0084] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0085] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the infrared imaging image processing method in the above embodiments.

[0086] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, radio frequency (RF), etc., or any suitable combination thereof.

[0087] The aforementioned computer-readable storage medium may be included in an infrared imaging image processing device; or it may exist independently and not assembled into an infrared imaging image processing device.

[0088] The aforementioned computer-readable storage medium carries one or more programs. When the aforementioned one or more programs are executed by the infrared imaging image processing device, the infrared imaging image processing device performs the following: image feature separation on the input image through filtering to obtain a base layer image and a detail residual image. The detail residual image contains the detail content and stripe content of the input image. A one-dimensional row guided filtering is performed on the detail residual image, and the output of the one-dimensional row guided filtering is subtracted from the detail residual to obtain a row-filtered high-frequency image. Using the row-filtered high-frequency image as the input image and the base layer image as the guiding image, a one-dimensional column guided filtering is performed to obtain a stripe layer image. The difference between the stripe layer image and the base layer image is calculated to obtain a destriped image of the input image.

[0089] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0090] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0091] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0092] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described infrared imaging image processing method, thereby solving the technical problem of lost texture information in infrared images. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the infrared imaging image processing method provided in the above embodiments, and will not be repeated here.

[0093] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. An image processing method for infrared imaging, characterized in that, The method includes the following steps: The input image is processed by an edge-preserving guided filter to separate image features, resulting in a base layer image and a detail residual image. The detail residual image contains the detail content and stripe content of the input image. A one-dimensional row guided filter is performed on the detail residual image, and the difference between the output of the one-dimensional row guided filter and the detail residual is calculated to obtain the row filtered high-frequency image. Using the row-filtered high-frequency image as the input image and the base layer image as the guide image, a one-dimensional column-guided filter is performed to obtain the stripe layer image; The difference between the stripe layer image and the base layer image is calculated to obtain the stripe-free image of the input image.

2. The method as described in claim 1, characterized in that, The step of performing image feature separation on the input image through filtering to obtain the base layer image and the detail residual image includes: Perform Gaussian filtering on the input image to obtain a smoothed image; The smoothed image is subjected to edge-preserving guided filtering to generate the base layer image; The difference between the smoothed image and the base layer is calculated to obtain the detail residual image.

3. The method as described in claim 1, characterized in that, After the step of subtracting the stripe layer image from the base layer image to obtain the destriped image of the input image, the method further includes: The low-frequency and high-frequency components of the destriped image are separated by a low-pass filter to generate a background image and a detail image. By adjusting the dynamic range, noise reduction and enhancement are performed on the background image to obtain the target background image; And / or, perform detail enhancement on the detail image to obtain a detail layer image; The detail layer image and the target background image are combined to generate a display image.

4. The method as described in claim 1, characterized in that, Before the step of performing image feature separation on the input image through filtering to obtain the base layer image and the detail residual image, the method further includes: Raw infrared data is collected using an infrared detector; From the non-uniformity correction table pre-stored in the infrared detector, read the gain coefficient and bias value corresponding to each pixel in the original infrared data; The original pixel value of the pixel is multiplied by the corresponding gain coefficient to obtain the intermediate pixel value; The intermediate pixel value is added to the corresponding bias value to obtain the corrected pixel value; The corrected pixel values ​​are arranged based on the original pixel positions to form the input image.

5. The method as described in claim 1, characterized in that, The step of subtracting the stripe layer image from the base layer image to obtain the destriped image of the input image includes: Align the pixel coordinates of the stripe layer image with the base layer image to determine the target pixel point of the pixel point in the stripe layer image that has the same pixel coordinates as the pixel point in the base layer image. Calculate the pixel difference between the stated pixel and the target pixel; Based on the pixel coordinates, the pixel differences are used to reconstruct the image, resulting in the striped image.

6. The method as described in claim 1, characterized in that, After the step of subtracting the stripe image from the base layer image to obtain the destriped image of the input image, the method further includes: Edge detection is performed on the destriped image to obtain an edge map; The high-frequency region of the destriped image is determined based on the edge map, and super-resolution reconstruction is performed in the high-frequency region to obtain a super-resolution image; Output the super-resolution image.

7. An infrared imaging image processing device, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the image processing method for infrared imaging as described in any one of claims 1 to 6.

8. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the infrared imaging image processing method as described in any one of claims 1 to 6.

Images