WO2020232710A1 - Haze image quality evaluation method and system, storage medium, and electronic device - Google Patents

Abstract

The present invention relates to a haze image quality evaluation method and system, a storage medium, and an electronic device, comprising the following steps: S1. obtaining a haze image, and obtaining a first transmittance map corresponding to the haze image on the basis of a dark channel prior method; S2. obtaining a non-sky region of the haze image; S3. obtaining a first incident light attenuation rate corresponding to the non-sky region according to the first transmittance map and the non-sky region, so as to obtain the image quality of the haze image according to the first incident light attenuation rate. According to the present invention, the interference of a sky region on the haze image quality evaluation is reduced by eliminating the sky region, thereby optimizing the image quality evaluation effect.

Description

Haze image quality evaluation method, system, storage medium and electronic equipment Technical field

The invention relates to the technical field of haze image quality evaluation, and more specifically, to a haze image quality evaluation method, system, storage medium and electronic equipment.

Background technique

Haze image quality evaluation has broad application prospects. For example, according to the haze image to estimate the haze concentration and impact in time, it can be used to predict the air quality level in weather forecasting, it can be used to estimate visibility in highway monitoring, and it can be used to calculate driving safety in the field of unmanned driving.

Similar to image quality evaluation methods, haze image quality evaluation is also divided into two categories: subjective methods and objective methods. Among them, subjective image quality evaluation takes a long time and is difficult to apply in real-time in embedded devices, so it cannot be directly applied to the field of video surveillance. In the objective haze image quality evaluation method, since there is no corresponding original haze-free image in the haze image itself, the focus of the research should be the non-reference haze image quality evaluation algorithm.

However, research in this area at home and abroad is relatively limited, and many methods still use traditional full-reference image quality evaluation algorithms to evaluate haze image quality. The following introduces several commonly used haze image quality evaluation algorithms.

(1) The structural similarity method (Structural similarity index, SSIM) is a reference image quality evaluation method. The greater the evaluation value, the greater the similarity between images. This method considers both brightness and contrast. The formula is as follows:

Among them, μ _x and μ _y represent the mean values of images x and y, respectively, and represent brightness information. σ _x and σ _y represent the variance of the image x and y, respectively, and represent contrast information, and C ₁ , C ₂ and C ₃ are constants.

(2) Peak signal to noise ratio (PSNR) is a traditional energy-based comparison method. The higher the PSNR value, the higher the similarity. The formula is as follows:

Among them, L is the total number of gray levels, usually 255, and σ represents the mean square error of the image.

(3) The Brenner gradient method is a non-reference image quality evaluation method. The larger the value, the higher the image quality. The formula is as follows:

Among them, f(x, y) is the gray value of the corresponding pixel of the image, and D(f) is the result of image quality evaluation.

(4) Point sharpness method. The point sharpness evaluation method belongs to the non-reference type image quality evaluation method. The higher the value, the better the image quality evaluation result. Xu et al. believe that the greater the edge gray level changes, the higher the sharpness and the lower the haze concentration. Therefore, the image quality can be evaluated by statistical point sharpness. The formula is as follows:

Among them, dI/dx represents the gray-scale derivative in the edge direction, and I(b)-I(a) represents the overall gray-scale change in the edge direction.

This method only counts specific image areas, and this area needs to be manually selected, which is not conducive to automation.

(5) The entropy method belongs to the non-reference image quality evaluation method. The greater the entropy of the image, the better the image quality. Image entropy is based on statistical features, used to measure the richness of image information, and is an important indicator of the amount of image information measured. The formula is as follows:

Among them, Pi is the probability that a pixel with a gray value of i appears in the image, and L is the total number of gray levels.

(6) Gray scale difference method (SMD) is a non-reference image quality evaluation method. The larger the gray-scale variance of the image, the better the image quality. An image with a lower degree of haze has more high-frequency components, so the change in gray level can be used as a basis for evaluating the quality of haze images. The formula is as follows:

Among them, f(x, y) represents the gray value of the pixel at the coordinate (x, y) on the image.

This method is convenient and fast to calculate, but the disadvantage is that the sensitivity is not high at dense gradients.

In the above-mentioned existing solutions, the image quality is evaluated by calculating the average gradient or point sharpness of the image. This kind of scheme fails to consider the physical model of the haze image degradation process. Moreover, images of different scenes have different average gradient or point sharpness characteristics. Therefore, such methods are difficult to generalize to the quality comparison of haze images in different scenes.

In summary, there is still a lot of room for improvement in the evaluation performance of haze image quality in existing schemes, and it is necessary to improve the haze image quality evaluation method and the scope of image collection.

Summary of the invention

The technical problem to be solved by the present invention is to provide a haze image quality evaluation method, system, storage medium and electronic equipment in view of the above-mentioned defects of the prior art.

The technical solution adopted by the present invention to solve its technical problems is: constructing a method for evaluating the quality of haze images, including the following steps:

S1. Acquire a haze image, and acquire a first transmittance map corresponding to the haze image based on a dark channel prior method;

S2. Obtain a non-sky area of the haze image;

S3. Acquire a first incident light attenuation rate corresponding to the non-sky area according to the first transmittance map and the non-sky area, so as to obtain the image quality of the haze image according to the first incident light attenuation rate .

Preferably, in the step S3, the calculating the first incident light attenuation rate corresponding to the non-sky area according to the first transmittance map and the non-sky area includes:

Acquiring the second incident light attenuation rate of the haze image according to the first transmittance map;

Obtaining the first incident light attenuation rate corresponding to the non-sky area according to the second incident light attenuation rate and the non-sky area; or

Acquiring a second transmittance map corresponding to the non-sky area according to the first transmittance map and the non-sky area;

Acquire the first incident light attenuation rate corresponding to the non-sky area according to the second transmittance map.

Preferably,

The first incident light attenuation rate D _{non_sky} satisfies:

Among them, A is the global atmospheric light value, I ^c (x) is the pixel value of the pixel x in the haze image in channel c, {r, g, b} represents the three-color channel, and Ω _{non_sky} represents the non- _{uniformity of the} haze image. Sky area.

Preferably, in the step S2, the obtaining the non-sky area of the haze image includes:

S21: Convert the haze image into a grayscale image;

S22: Acquire a gradient map of the grayscale image according to edge detection, and convert to generate a binary image;

S23. Perform minimum filtering on the binary image to obtain a non-sky area of the haze image.

Preferably, in the step S22, the acquiring a gradient map of the grayscale image according to edge detection and converting to generate a binary image includes:

S221: Convert the gradient map according to a preset gradient threshold to generate a first binary map;

S222: Convert the grayscale image according to a preset brightness threshold to generate a second binary image;

S223. Combine the first binary graph and the second binary graph to generate the binary graph.

Preferably, the gradient threshold value is the average value of the gradient of the gradient map; the brightness threshold value is the average value of the brightness of the gray image.

Preferably, in the step S21, the converting the haze image into a grayscale image includes:

Acquiring an RGB image of the haze image, and adjusting the RGB image according to a preset RGB ratio to acquire the corresponding grayscale image; and/or

In the step S22, the acquiring the gradient map of the grayscale image according to edge detection includes:

Adopting Sobel operator, Prewitt operator or Laplacian operator to perform edge detection to obtain the initial gradient map of the gray image;

Median filtering is applied to the initial gradient image to obtain the final gradient image of the grayscale image.

The present invention also constructs a haze image quality evaluation system, including:

An acquiring unit for acquiring a haze image;

A first processing unit, configured to obtain a first transmittance map corresponding to the haze image based on a dark channel prior method;

The second processing unit is used to obtain the non-sky area of the haze image;

A third processing unit, configured to obtain a first incident light attenuation rate corresponding to the non-sky area according to the first transmittance map and the non-sky area;

The output unit is configured to output the image quality of the haze image according to the first incident light attenuation rate.

The present invention also constructs a computer storage medium on which a computer program is stored, and when the computer program is executed by a processor, the method for evaluating the quality of the haze image described in any one of the above is realized.

The present invention also constructs an electronic device, including a memory and a processor;

The memory is used to store computer programs;

The processor is configured to execute the computer program to implement the haze image quality evaluation method as described in any one of the above.

The implementation of the haze image quality evaluation method, system, storage medium and electronic equipment of the present invention has the following beneficial effects: by eliminating the sky area, the interference of the sky area on the haze image quality evaluation is reduced to optimize the image quality evaluation effect.

Description of the drawings

The present invention will be further described below in conjunction with the accompanying drawings and embodiments. In the accompanying drawings:

FIG. 1 is a program flow chart of an embodiment of the method for evaluating haze image quality according to the present invention;

2 is a program flowchart of another embodiment of the method for evaluating the quality of a haze image of the present invention;

FIG. 3 is a program flow chart of another embodiment of the method for evaluating haze image quality according to the present invention;

Figure 4 is a schematic diagram of the identification process of non-sky areas in the haze image;

Figure 5 shows the NSDark values corresponding to the non-sky areas of different haze images;

Figure 6 and Figure 7 show the performance comparison of different haze images;

FIG. 8 is a logical block diagram of the first embodiment of the haze image quality evaluation system for detecting the sky area of the present invention.

Detailed ways

In order to have a clearer understanding of the technical features, objectives and effects of the present invention, the specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

As shown in Fig. 1, in the first embodiment of the haze image quality evaluation method of the present invention, the following steps are included:

S1. Obtain the haze image, and obtain the first transmittance map corresponding to the haze image based on the dark channel prior method; specifically, the atmospheric scattering model widely used in the field of image processing under the haze weather describes the atmospheric scattering process And the principle of ambient light attenuation. The detailed technical principle of the atmospheric scattering model under a single light source is shown in Figure 5. The model divides the light reaching the imaging device into two parts: one is directly attenuating the light, and the reflected light of the scene propagates to the imaging device. The scattering effect of particles in the air causes incident light attenuation, which is called direct light attenuation; the other part is that atmospheric light directly acts on suspended particles in the air, and is received by the imaging device after being scattered, and overlaps on the target image, which is called Additional scattered light. Usually these two parts of light exist, the lower the haze degree, the higher the proportion of the direct attenuation light in the image; the higher the haze degree, the higher the proportion of additional scattered light in the image. Based on the above, in the field of image defogging, the relationship between the foggy image and the defogging image can be represented by the following model:

I(x)=J(x)t(x)+A(1-t(x)) (1)

Among them, I(x) is the haze image, J(x) is the clear image, t(x) is the transmittance map, and A is the global atmospheric light value.

On the basis of the above, the dark channel prior method is used to estimate the first transmittance map corresponding to the haze image, and the first transmittance map can be expressed as:

Among them, A represents the global atmospheric light value, I ^c (x) represents the pixel value of the pixel x in the haze image in channel c, {r, g, b} represents the three-color channel, and Ω represents the overall area of the haze image .

The specific operation: For the non-sky area of the outdoor clear image, each pixel must have a pixel value close to 0 in the three primary color channels, that is, the elements in the channel composed of the minimum values of the image channels are close to 0. The formula is as follows:

Among them, J ^c (x) represents the pixel value of pixel x in the c channel, {r, g, b} represents the three color channels, Ω represents the overall area of the image, at this time the image is a clear image, J _dark (x) represents the image The dark channel composed of the minimum of the three color channels.

Substitute formula (3) into formula (1) to get the following formula:

Since the dark channel value of a fog-free image is close to 0, substituting formula (3) into (4) can get:

The first transmittance diagram can be obtained as formula (2).

Among them, A represents the global atmospheric light value, and the average value of the brightness of each channel can be calculated by taking the first 0.01% pixels of the haze image with the maximum brightness.

S2. Acquire the non-sky area of the haze image; specifically, the first transmittance map of the haze image acquired above corresponds to the transmittance of the entire image, and because in some scenes, for example, the sky area is cloudy or cloudy In the case of, the sky area in the haze image is too close to the haze, which will affect the judgment of the quality of the haze image. Therefore, it is necessary to further remove the sky area from the haze image to obtain the haze image The non-sky area in

S3. Obtain a first incident light attenuation rate corresponding to the non-sky area according to the first transmittance map and the non-sky area, so as to obtain the image quality of the haze image according to the first incident light attenuation rate. Specifically, after the non-sky area of the haze image is obtained, conversion can be performed according to the expression of the first transmittance map of the entire haze image that has been obtained, and the first incidence corresponding to the non-sky area in the haze image can be obtained. The light attenuation rate, which is used to obtain the image quality of the haze image through the first incident light attenuation rate, can also be understood as the image quality evaluation result of the haze image.

Further, in step S3, the specific process of calculating the first incident light attenuation rate corresponding to the non-sky area according to the first transmittance map and the non-sky area, in one embodiment, includes: obtaining the haze according to the first transmittance map The second incident light attenuation rate of the image; the first incident light attenuation rate corresponding to the non-sky area is obtained according to the second incident light attenuation rate and the non-sky area; in another embodiment, it includes: according to the first transmittance map and the non-sky area Obtain the second transmittance map corresponding to the non-sky area in the sky area; obtain the first incident light attenuation rate corresponding to the non-sky area according to the second transmittance map.

Specifically, in the atmospheric scattering model, the relationship between transmittance and scene depth and atmospheric extinction coefficient is:

t(x)=e ^-β(λ)d (6)

Among them, λ represents the wavelength, β (λ) represents the atmospheric extinction coefficient, and its physical meaning is the relative attenuation rate of electromagnetic wave radiation per unit distance in the atmosphere, and d is the distance between the observation point and the target.

Under single scattering conditions, aerosols are uniformly distributed in the air. β(λ) represents the relative attenuation rate per unit distance, β(λ)×d represents the total attenuation rate of the light from the scene to the imaging device, that is, the incident light attenuation rate. The atmospheric extinction coefficient is a parameter related to the properties and density of aerosols. It is constant when the aerosol properties are constant and uniformly distributed. Therefore, in an image with a constant scene depth d, the incident light attenuation rate D(x) is related to the haze density, and its formula is:

D(x)=β(λ)d=-ln(t(x)) (7)

Among them, D(x) represents the attenuation rate of incident light, λ represents the wavelength, β(λ) represents the atmospheric extinction coefficient, and t(x) represents the transmittance graph. That is, on the basis of the above, the second incident light attenuation rate D(x) of the haze image can be obtained according to formula (7), and converted to the corresponding first incident light attenuation rate D _{non_sky} according to the obtained non-sky area. It is also possible to first convert the first transmittance map corresponding to the haze image to obtain the second transmittance map t _{non_sky} corresponding to the non-sky area. _Based on the above, the converted formula can be:

Then, obtain the corresponding first incident light attenuation rate D _{non_sky} according to formula (7).

Further, the first incident light attenuation rate D _{non_sky} satisfies:

Among them, A is the global atmospheric light value, I ^c (x) is the pixel value of the pixel x in the haze image in channel c, {r, g, b} represents the three-color channel, and Ω _{non_sky} represents the non- _{uniformity of the} haze image. Sky area. The specific process can refer to the above description. The attenuation rate D _{non_sky} of the incident light in the non-sky area can be used as the evaluation value of the haze image quality, and the evaluation value can be defined as the NSDark value.

Optionally, as shown in FIG. 2, in step S2, obtaining the non-sky area of the haze image includes:

S21. Convert the haze image to a grayscale image; specifically, convert the acquired color haze image into a grayscale image. During the conversion process, in order to retain more edge information, a decolorization algorithm with contrast retention function can be used. For example, an energy function decolorization algorithm is used.

S22. Obtain a gradient map of the grayscale image according to the edge detection, and convert to generate a binary image; specifically, a variety of edge detection operators can be used to solve the gradient map. The obtained gradient map is divided into pixels to obtain a binary map.

S23. Perform minimum filtering on the binary image to obtain a non-sky area of the haze image. Specifically, the minimum filter is used to denoise, and the local scattered noise of the image is covered to make the overall segmentation more reasonable. Finally, the segmented non-sky area and sky area are obtained, so as to calculate the corresponding incident light attenuation rate according to the non-sky area of the haze image. In some embodiments, the diameter of the minimum filter can be set to 3.

Further, as shown in FIG. 3, in step S22, acquiring a gradient map of a grayscale image according to edge detection, and converting to generate a binary image includes:

S221: Convert the gradient map according to the preset gradient threshold to generate a first binary map;

S222: Convert the gray image according to the preset brightness threshold to generate a second binary image;

S223. Combine the first binary graph and the second binary graph to generate a binary graph.

Specifically, under the interference of clouds, edge information may appear in the sky area. The interference of artificial light sources may also cause the brightness of some areas on the ground to exceed the threshold. Therefore, a preset gradient threshold and a preset brightness threshold need to be set. Convert the first binary image corresponding to the gradient map according to the preset gradient, generate the second binary image corresponding to the brightness of the grayscale image according to the preset brightness threshold, and then merge the gradient and the brightness to generate the final binary image Figure.

Further, the gradient threshold value is the gradient average value of the gradient map; the brightness threshold value is the brightness average value of the grayscale image. Specifically, based on the above, the preset gradient threshold and the preset brightness threshold may be set as the gradient average value of the gradient map and the brightness average value of the grayscale image, respectively. Then, take 0 or 255 for the divided pixels to obtain a binary image. Since the goal of sky recognition is to find the average transmittance of non-sky areas, the slight error at the junction of the sky and the ground caused by the setting of the preset gradient threshold and the preset brightness threshold has little effect on the calculation results.

Optionally, in step S21, converting the haze image into a grayscale image includes: obtaining an RGB image of the haze image, and adjusting the RGB image according to a preset RGB ratio to obtain the corresponding grayscale image; specifically, considering some algorithms The performance improvement of the sky detection effect is limited. In order to improve the efficiency, in some embodiments, the conversion of the gray image may adopt a method of converting an RGB image into a gray image proportionally. Among them, the preset RGB ratio can be used, the weight of the R channel is set to 0.299, the weight of the G channel is set to 0.587, and the weight of the B channel is set to 0.114.

Optionally, in step S22, acquiring the gradient map of the grayscale image according to the edge measurement includes: using Sobel operator, Prewitt operator or Laplacian operator to perform edge detection to acquire the initial gradient map of the grayscale image; The gradient map uses median filtering to obtain the final gradient map of the grayscale image. Specifically, obtaining the gradient map of the gray image through a specific edge detection operator can adopt any one of the Sobel operator, the Prewitt operator, or the Laplacian operator. At the same time, when the gradient map of the grayscale image is calculated based on the above-mentioned operator, the gradation map can be denoised, so that subsequent operations can be performed on the denoised gradient map. Take the specific Laplacian operator as an example.

Use Laplacian operator for edge detection, the formula is:

Among them, L(f) represents the Laplacian detection value, f represents the gray value of the pixel on the image, and x and y represent the abscissa and ordinate of the pixel respectively.

The discrete form of Laplace operator is:

L(f)=[f(x+1,y)+f(x-1,y)+f(x,y-1)]-4f(x,y) (11)

Among them, L(f) represents the Laplacian detection value, and f(x, y) represents the gray value of the pixel with the coordinate (x, y) on the gray image.

The sky area is bright and smooth as a whole, and it generally appears as a white area in the Laplacian edge detection image. However, the edge of the cloud, the dust on the imaging device, etc., may all cause interference noise. In order to reduce noise, median filtering is needed to reduce noise. By sliding the window, the value of the center pixel of the window is taken as the window average value, which can effectively eliminate noise interference.

As shown in Figure 4, it shows the recognition process of non-sky areas in the haze image. Where A represents the original image, by converting it into a gray image B, the edge detection of the gray image B is performed by Laplacian operator, and the conversion is performed to obtain the binary image C, and the minimum value filter is performed on the binary image C to obtain the final The non-sky area corresponding to the haze object corresponds to the white area in Figure D.

Figure 5 shows the attenuation rate of incident light corresponding to the non-sky area of different haze images in the HID2018 database, that is, the NSDark value. It can be seen from Figure 5 that the NSDark values of Figure 5 (d) and (e) are small, and the image is less affected by haze. Figure 5(g) and (h) show that the NSDark value is large, and the image is seriously affected by the haze. The NSDark value is consistent with the subjective evaluation result.

At the same time, it can be understood that the process of obtaining the attenuation rate of incident light in the non-sky area D _{non_sky} corresponding to the haze image quality evaluation value, that is, the NSDark value, to evaluate the haze image image quality is a reference-free image quality evaluation process. Here, you can choose the commonly used non-reference image quality evaluation methods for comparison. Specifically, three commonly used non-reference image quality evaluation methods, entropy function (Entropy), gray-scale variance (SMD) and Laplacian gradient, are used for performance comparison. At the same time, the subjective score (MOS) of the haze image is used as a benchmark for comparison. Refer to the following table for details:

Table 1 Comparison of performance of haze image quality evaluation methods

Image number	MOS	Entropy	SMD	Laplacian		NSDark
1	3.2500	7.016	3597.809	563.637	0.159
2	2.2500	6.933	2593.533	224.517	0.295
3	4.1875	6.88	4401.999	796.295	0.103
4	5.0000	7.472	2,897.599	1994.336	0.092
5	5.0000	7.466	2925.595	2097.067	0.087
6	3.7500	6.7	2746.869	783.437	0.182
7	4.0625	6.742	3,252.525	1006.469	0.197
8	2.6250	6.933	2641.013	395.499	0.241
9	4.9375	7.459	2,758.44	2078.917	0.113
10	3.5625	6.666	3203.094	756.269	0.113
11	5.0000	7.481	2967.521	2211.008	0.206
12	4.6250	7.44	2,749.59	1,762.722	0.113
13	4.9375	7.272	2265.036	1711.025	0.156
14	3.5000	6.467	2529.061	727.178	0.261
15	2.2500	6.636	2,137.632	309.936	0.362
16	2.4375	6.682	2516.218	346.830	0.276
17	1.6250	6.818	2035.718	115.954	0.532
18	1.6875	7.072	2,654.255	86.571	0.575
19	1.8125	7.096	2739.016	116.050	0.405

20	4.7500	7.29	2,398.71	1701.762	0.15
twenty one	4.8750	7.348	2367.826	1,680.980	0.146
twenty two	2.7500	6.977	3036.286	307.178	0.259
twenty three	2.1250	6.424	1,477.099	305.824	0.439
twenty four	3.1250	6.604	2,285.189	652.431	0.278
25	5.0000	7.353	2238.621	1822.243	0.15
26	4.6250	7.235	2917.264	1,189.022	0.145
27	2.7500	6.642	2752.827	403.511	0.244

Randomly extract some MOS values and NSDark values from Table 1 for analysis. For example, image number 7 has a slight haze, and image number 8 has more severe haze. The NSDark value of image number 7 is 0.197, and the NSDark value of image number 7 is 0.241. From the NSDark value, it can be judged that the smog of image 8 is more serious than that of image 7.

The data in Table 1 is normalized to obtain the performance comparison of Mos, NSDark and Laplacian shown in Figure 6, where A1 represents the MOS value, A2 represents the NsDark value, A3 represents the Laplacian value, and the Mos and SMD shown in Figure 7. Compared with Entropy performance, B1 is the MOS value, B2 is the Entropy value, and B3 is the SMD value. It can be seen from Figure 6 and Figure 7 that the overall trend of the NSDark method is consistent with the Laplacian gradient method and MOS, and the monotonicity of its change is better than that of the SMD and Entropy methods.

Further, in order to compare different image quality evaluation methods, the mean square error between different image quality evaluation methods and MOS values is calculated. Among them, the mean square error between the Laplician value and the MOS value is 4.14, the mean square error between the Entropy value and the MOS value is 1.13, the mean square error between the SMD value and the MOS value is 1.91, and the mean square error between the NSDark value and the MOS value is 0.5. The mean square error of NSDark and MOS values of all images is the smallest, and the changes of NSDark and MOS values are relatively consistent. Therefore, the evaluation result of NSDark value is slightly better than other quality evaluation methods.

In addition, in the embodiment shown in FIG. 8, a haze image quality evaluation system of the present invention includes:

The obtaining unit 10 is used to obtain a haze image;

The first processing unit 20 is configured to obtain a first transmittance map corresponding to the haze image based on the dark channel prior method;

The second processing unit 30 is used to obtain the non-sky area of the haze image;

The third processing unit 40 is configured to obtain the first incident light attenuation rate corresponding to the non-sky area according to the first transmittance map and the non-sky area;

The output unit 50 is configured to output the image quality of the haze image according to the first incident light attenuation rate.

Specifically, the specific coordination operation process between the units of the haze image quality evaluation system can refer to the above-mentioned haze image quality evaluation method, which will not be repeated here.

In addition, an electronic device of the present invention includes a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program to implement any of the above haze image quality evaluation methods. Specifically, according to an embodiment of the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart. In such an embodiment, when the computer program can be downloaded and installed by an electronic device and executed, it executes the above-mentioned functions defined in the method of the embodiment of the present invention. The electronic device in the present invention can be a terminal such as a notebook, a desktop computer, a tablet computer, a smart phone, etc., or a server.

In addition, a computer storage medium of the present invention has a computer program stored thereon, and when the computer program is executed by a processor, any one of the above haze image quality evaluation methods is realized. Specifically, it should be noted that the above-mentioned computer-readable medium of the present invention may be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. The computer-readable storage medium may be, for example, but not limited to, an electric, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections with 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 fiber, portable compact disk read only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination of the above. In the present invention, the computer-readable storage medium may be any tangible medium that contains or stores a program, and the program may be used by or in combination with an instruction execution system, apparatus, or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as a part of a carrier wave, and a computer-readable program code is carried therein. This propagated data signal can take many forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer-readable signal medium may also be any computer-readable medium other than the computer-readable storage medium. The computer-readable signal medium may send, propagate, or transmit the program for use by or in combination with the instruction execution system, apparatus, or device . The program code contained on the computer-readable medium can be transmitted by any suitable medium, including but not limited to: wire, optical cable, RF (Radio Frequency), etc., or any suitable combination of the above.

The above-mentioned computer-readable medium may be included in the above-mentioned electronic device; or it may exist alone without being assembled into the electronic device.

It is understandable that the above examples only express the preferred embodiments of the present invention, and the descriptions are more specific and detailed, but they should not be construed as limiting the scope of the present invention; it should be pointed out that for those of ordinary skill in the art In other words, without departing from the concept of the present invention, the above technical features can be freely combined, and several modifications and improvements can be made. These all belong to the scope of protection of the present invention; therefore, everything that follows the scope of the claims of the present invention All equivalent changes and modifications shall fall within the scope of the claims of the present invention.

Claims (10)

1. A method for evaluating haze image quality, characterized in that it comprises the following steps:

   S1. Acquire a haze image, and acquire a first transmittance map corresponding to the haze image based on a dark channel prior method;

   S2. Obtain a non-sky area of the haze image;

   S3. Acquire a first incident light attenuation rate corresponding to the non-sky area according to the first transmittance map and the non-sky area, so as to obtain the image quality of the haze image according to the first incident light attenuation rate .
2. The method for evaluating haze image quality according to claim 1, wherein:

   In the step S3, the calculating the first incident light attenuation rate corresponding to the non-sky area according to the first transmittance map and the non-sky area includes:

   Acquiring the second incident light attenuation rate of the haze image according to the first transmittance map;

   Obtaining the first incident light attenuation rate corresponding to the non-sky area according to the second incident light attenuation rate and the non-sky area; or

   Acquiring a second transmittance map corresponding to the non-sky area according to the first transmittance map and the non-sky area;

   Acquire the first incident light attenuation rate corresponding to the non-sky area according to the second transmittance map.
3. The method for evaluating haze image quality according to claim 1, wherein:

   The first incident light attenuation rate D _{non_sky} satisfies:

   

   Among them, A is the global atmospheric light value, I ^c (x) is the pixel value of the pixel x in the haze image in channel c, {r, g, b} represents the three-color channel, and Ω _{non_sky} represents the non- _{uniformity of the} haze image. Sky area.
4. The method for evaluating the quality of a haze image according to claim 2, wherein, in the step S2, the obtaining the non-sky area of the haze image comprises:

   S21: Convert the haze image into a grayscale image;

   S22: Acquire a gradient map of the grayscale image according to edge detection, and convert to generate a binary image;

   S23. Perform minimum filtering on the binary image to obtain a non-sky area of the haze image.
5. The method for evaluating the quality of a haze image according to claim 4, wherein in the step S22, the obtaining a gradient map of the grayscale image according to edge detection, and converting to generate a binary image comprises:

   S221: Convert the gradient map according to a preset gradient threshold to generate a first binary map;

   S222: Convert the grayscale image according to a preset brightness threshold to generate a second binary image;

   S223. Combine the first binary graph and the second binary graph to generate the binary graph.
6. The method for evaluating the quality of a haze image according to claim 4, wherein the gradient threshold value is the average value of the gradient of the gradient map; the brightness threshold value is the average value of the brightness of the gray image.
7. The method for evaluating the quality of a haze image according to claim 4, wherein in the step S21, the converting the haze image into a grayscale image comprises:

   Acquiring an RGB image of the haze image, and adjusting the RGB image according to a preset RGB ratio to acquire the corresponding grayscale image; and/or

   In the step S22, the acquiring the gradient map of the grayscale image according to edge detection includes:

   Adopting Sobel operator, Prewitt operator or Laplacian operator to perform edge detection to obtain the initial gradient map of the gray image;

   Median filtering is applied to the initial gradient image to obtain the final gradient image of the grayscale image.
8. A haze image quality evaluation system, characterized in that it includes:

   An acquiring unit for acquiring a haze image;

   A first processing unit, configured to obtain a first transmittance map corresponding to the haze image based on a dark channel prior method;

   The second processing unit is used to obtain the non-sky area of the haze image;

   A third processing unit, configured to obtain a first incident light attenuation rate corresponding to the non-sky area according to the first transmittance map and the non-sky area;

   The output unit is configured to output the image quality evaluation of the haze image according to the first incident light attenuation rate.
9. A computer storage medium having a computer program stored thereon, wherein the computer program implements the haze image quality evaluation method according to any one of claims 1 to 7 when the computer program is executed by a processor.
10. An electronic device, characterized in that it includes a memory and a processor;

    The memory is used to store computer programs;

    The processor is configured to execute the computer program to implement the haze image quality evaluation method according to any one of claims 1 to 7.

Images