CN104008527A - Method for defogging single image - Google Patents

Abstract

A method for defogging a single image of the present invention uses the regional color mean vector and L2 norm of the RGB image I to be defogged to obtain a propagation map, and then calculates the defogged image according to the propagation map and the atmospheric light value; The method can directly use a single RGB color image without any other information provided by the outside world; at the same time, the invention can directly obtain the defogged image by using coefficients in the process of calculating the defogged image, without further increasing the brightness of the image after defogging to meet the display requirements.

Description

A Single Image Dehazing Method

technical field

The invention belongs to the field of digital image processing, is suitable for preprocessing in the early stage of computer vision applications, and can be widely used in the fields of road, square video monitoring, driving recorder and the like, and more specifically is a method for defogging a single image.

Background technique

In recent years, my country's air pollution has become more and more serious. The resulting smog has reduced the visibility of the atmosphere. It is difficult for the outdoor video surveillance system to directly obtain image information with a certain degree of visibility, which directly leads to the failure of the surveillance system. Therefore, on the basis of the image obtained by the camera, it is of great significance to use the software algorithm to realize the image defogging process and increase the visibility of the image.

The existing image defogging algorithms mainly include: methods based on multi-view, methods based on prior knowledge, and methods that directly use a single image to defog. Methods based on multi-view and prior knowledge need prior information such as depth provided by the outside world. In practical applications, these depth information are difficult to obtain, so the applicability of these two types of methods is not strong. Directly using a single image to dehaze requires only a single image, and all the information required for dehazing is obtained by the algorithm from the single image to be dehazed, which is a commonly used method at present.

However, in the currently commonly used single image defogging methods, most of them use the atmospheric light value and the propagation map reflecting the depth information of the image to calculate the defogged image. However, the methods for obtaining the propagation map are mostly based on dark channels. 201210011326.5, an invention patent titled "a method and device for defogging a single image", but the use of dark channels has the following defects:

The dark channel method needs to calculate the minimum value twice for each sliding window image block. The operation of calculating the minimum value requires sorting operation of all pixels in the image block by pixel value, so the calculation is more time-consuming.

Contents of the invention

The main inventive point of the present invention is to obtain the propagation map through the statistical characteristics of the regional RGB color vectors. This method can directly use a single RGB color image without providing any other information from the outside; at the same time, the present invention can use the coefficient α in the process of calculating the defogged image. Obtain the defogged image directly without further increasing the brightness of the image after defogging to meet the display requirements.

In order to solve the above-mentioned technical problems, the present invention provides a method for defogging a single image, which calculates the image after defogging through the atmospheric light value of the foggy RGB color image and the propagation map reflecting the image depth information, and the propagation in the method The graph is obtained by:

Step A, obtaining the regional color mean vector and L2 norm of the RGB image I to be defogged;

Step A-1. Input the RGB image I to be defogged with a size of m×n. Each pixel in the image I corresponds to a 1×3 color vector [I ^R (i,j), I ^G (i,j), I ^B (i, j)], where i, j represent pixel coordinates, i ∈ [1, m], j ∈ [1, n], m, n are both positive integers;

Step A-2: Obtain all area windows Ω(i,j) in the image I that meet the following conditions: take the i-th row and the j-th column as the center pixel, and construct a size of (2r+1)×(2r+1) Area window; wherein, the value range of i is an integer between r+1 and m-r; the value range of j is an integer between r+1 and n-r, and r is the radius of the area window;

Step A-3: Use the following formula to obtain the image I region color mean vector E(i,j) corresponding to each region window Ω(i,j)

E(i,j)=[e ^R (i,j),e ^G (i,j),e ^B (i,j)] ^T , where C∈{R,G,B};

Step A-4: Obtain the L2 norm u(i,j) of the color mean vector of the corresponding image I area according to each E(i,j), and its expression form: u(i,j)=||E(i ,j)|| ₂ ;

Step B. Obtain a propagation map T reflecting the depth information of the image;

Step B-1, use the following formula to calculate the initial value of the propagation map T

$\stackrel{<span\; class="notranslate">\; ~</span>}{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\frac{{\mathrm{<span\; class="notranslate">\; A</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}{\mathrm{<span\; class="notranslate">\; E.</span>}}_{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}{{<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>}_{\mathrm{<span\; class="notranslate">\; 2</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; A</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}{\mathrm{<span\; class="notranslate">\; E.</span>}}_{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span>}}$

In the formula, A is the atmospheric light value, A ^T represents the transpose operation on A;

Step B-2. Use the guided filtering algorithm to further optimize the initial value of the propagation map A propagation map T reflecting the depth information of the image is obtained.

Another inventive point of the present invention, the atmospheric light value A of the foggy RGB color image can adopt the following steps to obtain:

a. Establish a matrix U of m×n, where the i-th row and j-th column element in the matrix U is the L2 norm u(i,j) of the color mean vector of the image I area, when i<r+1 or i> When m-r and j<r+1 or j>n-r, u(i,j)=0;

b. Select i _max and j _max corresponding to the largest element in the matrix U;

C, in image I, take i _max , j _max as center pixel point, set up the area window Ω _max (i _max , j _max ) that size is (2r+1) * (2r+1), r is the area window radius;

d. Calculate the L2 norm of the 1×3 color vector corresponding to all pixels in Ω _max (i _max , j _max ), where the color vector corresponding to the maximum value in the L2 norm of the color vector is the atmospheric light value A.

In order to directly obtain the defogged image, there is no need to further increase the image brightness after defogging to meet the display

Requirements, in the single image defogging method of the present invention, the image J after defogging is calculated according to the following formula:

$\mathrm{<span\; class="notranslate">\; J</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; A</span>}}{\mathrm{<span\; class="notranslate">\; max</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; t</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; \&alpha;A</span>}$

In the formula, t ₀ =0.3, a∈[0.6,0.9], T(i,j) is the propagation map, I(i,j) is the original image to be dehazed; max(T(i,j),t ₀ ) means that when T(i,j)≥t ₀ , max(T(i,j),t ₀ )=T(i,j); otherwise max(T(i,j),t ₀ )=t ₀ .

Compared with the prior art, the present invention has the following significant advantages: (1) The image defogging method based on the statistical characteristics of the regional RGB color vector can directly use a single RGB color image without any other information provided by the outside world; (2) A large number of search minimum calculations are performed to improve the real-time performance of image defogging; (3) The defogged image is directly obtained without further increasing the image brightness after defogging to meet the display requirements.

Description of drawings

Fig. 1 is a schematic diagram of obtaining all the region windows Ω(i, j) satisfying the conditions in the image I;

Fig. 2 (a) is the original fog image I of the embodiment; Fig. 2 (b) is the initial value of the transmission map in the embodiment Fig. 2(c) is the transmission map T after guided filtering in the embodiment; Fig. 2(d) is the image J after defogging in the embodiment;

Detailed ways

The present invention is a method for defogging a single image. The method calculates the image after defogging through the atmospheric light value of the foggy RGB color image and the propagation map reflecting the depth information of the image. The propagation map in the method is obtained by the following method:

Step A, obtaining the regional color mean vector and L2 norm of the RGB image I to be defogged;

Step A-1. Input the RGB image I to be defogged with a size of m×n. Each pixel in the image I corresponds to a 1×3 color vector [I ^R (i,j), I ^G (i,j), I ^B (i, j)], where i, j represent pixel coordinates, i ∈ [1, m], j ∈ [1, n], m, n are both positive integers;

Step A-2: Obtain all area windows Ω(i,j) in the image I that meet the following conditions: take the i-th row and the j-th column as the center pixel, and construct a size of (2r+1)×(2r+1) Area window; wherein, the value range of i is an integer between r+1 and m-r; the value range of j is an integer between r+1 and n-r, and r is the radius of the area window;

Step A-3: Use the following formula to obtain the image I region color mean vector E(i,j) corresponding to each region window Ω(i,j)

E(i,j)=[e ^R (i,j),e ^G (i,j),e ^B (i,j)] ^T , where C∈{R,G,B};

Step A-4: Obtain the L2 norm u(i,j) of the color mean vector of the corresponding image I area according to each E(i,j), and its expression form: u(i,j)=||E(i ,j)|| ₂ ;

Step B. Obtain a propagation map T reflecting the depth information of the image;

Step B-1, use the following formula to calculate the initial value of the propagation map T

$\stackrel{<span\; class="notranslate">\; ~</span>}{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\frac{{\mathrm{<span\; class="notranslate">\; A</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}{\mathrm{<span\; class="notranslate">\; E.</span>}}_{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}{{<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>}_{\mathrm{<span\; class="notranslate">\; 2</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; A</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}{\mathrm{<span\; class="notranslate">\; E.</span>}}_{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span>}}$

In the formula, A is the atmospheric light value, A ^T represents the transpose operation on A;

Step B-2. Use the guided filtering algorithm to further optimize the initial value of the propagation map Obtain the propagation map T that reflects the depth information of the image; use the guided filtering algorithm to further optimize the initial value of the propagation map It is a well-known technology in the art. For details, see He, KM, Sun, J., and Tang, XOGuided ImageFiltering, IEEE Transactions on Pattern Analysis and Machine Intelligence, 2013, 35(6): 1397-1409, and will not repeat them here!

Obtaining the atmospheric light value of the foggy RGB color image can be obtained by traditional methods or by the following steps:

a. Establish a matrix U of m×n, where the i-th row and j-th column element in the matrix U is the L2 norm u(i,j) of the color mean vector of the image I area, when i<r+1 or i> When m-r and j<r+1 or j>n-r, u(i,j)=0;

b. Select i _max and j _max corresponding to the largest element in the matrix U;

c. In the image I, take i _max and j _max as the central pixel to establish a regional window Ω _max (i _max , j _max ), the size of the regional window is (2r+1)×(2r+1), and r is the radius of the regional window ;

d. Calculate the L2 norm of the 1×3 color vector corresponding to all pixels in Ω _max (i _max , j _max ), where the color vector corresponding to the maximum value in the L2 norm of the color vector is the atmospheric light value A.

When calculating the dehazed image using the atmospheric light value and the propagation map, the traditional formula can be used However, the effect of the defogged image is not good, and further image brightening processing is required to achieve the display effect;

Also can adopt defogging formula of the present invention $J(i,j)=\frac{I(i,j)-A}{\max (T(i,j),t_0)}+\mathrm{\&alpha;A},$ Where t ₀ =0.3,α∈[0.6,0.9], T(i,j) is the propagation map, I(i,j) is the original image to be dehazed; max(T(i,j),t ₀ ) means When T(i,j)≥t ₀ , max(T(i,j),t ₀ )=T(i,j); otherwise max(T(i,j),t ₀ )=t ₀ , α =0.8 is the best effect.

Example

As shown in Figure 2(a), the dehazing process of the color image I to be defogged is 470×350 in detail as follows: Step A, obtain the regional color mean vector and L2 norm of the RGB image I to be defogged;

Step A-1. Input the RGB image I to be defogged with a size of 470×350. Each pixel in the image I corresponds to a 1×3 color vector [I ^R (i,j), I ^G (i,j), I ^B (i, j)], where i, j represent pixel coordinates, i∈[1,470], j∈[1,350];

Step A-2: Get all the area windows Ω(i,j) in the image I that meet the following conditions: the area window takes the i-th row and j-th column as the center pixel point, and the size is (2r+1)×(2r+1 ); wherein, r is the radius of the region window, r=7 means to take 7 pixels, that is, the size of the region window is 15×15, and the value range of i is an integer between 8 and 463; the value range of j is an integer between 8 and 354;

Step A-3: Use the following formula to obtain the image I region color mean vector E(i,j) corresponding to each region window Ω(i,j)

E(i,j)=[e ^R (i,j),e ^G (i,j),e ^B (i,j)] ^T , where C∈{R,G,B};

Step A-4: Obtain the L2 norm u(i,j) of the color mean vector of the corresponding image I area according to each E(i,j), and its expression form: u(i,j)=||E(i ,j)|| ₂ ;

Step B. Obtain a propagation map T reflecting the depth information of the image;

Step B-1, use the following formula to calculate the initial value of the propagation map T As shown in Figure 2(b):

$\stackrel{<span\; class="notranslate">\; ~</span>}{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\frac{{\mathrm{<span\; class="notranslate">\; A</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}{\mathrm{<span\; class="notranslate">\; E.</span>}}_{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}{{<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>}_{\mathrm{<span\; class="notranslate">\; 2</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; A</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}{\mathrm{<span\; class="notranslate">\; E.</span>}}_{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span>}}$

In the formula, A is the atmospheric light value, and A ^T represents the transposition operation on A, where the calculation process of the atmospheric light value A is as follows:

(1) Establish a 470×350 matrix U, where the i-th row and j-th column element in the matrix U is the L2 norm u(i,j) of the color mean vector of the image I region, when i<8 or i>463 And when j<8 or j>354, u(i,j)=0;

(2) Select i _max , j _max corresponding to the largest element in the matrix U;

(3) Establish a region window Ω _max (i _max , j _max ) with i _max and j _max as the center pixel in image I, the size of the region window is (2r+1)×(2r+1), r is the region window Radius, r=7, that is, the window size is 15×15;

(4) Calculate the L2 norm of the color vectors of all elements in Ω _max (i _max , j _max ), where the color vector corresponding to the maximum value in the L2 norm of the color vector is the atmospheric light value A;

Step B-2. Use the guided filtering algorithm to further optimize the initial value of the propagation map The transmission map T reflecting the depth information of the image is obtained as shown in Figure 2(c);

Step C, use the transmission map T and the atmospheric light value A to calculate the image J after defogging, as shown in Figure 2(d):

$\mathrm{<span\; class="notranslate">\; J</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; A</span>}}{\mathrm{<span\; class="notranslate">\; max</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; t</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; \&alpha;A</span>}$

In the formula, t ₀ =0.3, α=0.8, A=[0.9294,0.9373,0.9333];

The single image defogging method of the present invention is an image defogging method based on the statistical characteristics of regional RGB color vectors, which can directly use a single RGB color image without any other information provided by the outside; at the same time, the defogged image can be directly obtained by using the coefficient α, There is no need to further increase image brightness after defogging to meet display requirements.

Claims (5)

1. A single image defogging method, the method calculates the image after the fog removal by the atmospheric light value of the foggy RGB color image and the propagation map reflecting the image depth information, it is characterized in that the propagation map in the method is passed The following methods are obtained: Step A, obtaining the regional color mean vector and L2 norm of the RGB image I to be defogged; Step A-1. Input the RGB image I to be defogged with a size of m×n. Each pixel in the image I corresponds to a 1×3 color vector [I ^R (i,j), I ^G (i,j), I ^B (i, j)], where i, j represent pixel coordinates, i ∈ [1, m], j ∈ [1, n], where m, n are both positive integers; Step A-2: Obtain all area windows Ω(i,j) in the image I that meet the following conditions: take the i-th row and the j-th column as the center pixel, and construct a size of (2r+1)×(2r+1) Area window; wherein, the value range of i is an integer between r+1 and m-r; the value range of j is an integer between r+1 and n-r, and the radius of the r area window; Step A-3: Use the following formula to obtain the image I region color mean vector E(i,j) corresponding to each region window Ω(i,j) E(i,j)=[e ^R (i,j),e ^G (i,j),e ^B (i,j)] ^T , where C∈{R,G,B}; Step A-4: Obtain the L2 norm u(i,j) of the color mean vector of the corresponding image I area according to each E(i,j), and its expression form: u(i,j)=||E(i ,j)|| ₂ ; Step B. Obtain a propagation map T reflecting the depth information of the image; Step B-1, use the following formula to calculate the initial value of the propagation map T $\stackrel{<span\; class="notranslate">\; ~</span>}{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\frac{{\mathrm{<span\; class="notranslate">\; A</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}{\mathrm{<span\; class="notranslate">\; E.</span>}}_{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}{{<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>}_{\mathrm{<span\; class="notranslate">\; 2</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; A</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}{\mathrm{<span\; class="notranslate">\; E.</span>}}_{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span>}}$ In the formula, A is the atmospheric light value, A ^T represents the transpose operation on A; Step B-2. Use the guided filtering algorithm to further optimize the initial value of the propagation map A propagation map T reflecting the depth information of the image is obtained. 2. The single image defogging method according to claim 1, wherein the acquisition steps of the atmospheric light value A of the foggy RGB color image are as follows: a. Establish an m×n matrix U, where the i-th row and j-th column element in the matrix U is the L2 norm u(i,j) of the color mean vector in the image I area, when i<r+1 or i> When m-r and j<r+1 or j>n-r, u(i,j)=0; b. Select i _max and j _max corresponding to the largest element in the matrix U; c. In the image I, take i _max and j _max as the center pixel, and set up a region window Ω _max (i _max , j _max ) whose size is (2r+1)×(2r+1); d. Calculate the L2 norm of the 1×3 color vector corresponding to all pixels in Ω _max (i _max , j _max ), where the 1×3 color vector corresponding to the maximum value in the L2 norm of the color vector is the atmospheric light value A . 3. The method for defogging a single image according to claim 1 or 2, wherein the image J after defogging is calculated according to the following formula: $\mathrm{<span\; class="notranslate">\; J</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; A</span>}}{\mathrm{<span\; class="notranslate">\; max</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; t</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; \&alpha;A</span>}$ In the formula, t ₀ =0.3,α∈[0.6,0.9], T(i,j) is the propagation map, I(i,j) is the original image to be dehazed; max(T(i,j),t ₀ ) means that when T(i,j)≥t ₀ , max(T(i,j),t ₀ )=T(i,j); otherwise max(T(i,j),t ₀ )=t ₀ . 4. The method for defogging a single image according to claim 3, wherein the α=0.8. 5. The method for defogging a single image according to claim 1 or 2, wherein the area window radius is r=7.