Low-illumination image enhancement method, system, equipment and medium

Abstract

The invention provides a low-illumination image enhancement method, system and device and a medium, and relates to the field of image processing. The method comprises the following steps: converting an original low-illumination image from an RGB color space to an HSV color space, and carrying out normalization processing to obtain first image data; decomposing the first image data through a first model to obtain an illumination component L and a reflection component R; wherein the first model is a model for performing logarithmic transformation and multi-scale decomposition on the input image on the basis of adding a noise item so as to estimate an illumination component L of the input image. And performing integral processing on the reflection component R and the illumination component obtained by adjusting the illumination component L to achieve the purpose of illuminating the dark area to generate an enhanced V-channel image. And converting the enhanced V-channel image from the HSV color space to the RGB color space to obtain an enhanced image. According to the method, the reflectivity displayed by the structure and the smooth illumination component can be estimated, so that the image enhancement effect is improved.

Description

Technical Field

[0001] The present application relates to the field of image processing, and in particular to a low-illumination image enhancement method, system, device and medium. Background Art

[0002] Due to the limitations of image acquisition equipment, the loss of transmission channels, randomness in image storage and processing, etc., image color distortion and noise are generated, which affects the visual effect of the image and reduces the quality and reliability of the image. With the development of digital image processing technology, image enhancement technology is also constantly improving. At present, the Retinex theoretical method has become a research hotspot for scholars at home and abroad due to its significant image enhancement effect and easy implementation. However, the traditional Retinex theoretical method can enhance the color of the image but is prone to distortion, and there is still a lot of room for improvement in image enhancement performance. Summary of the invention

[0003] The purpose of the present application is to provide a low-light image enhancement method, system, device and medium, which can estimate the reflectivity and smooth illumination components of a structure display, thereby improving the image enhancement effect.

[0004] This application is implemented as follows:

[0005] In a first aspect, the present application provides a low-light image enhancement method, comprising the following steps:

[0006] The original low-light image is converted from RGB color space to HSV color space and normalized to obtain first image data. The first image data is decomposed by a first model to obtain an illumination component L and a reflection component R; wherein the first model is a model that performs logarithmic transformation and multi-scale decomposition on the input image on the basis of adding a noise term to estimate the illumination component L of the input image. The reflection component R and the adjusted illumination component L are integrated to generate an enhanced V channel image; the adjusted illumination component L is an illumination component obtained by adjusting the illumination component L to illuminate the dark area. The enhanced V channel image is converted from HSV color space to RGB color space to obtain an enhanced image.

[0007] Further, based on the aforementioned scheme, the mathematical expression of the above-mentioned first model is S(x, y)=L(x, y)·R(x, y)+N; wherein S(x, y) is the original image, L(x, y) is the illumination component, R(x, y) is the reflection component, and N is the noise term.

[0008] Further, based on the above scheme, in the step of decomposing the first image data by the first model, the decomposition constraint model of the illumination component L and the reflection component R of the first image is:

[0009]

[0010] Wherein, R is the reflection component, L is the illumination component, G represents the adjustment gradient of the first image data I, N is the noise term, is the intensity of the overall noise, β and ω are regularization parameters, ||·|| F and ||·||1 represent F-norm and L-norm respectively, represents the first-order differential operator, is the fidelity between the original image and the reconstructed image, is the overall sparsity and regional smoothness of the illumination component L, The distance between R and G of the reflection estimate map.

[0011] Further, based on the above scheme, in the step of decomposing the first image data by the first model, the gradient of the input image is amplified by the factor K to obtain the G matrix, and the adjustment formula is:

[0012]

[0013]

[0014] in, λ controls the degree of amplification, and σ controls the amplification rate of different gradients.

[0015] Furthermore, based on the aforementioned scheme, in the process of solving the decomposition constraint model of the illumination component L and the reflection component R of the first image, the ADMM optimization framework is used and the Lagrange multiplier is introduced to decompose the original optimization problem into multiple sub-problems.

[0016] Further, based on the above scheme, the above-adjusted lighting component L is the lighting component obtained by adjusting the lighting component L through Gamma correction to achieve the purpose of illuminating the dark area. The formula for Gamma correction adjustment is: in, is the illumination component after adjustment, L is the illumination component before adjustment, and γ is the correction parameter.

[0017] Furthermore, based on the above scheme, the above method also includes: using information entropy, visible edge ratio, and visible edge normalized gradient mean as image quality evaluation indicators to evaluate the quality of the enhanced image.

[0018] In a second aspect, the present application provides a low-light image enhancement system, comprising:

[0019] The first conversion module is configured to convert the original low-light image from the RGB color space to the HSV color space, and perform normalization processing to obtain the first image data. The image decomposition module is configured to decompose the first image data through the first model to obtain the illumination component L and the reflection component R. The integral processing module is configured to perform integral processing on the reflection component R and the adjusted illumination component L to generate an enhanced V channel image; the adjusted illumination component L is the illumination component obtained by adjusting the illumination component L to achieve the purpose of illuminating the dark area. The second conversion module is configured to convert the enhanced V channel image from the HSV color space to the RGB color space to obtain the enhanced image.

[0020] In a third aspect, the present application provides an electronic device, comprising a memory for storing one or more programs; a processor; when the one or more programs are executed by the processor, a method as described in any one of the first aspects is implemented.

[0021] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a method as described in any one of the above-mentioned first aspects.

[0022] Compared with the prior art, the present invention has at least the following advantages or beneficial effects:

[0023] This application proposes a low-light image enhancement method, which estimates the illumination component of the image more robustly by performing logarithmic transformation and multi-scale decomposition on the original image. Then, by performing a subtraction operation on the estimated illumination component, the reflection component is effectively extracted to capture the local details and structure of the image. It simplifies the calculation by using a logarithmic transformation operation, and finally generates an enhanced image through an inverse logarithmic transformation, which is more robust in the face of illumination changes and noise, and improves the image quality and detail performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0025] Figure 1 This is a flow chart of an embodiment of a low-light image enhancement method of the present application;

[0026] Figure 2 This is a schematic diagram of the traditional Retinex algorithm model;

[0027] Figure 3 This is a structural block diagram of an embodiment of a low-light image enhancement system of the present application;

[0028] Figure 4 A structural block diagram of an electronic device provided in an embodiment of the present application.

[0029] Icon: 101, first conversion module; 102, image decomposition module; 103, integration processing module; 104, second conversion module; 201, processor; 202, memory; 203, communication interface. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various different configurations.

[0031] In conjunction with the accompanying drawings, some implementation methods of the present application are described in detail below. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0032] Example 1

[0033] The overall brightness of low-light AR color images is low. The traditional Retinex algorithm can enhance the color of the image, but it is prone to distortion, is sensitive to noise, and some image details are submerged in pixels with lower gray values. To this end, the embodiment of the present application provides a low-light image enhancement method that can estimate the reflectivity and smooth illumination components of a structure display, thereby improving the image enhancement effect.

[0034] See also Figure 1 , the low illumination image enhancement method comprises the following steps:

[0035] Step S101: converting the original low-light image from the RGB color space to the HSV color space, and performing normalization processing to obtain first image data.

[0036] In order to more closely match the visual characteristics of the human eye and conduct an intuitive analysis of the original low-light image, the original low-light image is converted from the RGB color space to the HSV color space, and normalized (mainly normalizing the V channel) to obtain the corresponding HSV image (i.e., the first image data). Among them, the full name of HSV is Hue, Saturation, Value, which means hue, saturation, and brightness. It is a space used for color processing and a color space that conforms to the visual experience of the human eye. Among them, hue (H) reflects the basic properties of color and describes the order of color from 0 to 360 degrees starting from red. Saturation (S) indicates the purity of the color, ranging from 0 to 1. Brightness (V) indicates the brightness of the color, ranging from 0 to 1.

[0037] Step S102: Decompose the first image data through a first model to obtain an illumination component L and a reflection component R; wherein the first model is a model that performs logarithmic transformation and multi-scale decomposition on the input image on the basis of adding a noise term to estimate the illumination component L of the input image.

[0038] like Figure 2 As shown in the figure, in the traditional Retinex algorithm, it is believed that the image can be decomposed into two parts: the incident component and the reflected component. The incident component captures the dynamic information of the image and corresponds to the low-frequency area in the image, while the reflected component reflects the essential characteristics of the image and corresponds to the high-frequency component in the image. Therefore, the algorithm improves the quality of the image by weakening the influence of the incident component and enhancing the reflected component at the same time. The image can be regarded as a model expression of the product of the incident component and the reflected component, and its mathematical expression can be expressed as: S(x, y) = L(x, y) · R(x, y). Among them, S(x, y) represents the original image collected, L(x, y) represents the incident component, also called the illumination component, and R(x, y) represents the reflected component. That is, the traditional Retinex method does not take model noise into account when processing the illumination smoothness and contrast of the original image. In the first model of the present application, assuming that the noise term is relatively evenly distributed in the original image, the corresponding mathematical expression is S(x, y) = L(x, y) · R(x, y) + N; wherein S(x, y) is the original image, L(x, y) is the illumination component, R(x, y) is the reflection component, and N is the noise term. Thus, the first model estimates the illumination component of the image in a more robust manner by performing a logarithmic transformation and multi-scale decomposition on the original image. By performing a subtraction operation on the estimated illumination component, the reflection component is effectively extracted to capture the local details and structure of the image. Logarithmic domain operations are used to simplify the calculation, and finally an enhanced image is generated by an inverse logarithmic transformation, which makes the algorithm more robust in the face of illumination changes and noise, and improves the image quality and detail performance.

[0039] Step S103: integrating the reflection component R and the adjusted illumination component L to generate an enhanced V channel image; the adjusted illumination component L is an illumination component obtained by adjusting the illumination component L to illuminate the dark area;

[0040] Step S104: converting the enhanced V channel image from the HSV color space to the RGB color space to obtain an enhanced image.

[0041] At present, most low-light image enhancement algorithms are designed for ordinary low-light images. When processing color images with extremely low illumination, such as AR images, problems such as insufficient image edge blur and severe color distortion are often faced. At the same time, low-light image enhancement will amplify noise. In the present application, after the first image data is decomposed by the first model to obtain the illumination component L and the reflection component R, in order to illuminate the dark area, the illumination component L is adjusted and the adjusted illumination component L is generated. The adjusted illumination component L and the reflection component R are then integrated to generate an enhanced V channel image. Finally, the enhanced V channel image is converted from the HSV color space to the RGB color space, and the enhanced HSV image is converted to the RGB color space to obtain the final enhanced image. Thus, the original color contrast can be relatively retained and the local clarity can be improved while denoising, that is, the color images with extremely low illumination such as AR images can be effectively enhanced, so as to better display the detailed information of objects such as scenery contained in the image.

[0042] In order to enable those skilled in the art to understand the present application more intuitively, a specific example will be used here to illustrate.

[0043] First, the model of the illumination component L and the reflection component R of the image input to the first model is decomposed as follows:

[0044]

[0045] Where R is the reflection component, L is the illumination component, G represents the adjustment gradient of the first image data I, β and ω are regularization parameters, ||·|| F and ||·||1 represent F-norm and L-norm respectively. represents the first-order differential operator, is the fidelity between the original image and the reconstructed image, is the overall sparsity and regional smoothness of the illumination component L, The distance between R and G of the reflection estimate map.

[0046] In the objective function, it is proposed To constrain and guide the fidelity between matrices G. The G matrix is ​​obtained by amplifying the gradient of the input image by factor K. In order to balance the overall size of G, factor K is adaptively adjusted in regions of different sizes. The formulas for G and K are as follows:

[0047]

[0048]

[0049] Among them, λ is the degree of control amplification, and σ is the amplification rate of different gradients.

[0050] Next, considering that the noise in natural low-light images is not just Gaussian white noise, it is difficult to evaluate the noise level, so a noise map is directly estimated from the input image (HSV space). It should be noted that the noise present in natural low-light images may have different distributions and properties. By directly estimating the noise map, it will not be limited to additive or multiplicative Gaussian white noise, so as to adapt to different types of noise more flexibly. In addition, since the noise map is directly estimated based on the content of the input image, and this estimation is performed in the HSV (hue, saturation, brightness) color space, the estimation of the noise map is related to the color and brightness information of the image. Exemplarily, the noise map can be estimated by analyzing the statistical characteristics of the input image, such as the distribution of pixel values, the correlation between color channels, etc. The noise map can also be estimated by fitting model parameters, and the fitted noise distribution model can be a Gaussian, Poisson or other distribution model.

[0051] That is to say, for the noise term N, it is a noise that can be directly estimated from the input image (first image data) at the first model. Accordingly, based on the above scheme, in an implementation of the present application, in the step of decomposing the first image data by the first model, the decomposition constraint model of the illumination component L and the reflection component R of the first image is:

[0052]

[0053] Wherein, R is the reflection component, L is the illumination component, G represents the adjustment gradient of the first image data I, N is the noise term, is the intensity of the overall noise, β and ω are regularization parameters, ||·|| F and ||·||1 represent F-norm and L-norm respectively, represents the first-order differential operator, is the fidelity between the original image and the reconstructed image, is the overall sparsity and regional smoothness of the illumination component L, The distance between R and G of the reflection estimate map.

[0054] Then, the matrix G is modified. That is, based on the above scheme, in one implementation of the present application, in the step of decomposing the first image data by the first model, the gradient of the input image is amplified by the factor K to obtain the G matrix, and the adjustment formula is:

[0055]

[0056]

[0057] in, λ controls the degree of amplification, and σ controls the amplification rate of different gradients. It is an absolute value function, so the noise level of small gradients is suppressed before amplification, which can cope with the situation of dense noise under extremely low illumination in the subsequent enhancement process.

[0058] Next, for non-concave optimization problems, in one implementation of the present application, in the process of solving the decomposition constraint model of the illumination component L and the reflection component R of the first image, the ADMM optimization framework is used, and the Lagrange multiplier is introduced to decompose the original optimization problem into multiple sub-problems.

[0059] The solution process can be divided into the following five steps:

[0060] Step 1: Transform the optimization problem to be solved into:

[0061]

[0062] where I is the original image, R is the reflection component, Ψ and Φ are regularization terms corresponding to contrast and smoothness respectively, and λ is a balancing parameter.

[0063] Step 2: Introduce Lagrange multipliers:

[0064] Introduce Lagrange multipliers L1 and L2:

[0065] L1=IR

[0066]

[0067] Step 3: Construct the augmented Lagrangian function:

[0068]

[0069] Among them, ρ is the penalty parameter, Ψ(I i，j ) is the contrast regularization term, and λ is the balance parameter used to control the balance between the contrast regularization term and the smoothness regularization term. i，j ) is the smoothness regularization term, which regularizes the smoothness of each pixel of the reflection component R by Φ. is the gradient of the image.

[0070] Step 4: ADMM optimization steps

[0071] Use ADMM to alternately minimize the augmented Lagrangian function:

[0072] Update I:

[0073]

[0074] Update R:

[0075]

[0076] Update L1 and L2:

[0077] L1 k+1 =L1 k +ρ(I k+1 -R k+1 )

[0078]

[0079] These steps will be iterated until the convergence condition is met. Finally, the enhanced image I is output. k+1 As the final result

[0080] Step 5: Lighting component adjustment to improve overall illumination

[0081] In one implementation of the present application, the adjusted illumination component L is an illumination component obtained by adjusting the illumination component L through Gamma correction to achieve the purpose of illuminating the dark area. The formula for Gamma correction adjustment is: in, is the illumination component after adjustment, L is the illumination component before adjustment, and γ is the correction parameter. For example, the γ value can be set to 2.2.

[0082] Finally, the denoising and enhancing image of the V channel in the HSV space can be converted to the RGB space according to the following formula:

[0083]

[0084]

[0085] is the enhanced image through the spatial V channel, R is the reflectivity component, and the final joint denoising and enhancement effect image can be obtained

[0086] Based on the above scheme, in an implementation of the present application, the method further includes: using information entropy, visible edge ratio, and visible edge normalized gradient mean as image quality evaluation indicators to evaluate the quality of the enhanced image, so as to objectively evaluate the quality of the image enhancement effect.

[0087] The size of image information entropy is proportional to the average amount of information. The larger the information entropy, the richer the image information and the better the image quality evaluation. The calculation formula of information entropy is as follows:

[0088]

[0089] Where N is the total number of gray levels, P i Represents the probability of pixel i appearing in the image.

[0090] The visible edge ratio of an image refers to the ratio of the visible boundary (area with obvious changes in color, brightness or texture) in the image to the entire image. The formula for the visible edge ratio is as follows:

[0091]

[0092] Where n0 is the number of visible edges in the original image, n r This is the number of visible edges collected after using the algorithm. As the number of visible edges increases, the image becomes clearer and more edges are visible.

[0093] The visible edge normalized gradient mean is an indicator used to evaluate the edge quality of an image. In the field of image processing, edge detection is an important step in identifying the boundaries of objects in an image. However, the edge quality obtained by different edge detection algorithms varies, so an objective evaluation method is needed to measure the edge quality. The main idea of ​​the visible edge normalized gradient mean is to compare the gradient of the edge in the image with the gradient distribution of the entire image to obtain a quantitative indicator of the edge quality. The calculation formula for the visible edge normalized gradient mean can be:

[0094]

[0095] In the formula, r represents the correlation coefficient of the visible edge, VL r and VL0 represent the visibility level of the target in the image collected by the improved algorithm and the original image, respectively, ΔI r and ΔI0 represent the gradient values ​​of the pixels on the visible edges in the image collected by the improved algorithm and the original image.

[0096] The applicant enhances the same target image group using the method of the present application, the traditional RetinexL method and the multi-scale RetinexL method, and uses information entropy, visible edge ratio, and visible edge normalized gradient mean as image quality evaluation indicators to evaluate the quality of the enhanced images. The table formed is as follows (the quality evaluation results of the three groups of processing results are given as examples below):

[0097] Information Entropy Visible edge ratio Visible edge normalized gradient mean Traditional RetinexL method 3.1698 0.58803 0.0619 Multi-scale RetinexL method 7.3823 0.99912 0.0415 Method of the present application 7.6024 0.95565 0.0207

[0098] Table 1. Index data of quality evaluation of the first group of target images using three algorithms

[0099] Information Entropy Visible edge ratio Visible edge normalized gradient mean Traditional RetinexL method 1.596 0.248 0.0466 Multi-scale RetinexL method 7.8477 0.9997 0.0469 Method of the present application 7.5444 0.9658 0.0307

[0100] Table 2. Index data of quality evaluation of the second group of target images using three algorithms

[0101]

[0102]

[0103] Table 3. Index data of quality evaluation of the third group of target images using three algorithms

[0104] From the data in Table 1, it can be seen that the information entropy of the low-light image enhancement method of the present application is higher than that of the multi-scale Retinex theoretical algorithm and the Retinex theory, indicating that the enhanced image processed by the low-light image enhancement method of the present application is rich in information and has richer detail texture. The visible edge ratio value of the low-light image enhancement method of the present application is slightly lower than that of the multi-scale Retinex algorithm, indicating that while it increases color contrast, the effect of improving image brightness is slightly less than ideal.

[0105] It can be seen from the data in Table 2 that the value of the normalized gradient mean of the visible edge of the low-illumination image enhancement method of the present application is smaller than that of the other two algorithms, indicating that the algorithm is effective in processing image details and retains the clearest original details of the image. However, the information entropy and visible edge values ​​are slightly lower than those of the multi-scale Retinex algorithm, which means that the brightness improvement and detail processing in individual areas of the image need to be strengthened.

[0106] It can be seen from the data in Table 3 that the information entropy value of the low-light image enhancement method adopted in the present application is the highest, and the values ​​of the visible edge ratio and the visible edge normalized gradient mean are relatively high. It can be seen that the low-light image enhancement method proposed in the present application has the best performance in image enhancement processing, and relatively retains the original color contrast of the image while denoising and improves the local clarity.

[0107] Therefore, the experimental results show that the various indicators of the low-light image enhancement method of this application are superior to those of other algorithms. The information entropy value is high, indicating that the image has the most information; the visible edge ratio is increased, that is, more edge details are retained, and the detail expression ability is strong; the visible edge normalized gradient mean is large, which improves the visibility of the image. Therefore, the low-light image enhancement method of this application is generally superior to the other two algorithms, can better display scene information, and can provide a certain theoretical basis for AR visual image enhancement.

[0108] Example 2

[0109] See also Figure 3 , the embodiment of the present application provides a low illumination image enhancement system, which includes:

[0110] The first conversion module 101 is configured to convert the original low-light image from the RGB color space to the HSV color space, and perform normalization processing to obtain the first image data. The image decomposition module 102 is configured to decompose the first image data through the first model to obtain the illumination component L and the reflection component R. The integral processing module 103 is configured to perform integral processing on the reflection component R and the adjusted illumination component L to generate an enhanced V channel image; the adjusted illumination component L is the illumination component obtained by adjusting the illumination component L to achieve the purpose of illuminating the dark area. The second conversion module 104 is configured to convert the enhanced V channel image from the HSV color space to the RGB color space to obtain the enhanced image.

[0111] For the specific implementation process of the above system, please refer to the low-illumination image enhancement method provided in Example 1, which will not be repeated here.

[0112] Example 3

[0113] See also Figure 4 , the embodiment of the present application provides an electronic device, which includes at least one processor 201 and at least one memory 202; wherein the processor 201 and the memory 202 are directly connected to each other, or communicate with each other through a communication interface 203, or are electrically connected through one or more communication buses or signal lines to achieve data transmission or interaction; the memory 202 stores program instructions that can be executed by the processor 201, and the processor 201 calls the program instructions to execute a low-light image enhancement method. For example, the following is implemented:

[0114] The original low-light image is converted from the RGB color space to the HSV color space and normalized to obtain the first image data. The first image data is decomposed by the first model to obtain the illumination component L and the reflection component R. The reflection component R and the adjusted illumination component L are integrated to generate an enhanced V channel image; the adjusted illumination component L is the illumination component obtained by adjusting the illumination component L to illuminate the dark area. The enhanced V channel image is converted from the HSV color space to the RGB color space to obtain an enhanced image.

[0115] Among them, the memory 202 can be, but is not limited to, a random access memory (RAM), a read only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable read-only memory (EEPROM), etc.

[0116] The processor 201 may be an integrated circuit chip with signal processing capability. The processor 201 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0117] Understandably, Figure 4 The structure shown is for illustration only. The electronic device may also include Figure 4 More or fewer components as shown, or with Figure 4 Different configurations shown. Figure 4 Each component shown in the figure can be implemented by hardware, software or a combination thereof.

[0118] Example 4

[0119] The present application provides a computer-readable storage medium having a computer program stored thereon, which implements a low-light image enhancement method when executed by the processor 201. For example, the following is implemented:

[0120] The original low-light image is converted from the RGB color space to the HSV color space and normalized to obtain the first image data. The first image data is decomposed by the first model to obtain the illumination component L and the reflection component R. The reflection component R and the adjusted illumination component L are integrated to generate an enhanced V channel image; the adjusted illumination component L is the illumination component obtained by adjusting the illumination component L to illuminate the dark area. The enhanced V channel image is converted from the HSV color space to the RGB color space to obtain an enhanced image.

[0121] If the above functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0122] It will be apparent to those skilled in the art that the present application is not limited to the details of the exemplary embodiments described above, and that the present application can be implemented in other specific forms without departing from the spirit or essential features of the present application. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the present application is defined by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims be included in the present application. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A low-light image enhancement method, characterized in that: The following steps are involved: The original low-light image is converted from the RGB color space to the HSV color space, and normalized to obtain first image data; Decomposing the first image data by a first model to obtain an illumination component L and a reflection component R; wherein the first model is a model that performs logarithmic transformation and multi-scale decomposition on the input image on the basis of adding a noise term to estimate the illumination component L of the input image; Integrating the reflection component R and the adjusted illumination component L to generate an enhanced V channel image; the adjusted illumination component L is an illumination component obtained by adjusting the illumination component L to illuminate the dark area; The enhanced V channel image is converted from the HSV color space to the RGB color space to obtain the enhanced image.

2. A low illumination image enhancement method as claimed in claim 1, characterized in that: The mathematical expression of the first model is S(x, y)=L(x, y)·R(x, y)+N; wherein S(x, y) is the original image, L(x, y) is the illumination component, R(x, y) is the reflection component, and N is the noise term.

3. A low-light image enhancement method as claimed in claim 2, characterized in that: In the step of decomposing the first image data by using the first model, the decomposition constraint model of the illumination component L and the reflection component R of the first image is: Wherein, R is the reflection component, L is the illumination component, G represents the adjustment gradient of the first image data I, N is the noise term, is the intensity of the overall noise, β and ω are regularization parameters, ||·|| F and ||·||1 represent F-norm and L-norm respectively, ▽ represents the first-order differential operator, is the fidelity between the original image and the reconstructed image, ||▽L||1 is the overall sparsity and regional smoothness of the illumination component L, The distance between R and G of the reflection estimate map.

4. A low-light image enhancement method as claimed in claim 3, characterized in that: In the step of decomposing the first image data by the first model, the gradient of the input image is amplified by the factor K to obtain the G matrix, and the adjustment formula is: in, λ controls the degree of amplification, and σ controls the amplification rate of different gradients.

5. The low-light image enhancement method according to claim 1, characterized in that: In the process of solving the decomposition constraint model of the illumination component L and the reflection component R of the first image, the ADMM optimization framework is used and the Lagrange multiplier is introduced to decompose the original optimization problem into multiple sub-problems.

6. A low-light image enhancement method as claimed in claim 1, characterized in that: The adjusted lighting component L is the lighting component obtained by adjusting the lighting component L through Gamma correction to achieve the purpose of illuminating the dark area. The formula for Gamma correction adjustment is: in, is the illumination component after adjustment, L is the illumination component before adjustment, and γ is the correction parameter.

7. The low-light image enhancement method according to claim 1, characterized in that: The method further comprises: using information entropy, visible edge ratio, and visible edge normalized gradient mean as image quality evaluation indicators to evaluate the quality of the enhanced image.

8. A low-light image enhancement system, characterized in that: include: The first conversion module is configured to: convert the original low-light image from the RGB color space to the HSV color space, and perform normalization processing to obtain first image data; An image decomposition module is configured to: decompose the first image data through a first model to obtain an illumination component L and a reflection component R; The integral processing module is configured to: perform integral processing on the reflection component R and the adjusted illumination component L to generate an enhanced V channel image; The adjusted lighting component L is a lighting component obtained by adjusting the lighting component L to achieve the purpose of illuminating the dark area; The second conversion module is configured to: convert the enhanced V channel image from the HSV color space to the RGB color space to obtain an enhanced image.

9. An electronic device, characterized in that: include: A memory for storing one or more programs; processor; When the one or more programs are executed by the processor, the method according to any one of claims 1 to 7 is implemented.

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

Images