Low-light image enhancement algorithm based on atmospheric scattering model

Abstract

The application discloses a low-brightness image enhancement algorithm based on an atmospheric scattering model, which comprises the following steps: calculating the maximum saturation promotion value of different color pixels in an original low-brightness image; establishing a mathematical model between the atmospheric light and the saturation promotion value to estimate the atmospheric light of the original low-brightness image, and performing smoothing processing on the estimated atmospheric light to obtain a locally smooth atmospheric light; removing the atmospheric light in the original low-brightness image to obtain an image without the atmospheric light; using the maximum brightness to estimate the brightness map of the image, performing smoothing processing on the estimated brightness map to obtain a smooth brightness map; and inputting the locally smooth atmospheric light and the smooth brightness map into a retinex model to obtain a reflection image with synchronous enhancement of brightness and saturation. The application can realize the enhancement of saturation and brightness for low-brightness images in different environments, effectively ensure the removal of the hazy visual effect of the image in a low-brightness scene containing atmospheric light, and significantly widen the application scene of the low-light image algorithm.

Description

Technical Field

[0001] This invention belongs to the field of visible light imaging and relates to a low-brightness image enhancement algorithm based on an atmospheric scattering model. Background Technology

[0002] Due to limitations in camera performance or low-light environments, the captured images exhibit low illumination. Low light blurs scene details, destroys the original characteristics of colors, reduces image contrast, and fails to provide a good visual experience. Furthermore, it can affect the use of other image processing tasks, such as infrared and visible light image fusion, object detection and tracking, and image segmentation.

[0003] To address this problem, scholars both domestically and internationally have conducted extensive research on low-brightness image enhancement algorithms, proposing numerous effective algorithms. However, existing low-brightness image enhancement algorithms still have some limitations:

[0004] (1) Traditional low-brightness image enhancement algorithms often focus on brightness enhancement, noise removal, and contrast enhancement, while neglecting the role of saturation enhancement. When applied to low-brightness scenes with a small amount of atmospheric light, the lack of saturation enhancement capability results in a hazy visual appearance, which is difficult to meet the requirements of practical applications.

[0005] (2) While deep learning-based low-brightness enhancement algorithms have good brightness enhancement capabilities, their saturation enhancement capabilities are limited. Supervised learning-based algorithms rely on paired input data for their effectiveness. As the application scenario changes, the algorithm's adaptability degrades. Unsupervised learning-based algorithms are free from dependence on paired data, but their performance is often inferior to supervised learning-based algorithms. Furthermore, deep learning-based low-brightness image enhancement algorithms have excessively high hardware requirements, limiting their practical application scope. Summary of the Invention

[0006] Purpose of the invention: The purpose of this invention is to provide a low-brightness image enhancement algorithm based on an atmospheric scattering model, which can enhance the saturation and brightness of low-brightness images under different environments.

[0007] Technical Solution: To achieve the above objectives, the present invention provides a low-brightness image enhancement algorithm based on an atmospheric scattering model, comprising the following steps:

[0008] Step 1: Calculate the maximum saturation boost value of different color pixels in the original low-brightness image;

[0009] Step 2: Establish a mathematical model between atmospheric light and saturation boost value to estimate the atmospheric light Q of the original low-brightness image, and use the norm to smooth the estimated atmospheric light Q to obtain locally smooth atmospheric light Q. o ;

[0010] Step 3: Utilizing locally smooth atmospheric light Q o Atmospheric light is removed from the original low-brightness image to obtain image S1 with atmospheric light removed. The brightness map t of image S1 is estimated using the maximum brightness prior, and the estimated brightness map t is smoothed using the norm to obtain a smooth brightness map t. o ;

[0011] Step 4: Transform the locally smooth atmospheric light Q o and smooth brightness diagram t o The image is input into the Retinex model to obtain a reflection image with simultaneous enhancement of brightness and saturation.

[0012] Preferably, the original low-brightness image mentioned in step one is a visible light image with RGB channels.

[0013] Preferably, the maximum saturation enhancement value of different color pixels in the original low-brightness image described in step one is calculated using an image color classification algorithm, including establishing a mathematical model for the maximum saturation enhancement value and using the mathematical model to calculate the color category w. i Maximum saturation boost

[0014] Preferably, the mathematical model for the maximum saturation enhancement value is shown in the following formula:

[0015]

[0016] In the formula, S represents the original low-brightness image, and k means This represents the k-means classification algorithm, w i To represent the color categories, the original low-brightness image has n color categories, denoted by w1, w2, ... w n express; Indicates color category w i The saturation of the pixels.

[0017] Preferably, the step two, which establishes a mathematical model between atmospheric light and saturation enhancement value to estimate the atmospheric light Q of the original low-brightness image, specifically involves:

[0018] The mathematical model between atmospheric light and the maximum saturation enhancement value is shown in the following equation:

[0019]

[0020] In the formula, S max ΔS represents the maximum pixel value in the RGB channels of the original low-brightness image. satThis represents the maximum saturation boost value for all color categories in the original low-brightness image, k0 represents the atmospheric light adjustment factor, M represents the ratio of the minimum to the maximum pixel value of the RGB channels in the original low-brightness image, k is an adjustment factor between 0 and 1, and ε is a constant to prevent the denominator from being 0.

[0021] Preferably, the norm model described in step two smooths the estimated atmospheric light as shown in the following equation:

[0022]

[0023] In the formula, λ1 represents the constraint factor, which takes values ​​ranging from 0 to 1.

[0024] Preferably, the image S1 = SQ after removing atmospheric light in step three is... o S represents the original low-brightness image.

[0025] Preferably, the brightness map t = max of image S1 in step three. rgb (S1).

[0026] Preferably, the norm model described in step three smooths the estimated brightness map t as shown in the following equation:

[0027]

[0028] In the formula, λ1 represents the constraint factor, which ranges from 0 to 1.

[0029] Preferably, the step four involves smoothing the locally glossy atmospheric light Q. o and smooth brightness diagram t o When input into the Retinex model, the resulting reflection image with simultaneous enhancement of brightness and saturation is shown in the following formula:

[0030] Q o =R*t o

[0031] In the formula, R represents the output image with high saturation and brightness after the original low-brightness image is enhanced.

[0032] Beneficial effects: The algorithm described in this invention can enhance the saturation and brightness of low-brightness images under different environments, effectively ensuring the removal of hazy visual effects in low-brightness scenes containing atmospheric light, and significantly broadening the application scenarios of low-light image algorithms. Attached Figure Description

[0033] Figure 1 This is a flowchart of the low-brightness image enhancement algorithm based on the atmospheric scattering model of the present invention.

[0034] Figure 2This is a flowchart illustrating the specific algorithm for low-brightness image enhancement based on the atmospheric scattering model of this invention. Detailed Implementation

[0035] The invention will now be described in further detail with reference to the accompanying drawings.

[0036] Please see Figure 1 and Figure 2 As shown, the low-brightness image enhancement algorithm based on the atmospheric scattering model of the present invention includes the following steps:

[0037] S1. Calculate the maximum saturation enhancement value of different color pixels in the original low-brightness image using an image color classification algorithm. This low-brightness image is a visible light image with RGB channels, and the atmospheric light in the environment in which the low-brightness image is located is very weak, specifically:

[0038] A mathematical model for the maximum saturation enhancement value is established, which is expressed as follows:

[0039]

[0040] In the formula, S refers to the original low-brightness image, and k means Refers to the k-means classification algorithm, w i The original low-brightness image is divided into n color categories, denoted by w1, w2, ... w n express. It is a color category w i The saturation of the pixels, The color category is w i Maximum saturation.

[0041] S2. Establish a mathematical model between atmospheric light and saturation enhancement value to estimate the atmospheric light Q of the original low-brightness image, and use the norm to smooth the estimated atmospheric light Q to obtain locally smooth atmospheric light Q. o Specifically:

[0042] S21. Establish a mathematical model between atmospheric light and the maximum saturation enhancement value, as shown in the following formula:

[0043]

[0044] In the formula, S max ΔS represents the maximum pixel value in the RGB channels of the original low-brightness image. satThis represents the maximum saturation boost value for all color categories in the original low-brightness image, k0 represents the atmospheric light adjustment factor, M represents the ratio of the minimum to the maximum pixel value of the RGB channels in the original low-brightness image, k is an adjustment factor between 0 and 1, and ε is a very small constant to prevent the denominator from being 0, which is set to 0.00001 here.

[0045] S21. The estimated atmospheric light is smoothed using a norm model to obtain locally smoothed atmospheric light, as shown in the following equation:

[0046]

[0047] In the formula, λ1 is a constraint factor, and its range is 0-1.

[0048] S3, utilizing locally smooth atmospheric light Q o Atmospheric light is removed from the original low-brightness image to obtain image S1 with atmospheric light removed. The brightness map t of image S1 is estimated using the maximum brightness prior, and the estimated brightness map t is smoothed using the norm to obtain a smooth brightness map t. o

[0049] S31, Image S1 after removing atmospheric light:

[0050] S1 = SQ o

[0051] S32. Based on S1, extract the brightness map using the maximum brightness prior, as shown in the following formula:

[0052] t = max rgb (S1)

[0053] S33. Smooth the estimated brightness map using a norm model, as shown in the following formula:

[0054]

[0055] In the formula, t o Refers to the output of a smooth brightness map.

[0056] S4, making the localized smooth atmospheric light Q o and smooth brightness diagram t o The input is fed into the retinex model to obtain a reflectance image with simultaneous enhancement of brightness and saturation, specifically:

[0057] Q o =R*t o

[0058] In the formula, R represents the output of the enhanced original low-brightness image, resulting in an output image with high saturation and brightness.

Claims

1. A low-brightness image enhancement algorithm based on an atmospheric scattering model, characterized in that, Includes the following steps: Step 1: Calculate the maximum saturation boost value of different color pixels in the original low-brightness image; Step 2: Establish a mathematical model between atmospheric light and saturation boost values ​​to estimate the atmospheric light in the original low-brightness image. And use norm models to estimate atmospheric light Smoothing is performed to obtain locally smooth atmospheric light. ; Step 3: Utilizing locally smooth atmospheric light The atmospheric light in the original low-brightness image is removed to obtain image S1 with atmospheric light removed; the brightness map of image S1 is estimated using the maximum brightness prior. And use the norm model to analyze the estimated brightness map Smoothing is performed to obtain a smooth brightness map. ; Step 4: Apply localized smooth atmospheric light and smooth brightness map The image is input into the Retinex model to obtain a reflection image with simultaneous enhancement of brightness and saturation. This involves using an image color classification algorithm to calculate the maximum saturation enhancement value of different color pixels in the original low-brightness image described in step one, including establishing a mathematical model for the maximum saturation enhancement value and using this mathematical model to calculate the color category. Maximum saturation boost ; The mathematical model for the maximum saturation enhancement value is shown in the following formula: ， ， In the formula, This represents the original low-brightness image. This refers to the k-means classification algorithm. To represent the color categories, the original low-brightness image has n color categories, which are represented by... express; Indicates color category as The saturation of the pixels; Step two involves establishing a mathematical model between atmospheric light and saturation boost values ​​to estimate the atmospheric light in the original low-brightness image. Specifically: The mathematical model between atmospheric light and the maximum saturation enhancement value is shown in the following equation: ， In the formula, This represents the maximum pixel value in the RGB channels of the original low-brightness image. This represents the maximum saturation boost value for all color categories contained in the original low-brightness image. Indicates atmospheric light modulator. This represents the ratio of the minimum to the maximum pixel value in the RGB channels of the original low-brightness image. It is a regulatory factor between 0 and 1. This is to prevent constants with a denominator of 0.

2. The low-brightness image enhancement algorithm based on the atmospheric scattering model according to claim 1, characterized in that, The original low-brightness image mentioned in step one is a visible light image with RGB channels.

3. The low-brightness image enhancement algorithm based on the atmospheric scattering model according to claim 1, characterized in that, The norm model described in step two smooths the estimated atmospheric light as shown in the following equation: ， In the formula, This represents the constraint factor, with a value range of 0-1.

4. The low-brightness image enhancement algorithm based on the atmospheric scattering model according to claim 1, characterized in that, The image with atmospheric light removed as described in step three. S represents the original low-brightness image.

5. The low-brightness image enhancement algorithm based on the atmospheric scattering model according to claim 1, characterized in that, The brightness map of image S1 mentioned in step three .

6. The low-brightness image enhancement algorithm based on the atmospheric scattering model according to claim 1, characterized in that, The norm model described in step three is used to estimate the brightness map. The smoothing process is shown in the following formula: ， In the formula, This represents the constraint factor, with a value range of 0-1.

7. The low-brightness image enhancement algorithm based on the atmospheric scattering model according to claim 1, characterized in that, Step four describes the process of applying locally smooth atmospheric light. and smooth brightness map When input into the Retinex model, the resulting reflection image with simultaneous enhancement of brightness and saturation is shown in the following formula: ， In the formula, This represents the output image with high saturation and brightness after the original low-brightness image has been enhanced.

Images