Low-illumination image enhancement method based on joint guidance total variation

Abstract

The invention discloses a low-illumination image enhancement method based on joint guidance total variation. The low-illumination image enhancement method comprises the following steps: acquiring a low-illumination image; converting the low-illumination image from an RGB space to an HSV color space, and performing Retinex variation processing to obtain a reflection component and an illumination component; and adjusting the illumination component based on arc tangent transformation, and multiplying the adjusted illumination component by the reflection component to obtain an enhanced image. According to the method provided by the invention, the brightness of the input image can be effectively enhanced, the noise is well suppressed, and meanwhile, the color contrast and detail information are kept.

Description

Technical Field

[0001] The present invention belongs to the technical field of low-light image enhancement, and particularly relates to a low-light image enhancement method based on jointly guided total variation. Background Art

[0002] Since the ambient brightness of the object during shooting is relatively dark, the captured image is difficult to have as clear and rich information as the image captured under normal lighting conditions. Such images are called low-light images. Low-light images usually exhibit phenomena such as low brightness, decreased contrast, obvious noise, and blurred details. These phenomena not only reduce the perception ability of the human eye but also reduce the recognition, analysis, and understanding abilities of intelligent perception devices, thus having a negative impact on people's production and life. Therefore, enhancing low-light images has important research significance and application value.

[0003] However, low-light image enhancement faces many challenges, such as how to enhance brightness and contrast without losing details, maintain color fidelity of the image while suppressing noise, and meet real-time computing requirements while ensuring the enhancement effect.

[0004] Current research methods are mainly divided into traditional algorithms and deep learning-based methods. Traditional algorithms such as histogram equalization and gamma correction enhance brightness by adjusting pixel distribution, but are prone to problems such as noise amplification or local over-enhancement. Deep learning methods have shown great potential in recent years. By learning the mapping from low-light to high-light images through convolutional neural networks, the enhancement effect can be significantly improved; generative adversarial networks generate more natural enhanced images through the game between the generator and the discriminator. However, due to the difficulty in obtaining the image pairs composed of low-light images and corresponding normal-light images, there is a lack of support for paired samples when training supervised low-light image enhancement networks. In addition, overfitting occurs when training the network on paired image datasets.

[0005] Aiming at the above problems in the prior art, it is urgent to propose a low-light image enhancement method based on jointly guided total variation. Summary of the Invention

[0006] To solve the above technical problems, the present invention proposes a low-light image enhancement method based on jointly guided total variation to solve the problems existing in the above prior art.

[0007] To achieve the above object, the present invention provides a low-light image enhancement method based on jointly guided total variation, including the following steps:

[0008] Obtain a low-light image;

[0009] Convert the low-light image from the RGB space to the HSV color space, and perform Retinex variational processing to obtain the reflection component and the illumination component;

[0010] Adjust the illumination component based on the arctangent transformation, and multiply the adjusted illumination component and the reflection component to obtain the enhanced image.

[0011] Optionally, the process of performing Retinex variational processing on the low-light image includes:

[0012] Keep the H component and the S component of the HSV color space unchanged, perform Retinex variational processing on the V component, use the joint guided total variation as the illumination component regularization term to constrain the illumination component, and use the logarithmic mean local variance as the reflection component regularization term to constrain the reflection component.

[0013] Optionally, the calculation formula of the joint guided total variation is as follows:

[0014]

[0015] where, H(I) p represents the joint guided total variation, ε is a decimal to avoid division by zero, ||·|| 1 represents the 1-norm, D x (p) and D y (p) represent the total variation of the window along the horizontal axis and the vertical axis of the pixel p, respectively.

[0016] Optionally, the calculation formula of the logarithmic mean local variance is as follows:

[0017]

[0018] where, Ω represents a 3*3 image block in the image, and are the gradient operators along the horizontal axis and the vertical axis respectively, and R is the reflection component in the image.

[0019] Optionally, the calculation formula for adjusting the illumination component based on the arctangent transformation is as follows:

[0020]

[0021] where, is the enhanced illumination component obtained using the arctangent transformation, I(x, y) is the illumination component obtained by Retinex variational processing, and δ is the parameter controlling the visibility.

[0022] The present invention also provides a computer device, including a memory, a processor, and a computer program stored on the memory, where the processor executes the computer program to implement the steps of the method.

[0023] The present invention also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the method are implemented.

[0024] The present invention also provides a computer program product, including a computer program, and when the computer program is executed by a processor, the steps of the method are implemented.

[0025] Compared with the prior art, the present invention has the following advantages and technical effects:

[0026] Based on the Retinex theory, the present invention proposes a low-light image enhancement method based on joint guided total variation. This method uses joint guided total variation as the illumination regularization term in the Retinex variation, and can obtain an illumination image with clear structure and smooth detail texture; uses the logarithmic mean local variance as the reflection component regularization term, and can obtain a reflection image with clear texture. In addition, the present invention applies the arctangent transform to enhance the color and contrast of extremely low illumination areas, and can effectively solve the problem of color distortion in extremely low illumination areas.

[0027] The algorithm proposed by the present invention has good effects in enhancing image contrast, maintaining image naturalness, reducing image noise, and reducing image distortion. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The drawings constituting a part of this application are used to provide a further understanding of this application. The schematic embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation to this application. In the drawings:

[0029] Figure 1 It is a flowchart of the method according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0030] It should be noted that, without conflict, the embodiments in this application and the features in the embodiments can be combined with each other. The following will refer to the drawings and combine the embodiments to detail this application.

[0031] It should be noted that the steps shown in the flowchart of the drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0032] Embodiment 1

[0033] As Figure 1As shown in the figure, in this embodiment, a low-light image enhancement method based on joint-guided total variation is provided, including the following steps:

[0034] Step 1: Convert the input image from the RGB color space to the HSV color space, keep the H component and the S component unchanged, and perform Retinex variational processing on the V component in the space to obtain the reflection component and the illumination component;

[0035] Step 2: Adjust the illumination component using the arctangent transformation;

[0036] Step 3: Multiply the adjusted illumination component and the reflection component to obtain the enhanced image;

[0037] Step 4: Convert the enhanced image from the HSV space back to the RGB color space to obtain the output image.

[0038] Glossary:

[0039] RGB color space:

[0040] The RGB color space is an additive color model based on light, consisting of three colors: red (R), green (G), and blue (B). By combining different intensities of red, green, and blue light, various colors can be produced. The RGB color space is widely used in digital image processing and display devices, such as computer monitors, mobile phone screens, etc.

[0041] HSV color space:

[0042] The HSV color space is a color model based on human visual perception, consisting of three components: hue, saturation, and value.

[0043] Among them, the hue represents the type of color, which is an angular value ranging from 0 to 360 degrees. In OpenCV, the hue value is usually mapped between 0 and 180 degrees. The saturation represents the purity of the color, ranging from 0 to 100%. 0% represents no color (i.e., gray), and 100% represents a fully saturated color. The value represents the brightness of the color, ranging from 0 to 100%. 0% represents black, and 100% represents white.

[0044] The HSV color space is particularly suitable for applications such as color segmentation, color detection, and filtering in image processing because it is more in line with human intuitive perception of colors.

[0045] Retinex variational processing:

[0046] Retinex variational processing is an image enhancement method based on the Retinex theory. By decomposing an image into an illumination component and a reflection component, it improves the visual effect of the image. Retinex variational processing has a wide range of applications in the field of image enhancement, including high-dynamic-range image tone mapping, non-uniform illumination enhancement, low-light image enhancement, etc.

[0047] Arc Tangent Transformation:

[0048] Arc Tangent Transformation is a non-linear transformation method used in image processing, mainly for enhancing the contrast and details of an image. It maps the pixel values of an image through the arctangent function (arctan), making the dynamic range of the image more uniform, thereby improving the visual effect of the image. Arc Tangent Transformation can effectively adjust the illumination component of an image, enhance the contrast and details of the image, and is particularly suitable for enhancing low-light images.

[0049] Total Variation (TV):

[0050] Total Variation is a regularization method used in image processing, mainly for image denoising and edge preservation. Total Variation regularization preserves the edge information of an image by minimizing the gradient magnitude of the image while removing noise. Total Variation regularization performs well in applications such as image denoising, image reconstruction, and image segmentation, and can effectively preserve the details and edge information of an image.

[0051] As a specific implementation manner, the process of performing Retinex variational processing on the V component in the input image space in step 1 includes: using the joint guided total variation as the illumination component regularization term in the Retinex variational model to constrain the illumination component, using the logarithmic mean local variance as the reflection component regularization term to constrain the reflection component, and the form of the objective function used is:

[0052]

[0053] where α and β are positive parameters used to balance the weights of each term; S is the input image, I is the illumination component in the image, R is the reflection component in the image, ||·|| 2 represents the 2-norm, is the data fidelity term; H(I) P is the illumination component regularization term using the joint guided total variation; E t (R) is the reflection component regularization term using the logarithmic mean local variance;

[0054] Furthermore, the joint guided total variation in the illumination component regularization term includes:

[0055]

[0056] Among them, H(I) p represents the joint-guided total variation, ε is a small number to avoid division by zero, ||·|| 1 represents the 1-norm, D x (p) and D y (p) represent the total window variation along the horizontal and vertical axes of pixel p, respectively, and the expression is:

[0057]

[0058] Among them, and are the gradient operators along the horizontal and vertical axes respectively, q belongs to the rectangular region R(p) centered on pixel p, I is the illumination component in the image, g k (||p - q||) is the weight of pixel q = (x q , y q ) in the spatial window R(p) under the Gaussian filter, and the expression is:

[0059]

[0060] Among them, k is the standard deviation of the Gaussian distribution, and x and y are the coordinate information of each pixel point in the image.

[0061] G x (p) and G y (p) represent the guided filtering along the horizontal and vertical axes of pixel p respectively, and the expression is:

[0062]

[0063] Among them, and are the gradient operators along the horizontal and vertical axes respectively, E(a k , b k ) is the loss function, and the expression is:

[0064]

[0065] Among them, w k represents the local window centered on pixel k, G is the guidance image, I is the illumination component in the image, λ is a regularization parameter, a k and b k represent the slope and intercept of the linear transformation respectively, and the expression is:

[0066]

[0067] Among them, μk and σ k respectively represent the mean and variance of the input image I, and ε is a decimal number to avoid division by zero. represents the pixel average of the illumination component I in the image.

[0068] Furthermore, the logarithmic mean local variance in the reflection component regularization term has the form:

[0069]

[0070] where Ω represents a 3*3 image block in the image. and are the gradient operators along the horizontal and vertical axes respectively, and R is the reflection component in the image.

[0071] As a specific implementation manner, the adjustment of the illumination component by using the arctangent transform in step 2 has the form:

[0072]

[0073] where is the enhanced illumination component obtained by using the arctangent transform, I(x,y) is the illumination component obtained by the Retinex variation, and δ is a parameter to control the visibility, which has the form:

[0074]

[0075] where mean(I) represents the average value of the illumination component I.

[0076] As a specific implementation manner, the multiplication of the illumination component and the reflection component in step 3 has the form:

[0077]

[0078] where is the enhanced image, is the enhanced illumination component obtained by using the arctangent transform, is the reflection component.

[0079] Example Two

[0080] This example also provides a computer device, including a memory, a processor, and a computer program stored on the memory. The processor executes the computer program to implement the steps of the method.

[0081] Example Three

[0082] This example also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method are implemented.

[0083] Example 4

[0084] This embodiment also provides a computer program product, including a computer program which, when executed by a processor, implements the steps of the method.

[0085] The above are only the preferred specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed in the present application should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A low-light image enhancement method based on joint guided total variation, characterized in that: The following steps are involved: Acquire low-light images; Convert the low-light image from the RGB color space to the HSV color space, and perform Retinex variational processing to obtain a reflection component and an illumination component; The illumination component is adjusted based on an inverse tangent transformation, and the adjusted illumination component and the reflection component are multiplied to obtain an enhanced image.

2. The method according to claim 1, characterized in that The process of performing Retinex variational processing on the low-light image includes: The H and S components of the HSV color space are kept unchanged, and the V component is subjected to Retinex variational processing. The joint guided total variation is used as the regularization term of the illumination component to constrain the illumination component, and the logarithmic mean local variance is used as the regularization term of the reflection component to constrain the reflection component.

3. The method according to claim 2, characterized in that The calculation formula of the joint guided total variation is as follows: Among them, H(I) p represents the joint guided total variation, ε is a small number that avoids division by 0, ||·||1 represents the 1-norm, and D x (p) and D y (p) represents the total variation of the window along the horizontal and vertical axes of pixel p, respectively.

4. The method according to claim 2, characterized in that: The calculation formula for the log mean local variance is as follows: Among them, Ω means that the 3*3 image in the image is fast, and are the gradient operators for the horizontal and vertical axes respectively, and R is the reflection component in the image.

5. The method according to claim 1, characterized in that The calculation formula for adjusting the illumination component based on the inverse tangent transformation is as follows: in, is the enhanced illumination component obtained using the inverse tangent transform, I(x,y) is the illumination component obtained by Retinex variation, and δ is the parameter that controls visibility.

6. A computer device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 5.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

8. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images