Image defogging method based on combined polarization process of target light and atmospheric light

Abstract

The invention provides an image defogging method based on a combined polarization process of target light and atmospheric light, and belongs to the technical field of image processing. The method comprises the following steps: constructing a combined polarization process of atmospheric light and target light based on a Stokes vector of probe light; representing a combined polarization process of the target light and the atmospheric light through vectors, and constructing a combined polarization model of the atmospheric light and the target light; solving the joint polarization model of the atmospheric light and the target light to obtain an atmospheric light image; and obtaining a target light image based on the atmospheric light image through an atmospheric scattering model, and taking the target light image as a defogged image. According to the method, the joint polarization model of the atmospheric light and the target light is constructed, and the estimation of the atmospheric light is optimized through maximum value screening and mean filtering, so that the atmospheric light better conforms to the real environment distribution, noise interference is avoided, the target light is solved, and the defogging robustness and accuracy are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, in particular to an image defogging method based on joint polarization process of target light and atmospheric light. BACKGROUND

[0002] The absorption, scattering and reflection characteristics of electromagnetic waves of different wavelengths are severely disturbed by heavy fog, so that the images collected by the sensor will have problems such as low contrast, color distortion, and blurred image details. In order to improve the image quality captured in heavy fog weather, researchers have proposed many methods for image defogging. The defogging method based on polarization imaging can obtain more information from different polarization images, so as to achieve better defogging results. Therefore, polarization defogging has been widely concerned and researched.

[0003] However, most of the current polarization defogging algorithms are usually based on the method of taking the target light as the natural light, and all the polarization information is provided by the atmospheric light. But in actual situation, the target usually has a degree of polarization, and in most scenes, the degree of polarization of the target is even much higher than that of the atmospheric light. The area with high degree of polarization will produce distortion in the image recovery process.

[0004] Therefore, there is an urgent need for a fast image defogging method based on the joint polarization process of target light and atmospheric light. SUMMARY

[0005] Therefore, the present application provides an image defogging method based on the joint polarization process of target light and atmospheric light, which realizes the conversion from physical model to mathematical model through vector representation and analytic geometry. Combined with the atmospheric scattering model, the image defogging and enhancement are realized.

[0006] To this end, the present application provides the following technical solutions: An image defogging method based on the joint polarization process of target light and atmospheric light, comprising: Constructing the joint polarization process of atmospheric light and target light based on the Stokes vector of the probe light; Representing the joint polarization process of atmospheric light and target light by vectors, and constructing a joint polarization model of atmospheric light and target light; Solving the joint polarization model of atmospheric light and target light to obtain an atmospheric light image; Obtaining a target light image based on the atmospheric light image through the atmospheric scattering model, as a defogging image.

[0007] Further, it further comprises: Enhancing the target light image to obtain a defogging image by using gray scale stretching.

[0008] Furthermore, the joint polarization process of the atmospheric light and the target light is constructed based on the Stokes vector of the detection light, including: Calculate the Stokes vector of the detection light based on images in different polarization directions; Calculate the polarization degree and polarization angle of the detection light based on the Stokes vector of the detection light; Based on the fact that the Stokes vector of the detection light is the superposition of the Stokes vector of the atmospheric light and the Stokes vector of the target light, the joint polarization process of the target light and the atmospheric light is constructed.

[0009] Furthermore, the joint polarization process of atmospheric light and target light is:

[0010]

[0011] Where, Indicates the polarization degree of the target light; Indicates the polarization degree of atmospheric light; Indicates the polarization angle of the target light; represents the polarization angle of atmospheric light; Stokes vector representing atmospheric light; Represents the Stokes vector of the target light.

[0012] Furthermore, the joint polarization process of target light and atmospheric light is characterized by vectors, including:

[0013] in, ; represents the coordinates of the detection light; represents the Stokes vector of the probe light.

[0014] Furthermore, the joint polarization model of atmospheric light and target light includes:

[0015]

[0016] in, represents the polarization degree of the detection light; .

[0017] Furthermore, the joint polarization model of atmospheric light and target light is solved to obtain an atmospheric light image, including: Combining polarization theory and dark channel prior theory, the polarization dark channel method is used to determine the light intensity value at infinity; Determine the polarization degree of the target light based on the light intensity value at infinity; Taking the light intensity at infinity as the boundary constraint, the joint polarization model of atmospheric light and target light is solved to obtain the atmospheric light image; and the atmospheric light image is filtered.

[0018] Furthermore, by combining polarization theory and dark channel prior theory, the polarization dark channel method is used to determine the light intensity value at infinity, including: Solve the minimum value of different polarization components in each pixel to generate a dark channel image; Select the window area with the largest light intensity in the dark channel image as the sky area; Calculate the average polarization degree of the sky region as the polarization degree of the sky region; calculate the average polarization angle of the sky region as the polarization angle of the sky region; The average light intensity of the sky area is taken as the light intensity value at infinity.

[0019] Furthermore, determining the polarization degree of the target light based on the light intensity value at infinity includes: When the polarization degree of the pixel point is less than or equal to the polarization degree of the sky area, the polarization degree of the target light is equal to the average polarization degree of the window area with the maximum light intensity.

[0020] When the polarization degree of the pixel point is greater than the polarization degree of the sky area, the polarization degree of the target light is equal to the polarization degree of the pixel point.

[0021] Furthermore, the light intensity at infinity is used as the boundary constraint to solve the joint polarization model of atmospheric light and target light to obtain the atmospheric light image, including:

[0022]

[0023] in, The root with the largest absolute value that fits the range.

[0024] Advantages and positive effects of the present invention: This method constructs a joint polarization model of atmospheric and target light to solve the target light image. It uses a polarization dark channel to accurately identify the sky area, reducing misidentification of bright areas. It also uses a maximum window polarization degree to enhance the boundary constraints of the target light solution, ensuring clear edge details. Furthermore, it optimizes the atmospheric light estimate through maximum filtering and mean filtering, making it more consistent with the actual environmental distribution and avoiding noise interference. This method then solves the target light, significantly improving the robustness and accuracy of dehazing. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0026] Figure 1 The flowchart of the image dehazing method based on the joint polarization process of target light and atmospheric light; Figure 2 Flowchart of an image defogging method based on a joint polarization process of target light and atmospheric light in an embodiment of the present invention; Figure 3 This is the vector representation diagram of the joint polarization dehazing algorithm in this implementation; Figure 4 Schematic diagram of the polarization unit array and data acquisition device in this embodiment; Figure 5 : This is the fog intensity map under different scenes in this embodiment; Figure 6 1 is an atmospheric window diagram under different scenarios in this embodiment; Figure 7 is a polarization degree diagram under different scenarios in this embodiment; Figure 8 is a polarization angle diagram under different scenarios in this embodiment; Figure 9 Filtered atmospheric light images under different scenarios in this embodiment; Figure 10 Target light images in different scenarios in this embodiment; Figure 11 These are defogging images under different scenarios in this embodiment. DETAILED DESCRIPTION

[0027] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0028] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0029] This invention provides an image dehazing method based on the joint polarization process of target and atmospheric light. A polarization camera is used to acquire the original polarization image. The sky region is determined using dark channel prior theory. The atmospheric light polarization information is then deduced based on the sky region. The atmospheric light polarization information is represented by a vector, and a joint polarization model of the atmospheric and target light is established. Boundary constraints are applied using the maximum window polarization degree to further enhance the accuracy of the target light decomposition. Finally, an atmospheric scattering model is incorporated to achieve image dehazing. Image normalization is also used to achieve image enhancement.

[0030] Combine Figure 1 The method comprises the steps of: S1. Construct a joint polarization model of atmospheric light and target light based on the Stokes vector of the detection light; 1) Based on different polarization directions Four images of , calculate the Stokes vector of the detection light; the calculation formula is:

[0031] in, Indicates the light intensity value corresponding to the 0 degree polarization direction, Indicates the light intensity value corresponding to the 45-degree polarization direction, Indicates the light intensity value corresponding to the 90-degree polarization direction, Indicates the light intensity value corresponding to the polarization direction of 135 degrees; Indicates the light intensity value, It represents the difference between the light intensity of the 0-degree component and the 90-degree component. Indicates the difference in light intensity between the 45-degree component and the 135-degree component; represents the Stokes vector of the probe light.

[0032] 2) Based on the Stokes vector of the detection light, calculate the polarization degree and polarization angle of the detection light:

[0033]

[0034] wherein, denotes the degree of polarization of the probe light; denotes the polarization angle of the probe light.

[0035] 3) Based on the superposition of the Stokes vector of the probe light and the Stokes vector of the atmospheric light and the Stokes vector of the target light, a joint polarization process of the target light and the atmospheric light is constructed;

[0036] wherein, denotes the Stokes vector of the atmospheric light; denotes the Stokes vector of the target light.

[0037] The formula of the joint polarization process of the target light and the atmospheric light is:

[0038]

[0039] In the formula, denotes the degree of polarization of the target light; denotes the degree of polarization of the atmospheric light; denotes the polarization angle of the target light; denotes the polarization angle of the atmospheric light.

[0040] 4) A joint polarization model of the atmospheric light and the target light is constructed by representing the joint polarization process of the target light and the atmospheric light by a vector; The joint polarization process of the target light and the atmospheric light is introduced into the coordinate system by representing it by a vector, and the formula is:

[0041] wherein, is expressed by the coordinate system:

[0042] The degree of polarization and the polarization angle of the probe light are calculated by the polarization camera, and the coordinates of the probe light are determined:

[0043] wherein, denotes the coordinates of the probe light; .

[0044] is expressed by the coordinate system:

[0045] By expressing the results in a simultaneous coordinate system, we can obtain:

[0046] 5) Determine the solvable form of the joint polarization model of atmospheric light and target light:

[0047]

[0048] in, represents the polarization degree of the detection light; .

[0049] S2. Solve the joint polarization model of atmospheric light and target light to obtain an atmospheric light image.

[0050] Traditional methods for calculating the sky region typically select the area with the highest light intensity as the sky region. This is because in foggy conditions, the atmospheric light is often much greater than the target light. However, this method is inaccurate and fails in light fog or when the target light is greater than the atmospheric light.

[0051] Under most conditions, the polarization degree of the target light is much higher than that of the atmospheric light. According to Malus's law, among different channels, the target light usually has a lower intensity value in one channel, while the atmospheric light intensity in each channel is relatively average.

[0052] 1) In order to ensure the accuracy of defogging, the polarization dark channel method is used to determine the light intensity at infinity by combining polarization theory and dark channel prior theory: Solve the minimum value of different polarization components in each pixel to generate a dark channel image; Select the window area with the largest light intensity in the dark channel image as the sky area; Calculate the average polarization degree of the sky region as the polarization degree of the sky region; calculate the average polarization angle of the sky region as the polarization angle of the sky region; The average light intensity of the sky area is taken as the light intensity value at infinity.

[0053] 2) Since the polarization degree reflects the object information to a certain extent, the polarization degree of the target light is determined based on the light intensity value at infinity: When the polarization degree of the pixel point is less than or equal to the polarization degree of the sky area, the atmospheric light information is higher than the target light, and the polarization degree of the target light is equal to the average polarization degree of the window area with the maximum light intensity.

[0054] When the polarization degree of the pixel point is greater than the polarization degree of the sky area, the information of the target light is higher than that of the atmospheric light, and the polarization degree of the target light is equal to the polarization degree of the pixel point.

[0055] Because the target light is usually covered by the atmospheric light in heavy fog, there is a large error in directly obtaining the target light. Therefore, the maximum atmospheric light is solved first, and the atmospheric light is filtered to approximate the real atmospheric light of the environment, and then the target light is solved. Because the dynamic range of the atmospheric light is small relative to the target light, it is aimed at the local area of the picture. The target light is determined by the material, reflectivity, distance, and other factors of the object, and has a large dynamic range, which is pixel-level. If the target light is directly obtained, the image will have a lot of noise interference.

[0056] 3) The light intensity value at infinity is used as a boundary constraint to ensure the accuracy of the atmospheric light solution. The joint polarization model of the atmospheric light and the target light is solved to obtain the atmospheric light image. The atmospheric light image is filtered to reduce noise.

[0057] The formula for solving the atmospheric light image is:

[0058]

[0059] wherein, is the root of the maximum absolute value in the range.

[0060] S3, obtain the target light image based on the atmospheric light image through the atmospheric scattering model, wherein the atmospheric scattering model is:

[0061] wherein, is the fog image captured by the detector; is the atmospheric light image; is the light intensity value at infinity.

[0062] S4, enhance the target light image to obtain a de-fog image by using gray scale stretching. Under heavy fog conditions, the target light is disturbed by the atmospheric light, so the light intensity value of the target light is small. Therefore, the target light is subjected to gray scale stretching to improve the brightness, so that the image is clearer. Specific embodiments S1, use a tripod and a gimbal to build an experimental shooting platform combined with a polarization camera, as shown in Figure 3 Each polarization unit is 2*2 pixels, arranged in 0, 45, 90, and 135 clockwise. According to the repeated polarization unit arrangement, four polarization images of different components are obtained, and the Stokes vector is calculated.

[0064] S2, determine the joint polarization model of the target light and the atmospheric light through the Stokes vector.

[0065] S3, solve the joint polarization model of the atmospheric light and the target light to obtain the atmospheric light image.

[0066] 1) Combining polarization theory and dark channel prior theory, the polarization dark channel method is used to determine the light intensity at infinity. To enhance robustness, a coefficient is added to ensure that the light intensity at infinity is the maximum value.

[0067] 2) Under the maximum window polarization constraint, solve the joint polarization model of the atmospheric light and the target light to obtain an atmospheric light image to ensure accurate atmospheric light solution. Filter the atmospheric light image to reduce noise. A bias coefficient is set before the target light polarization, ensuring that the biased target light polarization is slightly larger than the target light polarization. In this example, the bias coefficient is set to 1.2.

[0068] S4. Use the atmospheric scattering model to obtain the target light map.

[0069] S5. Perform grayscale stretching on the target light image to increase brightness and obtain an enhanced target light image.

[0070] The effectiveness of this method is verified through comparative experiments: Indicators for quantitative analysis of image dehazing results: information entropy, average gradient, image standard deviation, and SSIM; Information entropy measures the amount of information in an image. A higher entropy indicates more information and better image quality. The mean gradient reflects the rate of contrast change in tiny image details and, to a certain extent, reflects image clarity. The image standard deviation reflects the degree of dispersion of pixel values ​​relative to the mean. A higher standard deviation indicates sharper edges and better image quality. However, due to image distortion caused by traditional dehazing algorithms, the standard deviation may be biased high. Therefore, the SSIM evaluation metric is also used. SSIM evaluates the similarity between two images based on brightness, contrast, and structure.

[0071] Different algorithms were evaluated against the original image to verify the effectiveness of dehazing while preserving the original image. The resulting dehazed image evaluation metrics are shown in Table 1. In most scenarios, this method achieved the highest information entropy and average gradient while maintaining the highest similarity to the original image. This demonstrates the effectiveness of this method for image dehazing and enhancement.

[0072] Table 1

[0073] Experiments have shown that compared with other models, this model can be applied to most scenarios, and while ensuring maximum structural similarity, it has excellent image dehazing and enhancement effects, and is closer to the natural state.

[0074] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An image defogging method based on the joint polarization process of target light and atmospheric light, characterized in that: include: The joint polarization process of atmospheric light and target light is constructed based on the Stokes vector of the detection light; The joint polarization process of target light and atmospheric light is represented by vectors, and a joint polarization model of atmospheric light and target light is constructed. Solve the joint polarization model of atmospheric light and target light to obtain the atmospheric light image; The target light image is obtained based on the atmospheric light image through the atmospheric scattering model as the defogging image.

2. The method according to claim 1, characterized in that Also includes: Grayscale stretching is used to enhance the target light image to obtain the dehazed image.

3. The method according to claim 1, characterized in that The process of constructing the joint polarization of atmospheric light and target light based on the Stokes vector of the detection light includes: Calculate the Stokes vector of the detection light based on images in different polarization directions; Calculate the polarization degree and polarization angle of the detection light based on the Stokes vector of the detection light; Based on the fact that the Stokes vector of the detection light is the superposition of the Stokes vector of the atmospheric light and the Stokes vector of the target light, the joint polarization process of the target light and the atmospheric light is constructed.

4. The method according to claim 1, wherein The joint polarization process of the atmospheric light and the target light is: Where, Indicates the polarization degree of the target light; Indicates the polarization degree of atmospheric light; Indicates the polarization angle of the target light; represents the polarization angle of atmospheric light; Stokes vector representing atmospheric light; Represents the Stokes vector of the target light.

5. The method according to claim 1, wherein The process of characterizing the joint polarization of target light and atmospheric light by vectors includes: in, ; represents the coordinates of the detection light; represents the Stokes vector of the probe light.

6. The method according to claim 1, characterized in that The joint polarization model of the atmospheric light and the target light includes: in, represents the polarization degree of the detection light; .

7. The method according to claim 1, characterized in that Solving the joint polarization model of atmospheric light and target light to obtain an atmospheric light image includes: Combining polarization theory and dark channel prior theory, the polarization dark channel method is used to determine the light intensity value at infinity; Determine the polarization degree of the target light based on the light intensity value at infinity; Taking the light intensity at infinity as the boundary constraint, the joint polarization model of atmospheric light and target light is solved to obtain the atmospheric light image; and the atmospheric light image is filtered.

8. The method according to claim 7, characterized in that The method of combining polarization theory and dark channel prior theory to determine the light intensity value at infinity using the polarization dark channel method includes: Solve the minimum value of different polarization components in each pixel to generate a dark channel image; Select the window area with the largest light intensity in the dark channel image as the sky area; Calculate the average polarization degree of the sky region as the polarization degree of the sky region; calculate the average polarization angle of the sky region as the polarization angle of the sky region; The average light intensity of the sky area is taken as the light intensity value at infinity.

9. The method according to claim 7, characterized in that The determining of the polarization degree of the target light based on the light intensity value at infinity includes: When the polarization degree of the pixel point is less than or equal to the polarization degree of the sky area, the polarization degree of the target light is equal to the average polarization degree of the window area with the maximum light intensity. When the polarization degree of the pixel point is greater than the polarization degree of the sky area, the polarization degree of the target light is equal to the polarization degree of the pixel point.

10. The method according to claim 7, characterized in that The method of solving the joint polarization model of atmospheric light and target light using the light intensity value at infinity as a boundary constraint to obtain the atmospheric light image includes: in, The root with the largest absolute value that fits the range.

Images