A method, apparatus and device for enhancing a low-light image

Abstract

The application discloses a low-brightness image enhancement method, device and equipment, and the method comprises the following steps: inputting a to-be-processed image into a pre-constructed image decomposition model, using the image decomposition model to constrain the illumination gradient in the to-be-processed image, obtaining a first illumination component, using the image decomposition model to preserve the reflection gradient in the to-be-processed image, obtaining a first reflection component; adjusting the illumination intensity of the first illumination component according to a predetermined adjustment rule, generating a second illumination component; and generating an enhanced image based on the second illumination component and the first reflection component; the enhanced image obtained through the above method can enhance the weak light area while retaining the visibility of the local high-intensity area, so that the robot can have good visibility in a complex weak light environment, and the working efficiency of the equipment can be greatly improved.

Description

Technical Field

[0001] This invention relates to the field of image processing, and more specifically to a method, apparatus, and device for enhancing low-brightness images. Background Technology

[0002] In real-world scenarios, lighting, perspective, and other factors can cause our photos to be quite dark. These dark images not only affect our observation but also significantly impact the performance of computer vision processing algorithms. Therefore, low-brightness image enhancement methods have emerged.

[0003] In related technologies, image brightness enhancement methods typically decompose the input image into illumination and reflection components based on the Retinex model; then, Gamma correction is used to globally enhance the illumination, thereby achieving image brightness enhancement. However, because the constraints used in current Retinex model-based enhancement methods are insufficient to obtain accurate illumination and reflection components, problems such as color distortion and halo artifacts appear in the final enhanced image. Summary of the Invention

[0004] Therefore, the technical problem to be solved by the present invention is to overcome the defect in the prior art that it is impossible to accurately decompose the illumination component and the reflection component, thereby providing a method, apparatus and device for enhancing low brightness images.

[0005] In a first aspect, the present invention provides a method for enhancing a low-brightness image, comprising:

[0006] The image to be processed is input into a pre-constructed image decomposition model. The image decomposition model is used to constrain the illumination gradient in the image to be processed to obtain the first illumination component. The image decomposition model is used to preserve the reflection gradient in the image to be processed to obtain the first reflection component. The illumination intensity of the first illumination component is adjusted according to a predetermined adjustment rule to generate the second illumination component. An enhanced image is generated based on the second illumination component and the first reflection component.

[0007] This invention utilizes a pre-constructed image decomposition model to constrain the illumination gradient and preserve the fidelity of the reflection gradient, resulting in more accurate first illumination and first reflection components corresponding to the image to be processed. The illumination intensity of the first illumination component is adjusted according to a predetermined adjustment rule to obtain a second illumination component. Then, an enhanced image is generated based on the second illumination component and the first reflection component. The enhanced image obtained by this method can enhance low-light areas while preserving the visibility of local high-intensity areas, thereby enabling robots to have good visibility in complex low-light environments and significantly improving the working efficiency of the equipment.

[0008] In conjunction with the first aspect, in a first embodiment of the first aspect, the image decomposition model includes:

[0009] The first fidelity sub-model, the second fidelity sub-model, and the constraint sub-model.

[0010] In conjunction with the first aspect, in the second embodiment of the first aspect, obtaining the first illumination component and the first reflection component includes:

[0011] An initial illumination component is determined based on the grayscale values ​​of the R, G, and B channels of the image to be processed; a first image decomposition submodel is constructed based on the first fidelity submodel and the constraint submodel, and a second image decomposition submodel is constructed based on the first fidelity submodel and the second fidelity submodel; the initial illumination component is input into the first image decomposition submodel to obtain a third illumination component; the third illumination component is input into the second image decomposition submodel to obtain a second reflection component; when the third illumination component satisfies a first preset condition and the second reflection component satisfies a second preset condition, the third illumination component is determined as the first illumination component, and the second reflection component is determined as the first reflection component.

[0012] In conjunction with the first aspect, in the third embodiment of the first aspect, the second image decomposition sub-model constrains the reflection gradient in the image to be processed using a preset enhanced image gradient.

[0013] In conjunction with the first aspect, in the fourth embodiment of the first aspect, the step of constructing the first image decomposition model based on the first fidelity sub-model and the constraint sub-model includes:

[0014] A third image decomposition submodel is constructed based on the first fidelity submodel and the constraint submodel; the first fidelity submodel in the third image decomposition submodel is simplified into a third fidelity submodel formed by the initial illumination component; the constraint submodel in the third image decomposition submodel is replaced with a third fidelity submodel equivalent to the constraint submodel to generate the first image decomposition submodel.

[0015] In conjunction with the first aspect, in the fifth embodiment of the first aspect, the first image decomposition sub-model includes a first decomposition term and a second decomposition term. The initial illumination component is input into the first image decomposition sub-model to obtain a third illumination component, including:

[0016] The initial illumination component and predefined constraint operators are input into the first decomposition term to obtain auxiliary variables, which are used to constrain the illumination gradient. The auxiliary variables, the initial illumination component, and the first preset coefficient are input into the second decomposition term to obtain the third illumination component, wherein the first preset coefficient is used to characterize the similarity between the illumination gradient and the auxiliary variables.

[0017] In conjunction with the first aspect, in the sixth embodiment of the first aspect, the method further includes:

[0018] When the third illumination component does not meet the first preset condition, and / or the second reflection component does not meet the second preset condition, the iteration count is updated; the updated iteration count is compared with a preset iteration threshold; when the updated iteration count reaches the preset iteration threshold, the third illumination component is determined as the first illumination component, and the second reflection component is determined as the first reflection component; or, when the updated iteration count does not reach the preset iteration threshold, the first preset coefficient is adjusted to the second preset coefficient according to a preset rule; the auxiliary variable, the initial illumination component, and the second preset coefficient are input into the second decomposition term, and the result output by the second decomposition term is input into the second image decomposition sub-model; until the final output illumination component meets the first preset condition, and the output reflection component meets the second preset condition, or, when the iteration count reaches the preset iteration threshold, the iteration operation is stopped, and the final output illumination component is determined as the first illumination component, and the final output reflection component is determined as the first reflection component.

[0019] In conjunction with the first aspect, in the seventh embodiment of the first aspect, adjusting the light intensity of the first light component according to a predetermined adjustment rule to generate the second light component includes:

[0020] Based on the adjustment rule, adjustment parameters corresponding to each pixel in the image to be processed are determined, and the adjustment rule is the mapping relationship between the adjustment parameters and the pixels; the illumination intensity of the first illumination component is adjusted based on all the adjustment parameters to generate a second illumination intensity.

[0021] Secondly, the present invention provides an enhancement device for low-brightness images, comprising:

[0022] The input module is used to input the image to be processed into a pre-constructed image decomposition model. Using the image decomposition model, the illumination gradient in the image to be processed is constrained to obtain the first illumination component. Using the image decomposition model, the reflection gradient in the image to be processed is preserved to obtain the first reflection component. The adjustment module is used to adjust the illumination intensity of the first illumination component according to a predetermined adjustment rule to generate the second illumination component. The generation module is used to generate an enhanced image based on the second illumination component and the first reflection component.

[0023] Thirdly, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory being used to store a computer program, and when the computer program is executed by the processor, causing the processor to perform a method for enhancing a low-brightness image as described in any of the claims of the present invention. Attached Figure Description

[0024] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0025] Figure 1 This is a flowchart of a low-brightness image enhancement method provided in an embodiment of the present invention;

[0026] Figure 2 This is a connection diagram of the low-brightness image enhancement device provided in an embodiment of the present invention;

[0027] Figure 3 This is a computer device connection diagram provided for an embodiment of the present invention. Detailed Implementation

[0028] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] In image enhancement techniques, the Retinex model is typically used to decompose the image into illumination and reflection components. Then, Gamma global illumination is used. However, the enhanced image obtained in this way will severely lose the structure and texture of local high-intensity areas.

[0030] This invention discloses a method for enhancing low-brightness images, such as... Figure 1 As shown, the specific steps include the following:

[0031] S11: Input the image to be processed into the pre-built image decomposition model. Use the image decomposition model to constrain the illumination gradient in the image to be processed to obtain the first illumination component. Use the image decomposition model to preserve the reflection gradient in the image to be processed to obtain the first reflection component.

[0032] Specifically, the image to be processed is a pre-acquired RGB color image, which is a low-brightness image.

[0033] Specifically, when decomposing the image to be processed, each pixel of the image to be processed is decomposed into a first illumination part and a reflection part. The first illumination component in S11 is a general term for the first illumination part decomposed from each pixel in the image to be processed. Similarly, the first reflection component is a general term for the reflection part decomposed from each pixel in the image to be processed.

[0034] For example, in an optional embodiment, the image decomposition model includes: a first fidelity sub-model, a second fidelity sub-model, and a constraint sub-model. This image decomposition model can be implemented using formula (1):

[0035]

[0036] in,‖*‖ F Let I be the Fourier norm, ||*||0 be the L0 norm, L be the illumination component, R be the reflection component, I be the image to be processed, and Q be an auxiliary variable. Let H be the reflection gradient, and H be the gradient of the image to be enhanced. It is a fidelity term based on the Retinex model between the "illumination component and reflection component" and the "image to be processed", where ||Q|0 is the constraint term for the illumination gradient, and λ and ω are balance coefficients. This is the fidelity term between the reflection gradient and the gradient of the image to be enhanced. The constraint condition is the illumination gradient.

[0037] In this embodiment, the first fidelity sub-model is embodied as The second fidelity sub-model is manifested as follows: The constraint sub-model is embodied in |Q|0, In the second fidelity sub-model, the reflection component can be constrained by the second fidelity term to suppress image noise.

[0038] S12: Adjust the light intensity of the first light component according to the predetermined adjustment rules to generate the second light component.

[0039] For example, after obtaining the first illumination component from the image decomposition model, the illumination intensity of the first illumination component needs to be adjusted. When adjusting the illumination intensity of the first illumination component, adjustment parameters corresponding to each pixel in the image to be processed are determined based on a predetermined adjustment rule. The illumination intensity of the first illumination component is adjusted based on all adjustment parameters to generate a second illumination component. That is, according to the adjustment parameters, the first illumination portion in each pixel is adjusted to the second illumination portion; the second illumination component is the collective term for the second illumination portion corresponding to each pixel in the image to be processed. The adjustment parameters corresponding to each pixel can be the same or different. The adjustment rule is the mapping relationship between the adjustment parameters and the pixels.

[0040] S13: Generate an enhanced image based on the second illumination component and the first reflection component.

[0041] Specifically, after obtaining the second illumination component and the first reflection component, the second illumination component and the first reflection component are multiplied element-wise to obtain the enhanced image. Element-wise multiplication involves multiplying the second illumination component and the first reflection component corresponding to each pixel to obtain the corresponding calculation result. Based on the calculation results of all pixels, the enhanced image is generated.

[0042] This invention utilizes a pre-constructed image decomposition model to constrain the illumination gradient and preserve the fidelity of the reflection gradient, resulting in more accurate first illumination and first reflection components corresponding to the image to be processed. The illumination intensity of the first illumination component is adjusted according to a predetermined adjustment rule to obtain a second illumination component. Then, an enhanced image is generated based on the second illumination component and the first reflection component. The enhanced image obtained by this method can enhance low-light areas while preserving the visibility of local high-intensity areas, thereby enabling robots to have good visibility in complex low-light environments and significantly improving the working efficiency of the equipment.

[0043] In an optional embodiment, obtaining the first illumination component and the first reflection component includes the following steps:

[0044] S21: Determine the initial illumination components based on the gray values ​​of the R, G, and B channels of the image to be processed.

[0045] For example, after acquiring the image to be processed, the grayscale values ​​corresponding to the R, G, and B channels of the image to be processed are acquired respectively, and the maximum value is selected from the three grayscale values ​​and determined as the initial illumination component. It should be noted that the initial illumination component here is also a general term for the initial illumination part corresponding to each pixel in the image to be processed. In this embodiment, the specific method for determining the initial illumination component can be achieved by formula (2):

[0046]

[0047] in, Let x be the initial illumination component, x be the number of pixels in the image to be processed, and c be the R, G, and B channels of the image to be processed.

[0048] S22: Based on the first fidelity sub-model and the constraint sub-model, construct the first image decomposition sub-model, and based on the first fidelity sub-model and the second fidelity sub-model, construct the second image decomposition sub-model.

[0049] For example, still taking the embodiment of S11 as an example, since the image decomposition model in this embodiment is a non-convex problem, the estimation of the first illumination component and the first reflection component needs to be obtained by further decomposing the image decomposition model. In this embodiment, the image decomposition model needs to be decomposed into a first image decomposition sub-model corresponding to the illumination component and a second image decomposition sub-model corresponding to the reflection component.

[0050] In one optional embodiment, the specific method for constructing the first image decomposition sub-model is as follows: First, a third image decomposition sub-model is constructed based on the first fidelity sub-model and the constraint sub-model in the image decomposition model; second, the first fidelity sub-model in the third image decomposition sub-model is simplified into a third fidelity sub-model formed by the initial illumination component; finally, the constraint sub-model in the third image decomposition sub-model is replaced with a third fidelity sub-model equivalent to the constraint sub-model to generate the first image decomposition sub-model.

[0051] Specifically, a third image decomposition sub-model is constructed based on the first fidelity sub-model and the constraint sub-model. The third image decomposition sub-model is implemented by formula (3).

[0052]

[0053] After constructing the third image sub-model, since the auxiliary variables of the illumination component, reflection component, and illumination gradient in this model are all unknown, solving for the first illumination component based solely on the third image decomposition sub-model would be extremely complex. Therefore, this scheme proposes to solve for the first illumination component based on the initial illumination component. Estimate the illumination component L. Specifically, a semi-decoupling method is used to simplify the third image decomposition sub-model. The initial illumination component is used to preserve the fidelity of the illumination component, thus obtaining the simplified third image decomposition sub-model, which is achieved through formula (4). In formula (4)... This is the third fidelity sub-model.

[0054]

[0055] To further simplify the solution process, this embodiment uses the constraint sub-model in the third image sub-model. Equivalent rewrite as This generates the first image sub-model. The first image sub-model is implemented using formula (5):

[0056]

[0057] The coefficient β is used to control the auxiliary variable Q and the constraints. The similarity between them. When β is large enough, formula (5) will be infinitely close to formula (4).

[0058] In an optional embodiment, the specific method for constructing the second image decomposition sub-model is as follows: extracting a first fidelity sub-model and a second fidelity sub-model from the image decomposition model, and constructing the second image decomposition sub-model based on the first fidelity sub-model and the second fidelity sub-model. The second image decomposition sub-model uses a preset enhanced image gradient to constrain the reflection gradient in the image to be processed. The second image decomposition sub-model is implemented by formula (6):

[0059]

[0060] S23: Input the initial illumination component into the first image decomposition sub-model to obtain the third illumination component.

[0061] For example, the first image decomposition sub-model is still a non-convex problem. Therefore, in order to obtain the third illumination component, the first image sub-model needs to be further decomposed by the alternating minimization method to obtain the first decomposition term related to the auxiliary variable Q constraining the illumination gradient and the second decomposition term related to the illumination component L. The first decomposition term is embodied in formula (7), and the second decomposition term is embodied in formula (8).

[0062]

[0063]

[0064] After obtaining the first decomposition term, the initial illumination components and predefined constraints will be... Input the values ​​into the first decomposition term to solve for the auxiliary variable Q under different conditions. Then, combine the auxiliary variable Q and the initial illumination component. The first preset coefficient β is input into the second decomposition term to obtain the third illumination component. It should be noted that the first preset coefficient β is used to characterize the constraint conditions. The similarity between the variable Q and the auxiliary variable Q.

[0065] S24: Input the third illumination component into the second image decomposition sub-model to obtain the second reflection component.

[0066] For example, taking the embodiment corresponding to S22 as an example, the first illumination component determined in S24 is substituted into formula (6). Since formula (6) is a classic least squares problem, the derivative of the first reflection component needs to be calculated in formula (6), and the derivative is set to 0. In this way, the equation shown in formula (9) can be obtained, thereby solving for the second reflection component.

[0067] (L 2 +ωD T D)R=L T I+ωD T H (9)

[0068] S25: When the third illumination component satisfies the first preset condition and the second reflection component satisfies the second preset condition, the third illumination component is determined as the first illumination component and the second reflection component is determined as the first reflection component.

[0069] For example, the second preset condition is the reflection components R of two adjacent reflections. (k) and R (k+1) The difference between them is less than the corresponding difference threshold ΔR.

[0070] In this embodiment, the third illumination component is the illumination component calculated for the first time after the image to be processed is input into the image decomposition model, i.e., L. (1) At this point, the previous illumination component adjacent to the third illumination component can be defaulted to L. (0) , will L (1) and L (0) The difference is compared with ΔL. When L (1) -L (0) When ΔL < ΔL, the third illumination component is determined as the first illumination component. Similarly, R is... (1) and R (0) The difference is compared with ΔR. When R (1) -R (0) When ΔR < ΔR, the second reflection component is determined as the first reflection component. In this embodiment, L is... (0) and R (0) The default value is 0; set both ΔL and ΔR to 10. -3 .

[0071] When the third illumination component does not meet the first preset condition, and / or the second reflection component does not meet the second preset condition, the iteration count is first obtained to determine whether the iteration count has reached the preset iteration threshold. When the iteration count reaches the preset iteration threshold, the third illumination component is directly determined as the first illumination component, and the second reflection component is determined as the first reflection component. When the iteration count has not reached the preset iteration threshold, the first preset coefficient is adjusted to the second preset coefficient according to the preset rules, and steps S23-S25 are executed again until the final illumination component meets the first preset condition and the reflection component meets the second preset condition. At this point, the iteration operation is stopped, and the first reflection component and the first reflection component are output.

[0072] In an optional embodiment, the first image decomposition sub-model includes a first decomposition term and a second decomposition term. An initial illumination component is input into the first image decomposition sub-model to obtain a third illumination component, including:

[0073] S31: Input the initial illumination components and predefined constraints into the first decomposition term to obtain auxiliary variables, which are used to constrain the illumination gradient.

[0074] For example, in this embodiment, for ease of calculation, the first decomposition term is assigned to E, resulting in formula (10):

[0075]

[0076] In this embodiment, the constraints are... Defined in the following form:

[0077]

[0078] Where the operator * represents the convolution operation, G σ (x,y) is a Gaussian kernel with standard deviation σ, which is formally represented by formula (12).

[0079]

[0080] Where K is the normalized value, and the role of K is to ensure that ∫∫G σ (x,y)dxdy=1, where x and y are the coordinates of each pixel.

[0081] In this embodiment, when Q≠0, the value of ||Q||0 is 1, and when Q=0, the value of ||Q||0 is 0.

[0082] Combining formulas (10) and (11), we analyze the auxiliary variable Q corresponding to the constraints under different conditions.

[0083] The analysis process is as follows:

[0084] when And when Q≠0, we can obtain:

[0085]

[0086] when And when Q = 0, we can obtain:

[0087]

[0088] According to formulas (13) and (14), This is obtained when Q = 0.

[0089] when And when Q≠0, we can obtain:

[0090]

[0091] According to formula (15), exist It was obtained at that time.

[0092] when And when Q = 0, we can obtain:

[0093]

[0094] because From formula (11), we can obtain that when At that time, one can obtain

[0095] Taking all the above factors into account, we can obtain the auxiliary variable Q:

[0096]

[0097] S32: Input the auxiliary variable, the initial illumination component, and the first preset coefficient into the second decomposition term to obtain the third illumination component.

[0098] Specifically, the first preset coefficient is used to characterize the similarity between the constraints and the auxiliary variables.

[0099] For example, since the solution process of the second decomposition term, i.e., formula (8), includes the solution of... The decomposition of is therefore extremely difficult to solve. It can be observed that as β approaches infinity, It tends to 0. Therefore, to simplify the solution process, this scheme will... Simplified to

[0100]

[0101] Since formula (18) involves quadratic terms, the illumination component can be differentiated in formula (18), and the derivative can be set to 0 to obtain the equation shown in formula (19), thereby solving for the third illumination component.

[0102]

[0103] Where 1 is an identity matrix of appropriate size, and D contains level D h and vertical D v D h and D v It is the Toeplitz matrix obtained based on the discrete gradient operator with positive difference.

[0104] In an optional embodiment, the method further includes:

[0105] When the third illumination component does not meet the first preset condition, and / or the second reflection component does not meet the second preset condition, the iteration count is updated; the updated iteration count is compared with a preset iteration threshold; when the updated iteration count reaches the preset iteration threshold, the third illumination component is determined as the first illumination component, and the second reflection component is determined as the first reflection component; or, when the updated iteration count does not reach the preset iteration threshold, the first preset coefficient is adjusted to the second preset coefficient according to a preset rule; the auxiliary variable, the initial illumination component, and the second preset coefficient are input into the second decomposition term, and the output of the second decomposition term is input into the second image decomposition sub-model; until the final output illumination component meets the first preset condition, and the output reflection component meets the second preset condition, or when the iteration count reaches the preset iteration threshold, the iteration operation is stopped, and the final output illumination component is determined as the first illumination component, and the final output reflection component is determined as the first reflection component.

[0106] For example, in this embodiment, the preset iteration threshold is 5 times. When the third illumination component does not meet the first preset condition, and / or the second reflection component does not meet the second preset condition, the iteration count is incremented by 1, and it is determined whether the updated iteration count reaches 5 times. When the updated iteration count is greater than or equal to 5 times, the third illumination component is directly determined as the first illumination component, and the second reflection component is determined as the first reflection component. When the iteration count is less than 5 times, the first preset coefficient β in the second decomposition term, i.e., formula (8), is updated to the second preset coefficient β according to the preset rule. ‘ The preset rule can be β ‘=2β. Then, the second preset coefficient, initial illumination variable and auxiliary variable are re-inputted into the second decomposition term, and the operation of S32 is continued. The output of the second decomposition term is input into the second image decomposition sub-model until the final output illumination component meets the first preset condition and the output reflection component meets the second preset condition; or, when the number of iterations reaches the preset iteration threshold, the iteration operation is stopped, and the output illumination component is determined as the first illumination component and the output reflection component is determined as the first reflection component.

[0107] In one optional embodiment, the illumination intensity of the first illumination component is adjusted according to a predetermined adjustment rule to generate the second illumination component, including:

[0108] Based on the adjustment rule, adjustment parameters corresponding to each pixel in the image to be processed are determined, and the adjustment rule is the mapping relationship between the adjustment parameters and the pixels; the illumination intensity of the first illumination component is adjusted based on all the adjustment parameters to generate a second illumination intensity.

[0109] For example, in this embodiment, the adjustment rule is embodied in the form of an adjustment function. The function model corresponding to the adjustment function is the optimal model selected after a large number of trials and tests, and the parameters in the adjustment function are the optimal parameters determined after fitting a large amount of data. In this embodiment, the adjustment function is an exponential function, as shown below:

[0110]

[0111] Where x represents different pixel values, and a, b, and d are fixed parameters. In this embodiment, a, b, and d are 3, 4, and 0.41, respectively.

[0112] Once the adjustment parameters for each pixel are determined, the first illumination component is adjusted using all the adjustment parameters to transform it into the second illumination component. Then, the second illumination component and the first reflection component are calculated element-wise to obtain the final enhanced image.

[0113]

[0114] in, To enhance the image, R represents the first reflection component, and L represents the first illumination component. This is the second illumination component. To adjust the rules, This is the initial illumination component.

[0115] This embodiment determines the adjustment parameters corresponding to different pixel values ​​in the image to be processed by fitting a pre-defined exponential function, and then adjusts the first illumination portion of the corresponding pixel using adjustment rules. In this way, targeted adjustments can be made based on pixel characteristics, thereby avoiding the loss of image structure that exists in related technologies, and thus enhancing the visibility of local high-intensity areas while enhancing low-light areas.

[0116] This invention discloses a device for enhancing low-brightness images, such as... Figure 2 As shown, the device includes the following modules:

[0117] The input module 21 is used to input the image to be processed into a pre-constructed image decomposition model. The image decomposition model is used to constrain the illumination gradient in the image to be processed to obtain the first illumination component. The image decomposition model is used to preserve the reflection gradient in the image to be processed to obtain the first reflection component.

[0118] The adjustment module 22 is used to adjust the light intensity of the first light component according to a predetermined adjustment rule to generate the second light component.

[0119] The generation module 23 is used to generate an enhanced image based on the second illumination component and the first reflection component.

[0120] In an optional embodiment, the image decomposition model in input module 21 includes:

[0121] The first fidelity sub-model, the second fidelity sub-model, and the constraint sub-model.

[0122] In an optional embodiment, the input module 21 includes:

[0123] The first determining submodule is used to determine the initial illumination component based on the gray values ​​of the R, G, and B channels of the image to be processed; the construction submodule is used to construct a first image decomposition submodel based on the first fidelity submodel and the constraint submodel, and to construct a second image sub-decomposition model based on the first fidelity submodel and the second fidelity submodel; the first input submodule is used to input the initial illumination component into the first image decomposition submodel to obtain the third illumination component; the second input submodule is used to input the third illumination component into the second image decomposition submodel to obtain the second reflection component; the second determining submodule is used to determine the third illumination component as the first illumination component and the second reflection component as the first reflection component when the third illumination component meets the first preset condition and the second reflection component meets the second preset condition.

[0124] In an optional embodiment, the second image decomposition sub-model in the construction sub-module constrains the reflection gradient in the image to be processed using a preset enhanced image gradient.

[0125] In an alternative embodiment, the construction submodule includes:

[0126] The first image decomposition submodel is constructed based on the first fidelity submodel and the constraint submodel. The second image decomposition submodel is simplified to the first fidelity submodel in the third image decomposition submodel and formed by the initial illumination component. The third image decomposition submodel is generated by replacing the constraint submodel in the third image decomposition submodel with the third fidelity submodel that is equivalent to the constraint submodel.

[0127] In an optional embodiment, the first image decomposition sub-model in the first input sub-module includes a first decomposition term and a second decomposition term. The first input sub-module includes:

[0128] The first input unit is used to input the initial illumination component and predefined constraints into the first decomposition term to obtain auxiliary variables, which are used to constrain the illumination gradient; the second input unit is used to input the auxiliary variables, the initial illumination component, and the first preset coefficient into the second decomposition term to obtain the third illumination component, wherein the first preset coefficient is used to characterize the similarity between the constraints and the auxiliary variables.

[0129] In an optional embodiment, the input module 21 further includes:

[0130] The update submodule is used to update the iteration count when the third illumination component does not meet the first preset condition and / or the second reflection component does not meet the second preset condition; the comparison submodule is used to compare the updated iteration count with a preset iteration threshold; the determination submodule is used to determine the third illumination component as the first illumination component and the second reflection component as the first reflection component when the updated iteration count reaches the preset iteration threshold; the adjustment submodule is used to adjust the first preset coefficient to the second preset coefficient according to a preset rule when the updated iteration count does not reach the preset iteration threshold; the input submodule is used to input the auxiliary variable, the initial illumination component, and the second preset coefficient into the second decomposition term, and input the output of the second decomposition term into the second image decomposition sub-model, until the final output illumination component meets the first preset condition and the output reflection component meets the second preset condition, or when the iteration count reaches the preset iteration threshold, the iteration operation is stopped, and the final output illumination component is determined as the first illumination component and the final output reflection component is determined as the first reflection component.

[0131] In an optional embodiment, the generation module 23 includes:

[0132] The determination submodule is used to determine the adjustment parameters corresponding to each pixel in the image to be processed based on the adjustment rules, wherein the adjustment rules are the mapping relationship between the adjustment parameters and the pixels; the adjustment submodule is used to adjust the light intensity of the first light component based on all the adjustment parameters to generate a second light intensity.

[0133] This embodiment provides a computer device, such as... Figure 3 As shown, the computer device may include at least one processor 31, at least one communication interface 32, at least one communication bus 33, and at least one memory 34. The communication interface 32 may include a display screen and a keyboard; optionally, the communication interface 32 may also include a standard wired interface or a wireless interface. The memory 34 may be high-speed RAM (Random Access Memory) or non-volatile memory, such as at least one disk drive. Optionally, the memory 34 may also be at least one storage device located remotely from the aforementioned processor 31. The processor 31 may be combined with... Figure 3 The described apparatus has an application program stored in memory 34, and the processor 31 calls the program code stored in memory 34 to perform the low-brightness image enhancement method of any of the above method embodiments.

[0134] The communication bus 33 can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The communication bus 33 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 3 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0135] The memory 34 may include volatile memory, such as random-access memory (RAM); the memory may also include non-volatile memory, such as flash memory, hard disk drive (HDD) or solid-state drive (SSD); the memory 34 may also include a combination of the above types of memory.

[0136] The processor 31 can be a central processing unit (CPU), a network processor (NP), or a combination of CPU and NP.

[0137] The processor 31 may further include a hardware chip. This hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The PLD may be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof. Optionally, the memory 34 is also used to store program instructions. The processor 31 can invoke the program instructions to implement the low-brightness image enhancement method in any embodiment of the present invention.

[0138] This embodiment provides a computer-readable storage medium storing computer-executable instructions that can execute the low-brightness image enhancement method in any of the above method embodiments. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk drive (HDD), or solid-state drive (SSD), etc.; the storage medium may also include combinations of the above types of memory.

[0139] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A method for enhancing a low-brightness image, characterized in that, include: The image to be processed is input into a pre-constructed image decomposition model. The illumination gradient in the image to be processed is constrained using the image decomposition model to obtain the first illumination component. The reflection gradient in the image to be processed is preserved using the image decomposition model to obtain the first reflection component. The light intensity of the first light component is adjusted according to a predetermined adjustment rule to generate the second light component; An enhanced image is generated based on the second illumination component and the first reflection component; The step of adjusting the light intensity of the first light component according to a predetermined adjustment rule to generate the second light component includes: Based on the adjustment rule, adjustment parameters corresponding to each pixel in the image to be processed are determined, and the adjustment rule is the mapping relationship between the adjustment parameters and the pixels; The light intensity of the first light component is adjusted based on all the aforementioned adjustment parameters to generate a second light intensity; The adjustment rule is expressed in the form of an adjustment function, which is an exponential function: ;in, The function is called the adjustment function, where x represents different pixel values, and a, b, and d are fixed parameters. The enhanced image is: ;in, For the enhanced image, For the first reflection component, This is the first illumination component. This is the second illumination component. The adjustment rule, This is the initial illumination component; The initial illumination component is determined based on the grayscale values ​​of the image to be processed in the R, G, and B channels.

2. The method for enhancing low-brightness images according to claim 1, characterized in that, The image decomposition model includes: The first fidelity sub-model, the second fidelity sub-model, and the constraint sub-model.

3. The method for enhancing low-brightness images according to claim 2, characterized in that, Obtaining the first illumination component and the first reflection component includes: Based on the first fidelity sub-model and the constraint sub-model, a first image decomposition sub-model is constructed, and based on the first fidelity sub-model and the second fidelity sub-model, a second image decomposition sub-model is constructed. The initial illumination component is input into the first image decomposition sub-model to obtain the third illumination component; The third illumination component is input into the second image decomposition sub-model to obtain the second reflection component; When the third illumination component satisfies the first preset condition and the second reflection component satisfies the second preset condition, the third illumination component is determined as the first illumination component and the second reflection component is determined as the first reflection component.

4. The method for enhancing low-brightness images according to claim 3, characterized in that, The second image decomposition sub-model uses a preset enhanced image gradient to constrain the reflection gradient in the image to be processed.

5. The method for enhancing low-brightness images according to claim 3, characterized in that, The construction of the first image decomposition model based on the first fidelity sub-model and the constraint sub-model includes: A third image decomposition submodel is constructed based on the first fidelity submodel and the constraint submodel; The first fidelity sub-model in the third image decomposition sub-model is simplified to the third fidelity sub-model formed by the initial illumination components; The constraint sub-model in the third image decomposition sub-model is replaced with a third fidelity sub-model that is equivalent to the constraint sub-model to generate the first image decomposition sub-model.

6. The method for enhancing a low-brightness image according to any one of claims 3-5, characterized in that, The first image decomposition sub-model includes a first decomposition term and a second decomposition term. The step of inputting the initial illumination component into the first image decomposition sub-model to obtain the third illumination component includes: The initial illumination component and predefined constraints are input into the first decomposition term to obtain auxiliary variables, which are used to constrain the illumination gradient. The auxiliary variable, the initial illumination component, and the first preset coefficient are input into the second decomposition term to obtain the third illumination component, wherein the first preset coefficient is used to characterize the similarity between the constraint and the auxiliary variable.

7. The method for enhancing low-brightness images according to claim 6, characterized in that, The method further includes: When the third illumination component does not meet the first preset condition, and / or the second reflection component does not meet the second preset condition, the iteration count is updated; The updated number of iterations is compared with a preset iteration threshold. When the updated number of iterations reaches the preset iteration threshold, the third illumination component is determined as the first illumination component, and the second reflection component is determined as the first reflection component. Alternatively, if the updated number of iterations does not reach the preset iteration threshold, the first preset coefficient is adjusted to the second preset coefficient according to a preset rule; The auxiliary variable, the initial illumination component, and the second preset coefficient are input into the second decomposition term, and the output of the second decomposition term is input into the second image decomposition sub-model; The iteration operation stops when the final output illumination component meets the first preset condition and the output reflection component meets the second preset condition, or when the number of iterations reaches the preset iteration threshold. The final output illumination component is then determined as the first illumination component, and the final output reflection component is also determined as the first reflection component.

8. A device for enhancing low-brightness images, characterized in that, include: The input module is used to input the image to be processed into a pre-constructed image decomposition model, use the image decomposition model to constrain the illumination gradient in the image to be processed to obtain the first illumination component, and use the image decomposition model to preserve the reflection gradient in the image to be processed to obtain the first reflection component. An adjustment module is used to adjust the light intensity of the first light component according to a predetermined adjustment rule to generate a second light component. The generation module is used to generate an enhanced image based on the second illumination component and the first reflection component; The step of adjusting the light intensity of the first light component according to a predetermined adjustment rule to generate the second light component includes: Based on the adjustment rule, adjustment parameters corresponding to each pixel in the image to be processed are determined, and the adjustment rule is the mapping relationship between the adjustment parameters and the pixels; The light intensity of the first light component is adjusted based on all the aforementioned adjustment parameters to generate a second light intensity; The adjustment rule is expressed in the form of an adjustment function, which is an exponential function: ;in, The function is called the adjustment function, where x represents different pixel values, and a, b, and d are fixed parameters. The enhanced image is: ;in, For the enhanced image, For the first reflection component, This is the first illumination component. This is the second illumination component. The adjustment rule, This is the initial illumination component; The initial illumination component is determined based on the grayscale values ​​of the image to be processed in the R, G, and B channels.

9. A computer device, characterized in that, include: A memory and a processor are communicatively connected, the memory being used to store a computer program, which, when executed by the processor, causes the processor to perform a low-brightness image enhancement method as described in any one of claims 1 to 7.

Images