An image exposure parameter adjustment method, device, and equipment

Abstract

The present invention discloses an image exposure parameter adjustment method, device and equipment, which includes: obtaining an image to be processed, and inputting the image to be processed into a trained exposure adjustment model for exposure parameter adjustment; inputting the image to be processed into the feature extraction module, and respectively extracting local features and global features of the image according to the convolutional neural network and the convolutional neural network with attention mechanism in the feature extraction module; inputting preset adjustment parameters into the degree modulation module to obtain a plurality of exposure modulation parameters; fusing the local features, the global features and the exposure modulation parameters to obtain an affine grid, and obtaining an affine matrix by upsampling the affine grid; inputting the image to be processed and the affine matrix into the matrix mapping module for mapping operation to obtain a generated image after exposure adjustment. It can obtain a more delicate image effect to achieve the purpose of beautifying the image.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and particularly to an image exposure parameter adjustment method, device, and equipment. Background Art

[0002] Exposure is usually a parameter that reflects the amount of light entering the photosensitive element of the lens during the image photography process. In the real scenario, limited by the limited hardware conditions of the mobile phone, users usually have difficulty taking images with satisfactory exposure. Therefore, users need to adjust and enhance the exposure of the image according to their own needs through software algorithms in the later stage of shooting to further beautify the image effect.

[0003] Currently, the existing image exposure adjustment mainly includes: (1) traditional image processing methods, such as the commonly used histogram equalization algorithm. Although this type of method can achieve the effect of exposure adjustment, it is easy to lose image details and generate regional breaks; (2) deep learning methods based on CNN. Although this type of method can obtain detailed image details, its exposure degree is usually fixed, and it is difficult to adjust the exposure degree based on the user's preference. Moreover, the number of network parameters is huge and the operation efficiency is low, which is difficult to meet the operation requirements of the mobile phone. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to propose an image exposure parameter adjustment method, device, and equipment to solve the above problems.

[0005] To achieve the above purpose, the present invention provides an image exposure parameter adjustment method, and the method includes:

[0006] Obtain an image to be processed, and input the image to be processed into a trained exposure adjustment model for exposure parameter adjustment. Among them, the network structure of the exposure adjustment model includes a feature extraction module, a degree modulation module, and a matrix mapping module, and the feature extraction module is constructed based on a convolutional neural network and a convolutional neural network with an attention mechanism;

[0007] Input the image to be processed into the feature extraction module, and respectively extract the local feature and the global feature of the image according to the convolutional neural network and the convolutional neural network with an attention mechanism in the feature extraction module;

[0008] Input a preset adjustment parameter into the degree modulation module to obtain a plurality of exposure modulation parameters;

[0009] Fuse the local feature, the global feature, and the exposure modulation parameters to obtain an affine grid, and obtain an affine matrix by upsampling the affine grid;

[0010] Input the to-be-processed image and the affine matrix into the matrix mapping module for mapping operations to obtain a generated image after exposure adjustment.

[0011] Preferably, extracting the local features of the image according to the convolutional neural network in the feature extraction module includes:

[0012] Performing different non-linear transformation combinations on the to-be-processed image according to the convolutional neural network in the feature extraction module to obtain local features.

[0013] Preferably, extracting the global features of the image according to the convolutional neural network with an attention mechanism in the feature extraction module includes:

[0014] Dynamically combining and extracting the extracted local features in an adaptive weighting manner according to the convolutional neural network with an attention mechanism in the feature extraction module to obtain global features.

[0015] Preferably, dynamically combining and extracting the extracted local features in an adaptive weighting manner according to the convolutional neural network with an attention mechanism in the feature extraction module to obtain global features includes:

[0016] According to the formula X feature =W ι *[softmax(X q ⊙X k )⊙X v for dynamic combination of the local features of the image, where ⊙ represents matrix operation, X q 、X k 、X v represent local feature matrices formed by different non-linear transformation combinations of the image X, W ι represents the weight parameter of the perceptron, and X feature represents the global features dynamically extracted by the convolutional neural network with an attention mechanism of the image.

[0017] Preferably, the degree modulation module is constructed based on a multi-layer perceptron; inputting the preset adjustment parameters into the degree modulation module to obtain multiple exposure modulation parameters includes:

[0018] Inputting the preset adjustment parameters into the multi-layer perceptron, outputting multiple scaling parameters and displacement parameters, and modulating the exposure degree through the multiple scaling parameters and the displacement parameters to obtain the multiple exposure modulation parameters.

[0019] Preferably, modulating the exposure degree through the multiple scaling parameters and the displacement parameters to obtain the multiple exposure modulation parameters includes:

[0020] The modulation of the exposure degree is performed according to the formula C′ = α*C + β, where α represents the scaling parameter, β represents the shift parameter, C represents the preset exposure adjustment parameter, and C′ represents the modulated exposure modulation parameter.

[0021] Preferably, the training process of the exposure adjustment model includes:

[0022] Using the preset loss function L = L1 + L 感知 + L fft to optimize the exposure adjustment model, where L1 represents the L1 loss, L 感知 represents the perceptual loss, and L fft represents the FFT loss. In the formula, x represents the target image, represents the generated image, and F represents the fast Fourier transform operation on the image.

[0023] To achieve the above object, the present invention also provides an image exposure parameter adjustment device, and the device includes:

[0024] An acquisition unit, configured to acquire an image to be processed, and input the image to be processed into a trained exposure adjustment model for exposure parameter adjustment. The network structure of the exposure adjustment model includes a feature extraction module, a degree modulation module, and a matrix mapping module. The feature extraction module is constructed based on a convolutional neural network and a convolutional neural network with an attention mechanism;

[0025] A feature extraction unit, configured to input the image to be processed into the feature extraction module, and respectively extract the local feature and the global feature of the image according to the convolutional neural network and the convolutional neural network with an attention mechanism in the feature extraction module;

[0026] A parameter modulation unit, configured to input a preset adjustment parameter into the degree modulation module to obtain a plurality of exposure modulation parameters;

[0027] A fusion processing unit, configured to perform fusion processing on the local feature, the global feature, and the exposure modulation parameters to obtain an affine grid, and obtain an affine matrix by upsampling the affine grid;

[0028] A mapping operation unit, configured to input the image to be processed and the affine matrix into the matrix mapping module for mapping operation to obtain a generated image after exposure adjustment.

[0029] To achieve the above object, the present invention also proposes an image exposure parameter adjustment device, including a processor, a memory, and a computer program stored in the memory. The computer program is executed by the processor to implement the steps of an image exposure parameter adjustment method as described in the above embodiment.

[0030] To achieve the above object, the present invention further provides a computer-readable storage medium, on which a computer program is stored, and the computer program is executed by a processor to implement the steps of an image exposure parameter adjustment method as described in the above embodiments.

[0031] Beneficial effects:

[0032] In the above solution, compared with the traditional image processing method, the exposure adjustment based on the convolutional neural network can obtain a more delicate image effect, and can generate controllable beautification effects with different exposure degrees according to user needs, so as to improve the generated images with better exposure adjustment effects.

[0033] In the above solution, the constructed exposure adjustment model mainly includes three parts: feature extraction, degree modulation, and matrix mapping. By using the attention mechanism to design a neural network and modulating it through an exposure coefficient, an affine grid with different exposure degrees of the image is obtained. Through a simple upsampling operation on the affine grid, an affine matrix can be obtained and used to map to obtain the image after exposure adjustment. This characteristic of the affine grid can further help to build a lightweight network, greatly reducing the number of parameters and improving the operation and processing effect on the mobile phone side.

[0034] In the above solution, when training the exposure adjustment model, in addition to adding the L1 loss and the perceptual loss, the FFT loss is also added to enhance the texture details of the generated image, making the exposure adjustment result more natural. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0036] Figure 1 It is a schematic flowchart of an image exposure parameter adjustment method provided by an embodiment of the present invention.

[0037] Figure 2 It is a schematic structural diagram of an exposure adjustment model provided by an embodiment of the present invention.

[0038] Figure 3 It is a schematic structural diagram of an image exposure parameter adjustment device provided by an embodiment of the present invention.

[0039] The realization of the invention object, functional features and advantages will be further described in combination with the embodiments with reference to the drawings. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0040] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed present invention, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention.

[0041] The content of the present invention will be elaborated in detail below in conjunction with embodiments.

[0042] Refer to Figure 1 The flowchart of a method for adjusting image exposure parameters provided by an embodiment of the present invention is shown.

[0043] In this embodiment, the method includes:

[0044] S11, obtaining an image to be processed, and inputting the image to be processed into a trained exposure adjustment model for adjusting exposure parameters, where the network structure of the exposure adjustment model includes a feature extraction module, a degree modulation module, and a matrix mapping module, and the feature extraction module is constructed based on a convolutional neural network and a convolutional neural network with an attention mechanism;

[0045] S12, inputting the image to be processed into the feature extraction module, and respectively extracting local features and global features of the image according to the convolutional neural network and the convolutional neural network with an attention mechanism in the feature extraction module;

[0046] S13, inputting a preset adjustment parameter into the degree modulation module to obtain a plurality of exposure modulation parameters;

[0047] S14, performing a fusion process on the local feature, the global feature, and the exposure modulation parameter to obtain an affine grid, and obtaining an affine matrix by performing an upsampling operation on the affine grid;

[0048] S15, inputting the image to be processed and the affine matrix into the matrix mapping module for a mapping operation to obtain a generated image after exposure adjustment.

[0049] In this embodiment, during actual operation, the expected effect of the user on exposure adjustment is mainly that as the degree is adjusted, the corresponding exposure effect also changes accordingly. The greater the positive degree, the higher the brightness of the image, while the greater the negative degree, the lower the brightness of the image. In addition, the user also hopes to keep the original details of the image as much as possible while adjusting the exposure degree to achieve a transparent image effect. Therefore, the exposure adjustment model constructed in this embodiment is used to achieve the above situation. The exposure adjustment model includes three modules: feature extraction, degree modulation, and matrix mapping. Refer to Figure 2 as shown. First, the image to be processed is input into the feature extraction module to obtain the local features and global features of the image; at the same time, the preset exposure adjustment parameters are input into the degree modulation module to obtain a series of exposure modulation parameters; then, the local features, global features, and exposure modulation parameters are fused to obtain an affine grid, and the affine grid is obtained through upsampling operation. Finally, the image to be processed and the affine matrix are mapped through the matrix mapping module to obtain the final generated image. Among them, the range of the preset exposure adjustment parameter C is from -100 to 100 (including -100, -90,..., -10, 10,..., 90, to 100; respectively representing the exposure degree of -100, the exposure degree of -90,..., the exposure degree of 90, the exposure degree of 100) these 20 parameter values. And the fusion process mainly fuses the local features, global features, and exposure modulation parameters through channel stacking operation.

[0050] Further, in step S12, the image to be processed is input into the feature extraction module, and the local features and global features of the image are respectively extracted according to the convolutional neural network and the convolutional neural network with attention mechanism in the feature extraction module, including:

[0051] S12-1, performing different non-linear transformation combinations on the image to be processed according to the convolutional neural network in the feature extraction module to obtain local features;

[0052] S12-2, dynamically combining and extracting the extracted local features through the convolutional neural network with attention mechanism in the feature extraction module in an adaptive weighted manner to obtain global features.

[0053] Among them, the dynamically combining and extracting the extracted local features through the convolutional neural network with attention mechanism in the feature extraction module in an adaptive weighted manner to obtain global features includes:

[0054] According to the formula X feature = W ι *[softmax(X q ⊙X k )⊙X vPerform dynamic combination of local features of an image, where ⊙ represents matrix operation, and X q , X k , X v represent local feature matrices obtained by combining different non - linear transformations of image X, W ι represents the weight parameter of the perceptron, and X feature represents the global feature dynamically extracted by the convolutional neural network of the attention mechanism for the image.

[0055] In this embodiment, the feature extraction module uses a convolutional neural network to extract local features of the image; in addition, the feature extraction module also uses a convolutional neural network based on the attention mechanism for construction to extract global features of the image. Specifically, the convolutional neural network of the attention mechanism combines local features of the image into dynamic global features in an adaptive weighting manner (the adaptive weighting is reflected in the learnable perceptron weight parameter W ι ), and the process of combining local features into global features is reflected by the following formula. In the formula, local features X q , X k , X v are combined through the learnable weight parameter W ι and softmax operation, which can significantly improve the expression ability of the affine grid. Its formula can be expressed as:

[0056] X feature = W ι * [softmax(X q ⊙ X k )⊙ X v

[0057] where ⊙ represents matrix operation, X q , X k , X v represent local feature matrices obtained by combining different non - linear transformations of image X (referring to the image to be processed or the original image), W ι represents the weight parameter of the perceptron (this perceptron is the MLP multi - layer perceptron, and the weight parameter W ι in the perceptron is obtained by using the gradient descent method to perform back - iterative update on the parameters in the neural network), and X feature is the global feature dynamically extracted by the convolutional neural network of the attention mechanism for image X.

[0058] Furthermore, the degree modulation module is constructed based on a multi - layer perceptron; inputting the preset adjustment parameters into the degree modulation module to obtain multiple exposure modulation parameters, including:

[0059] ​Input the preset adjustment parameters into the multi-layer perceptron, and output multiple scaling parameters and displacement parameters. Modulate the exposure level through the multiple scaling parameters and the displacement parameters to obtain multiple exposure modulation parameters.

[0060] Among them, the modulating the exposure level through the multiple scaling parameters and the displacement parameters to obtain multiple exposure modulation parameters includes:

[0061] Modulate the exposure level according to the formula C′ = α * C + β, where α represents the scaling parameter, β represents the shift parameter, C represents the preset exposure adjustment parameter, and C‘ represents the modulated exposure modulation parameter.

[0062] In this embodiment, the degree modulation module is constructed using a multi-layer perceptron (MLP). The MLP has good pixel independence, and operating on a certain pixel will not affect the values of adjacent pixels. After the input preset exposure level adjustment parameters pass through the MLP, a series of scaling parameters α and shift parameters β for modulation are output, and its formula can be expressed as:

[0063] C′ = α * C + β where C represents the preset exposure adjustment parameter, and C‘ represents the modulated exposure modulation parameter.

[0064] The modulation termination condition of the above exposure modulation parameters is the same as the termination condition of the training of the entire network framework. When the generated image and the target image are close enough, the training is terminated.

[0065] Furthermore, a simple upsampling operation is used in the matrix mapping module to upsample the affine grid into an affine matrix with the same size as the input original image, and the input original image is mapped into the final generated image through a series of multiplication and addition operations.

[0066] Furthermore, the training process of the exposure adjustment model includes:

[0067] Select images with landscapes and portraits as the original images, and perform exposure adjustment on the original images through a preset algorithm to obtain target images;

[0068] Train according to the original images, target images, and a preset loss function to obtain the exposure adjustment model.

[0069] In this embodiment, by selecting image data including two types of scenes, namely landscapes and portraits, as the original images, the original images and the images after exposure adjustment using a traditional exposure adjustment algorithm are used as the image pairing data of the original image - target image. Geometric transformations such as random flipping, translation, rotation, and scaling are performed on the original images and the target images to improve the availability of the image data for network training. Furthermore, a loss function of the exposure adjustment model is constructed, and the loss function includes L1 loss, perceptual loss, and FFT loss. Among them,

[0070] The FFT loss is:

[0071] In the formula, x represents the target image, represents the generated image, and F represents the fast Fourier transform (FFT) operation on the image. Therefore, the overall loss function of the network is expressed as:

[0072] L = L1 + L 感知 + L fft

[0073] The overall network structure is jointly trained, and the generated image and the target image output by the network are input into the loss function of the overall network structure to calculate the loss, and the network parameters are iteratively updated. Through the above method, the learning ability of the network model can be better trained, and by using this exposure adjustment model to adjust the image exposure parameters, images with better exposure adjustment effects can be generated.

[0074] Compared with traditional image processing methods, the deep learning method based on CNN proposed in this embodiment obtains more delicate image effects; moreover, existing methods can only obtain exposure images with a fixed degree and cannot controllably generate images with different exposure degrees. Based on the above disadvantages, in this embodiment, an exposure adjustment method with controllable adjustment degree is constructed based on an attention neural network model, and three parts, namely a feature extraction module, a degree modulation module, and a matrix mapping module, are designed in the network structure. The exposure degree is modulated using an MLP to obtain the corresponding exposure effect. In addition, in addition to L1 loss and perceptual loss, FFT loss is also added to enhance the texture details of the generated image, and a lightweight network structure is constructed based on an affine matrix, greatly reducing the number of parameters and improving the processing effect of mobile devices such as mobile phones.

[0075] Refer to Figure 3 which shows a schematic structural diagram of an image exposure parameter adjustment device provided by an embodiment of the present invention.

[0076] In this embodiment, the device 30 includes:

[0077] An acquisition unit 31, configured to acquire an image to be processed, and input the image to be processed into a trained exposure adjustment model to adjust exposure parameters. The network structure of the exposure adjustment model includes a feature extraction module, a degree modulation module, and a matrix mapping module. The feature extraction module is constructed based on a convolutional neural network and a convolutional neural network with an attention mechanism;

[0078] A feature extraction unit 32, configured to input the image to be processed into the feature extraction module, and extract local features and global features of the image according to the convolutional neural network and the convolutional neural network with an attention mechanism in the feature extraction module respectively;

[0079] A parameter modulation unit 33, configured to input a preset adjustment parameter into the degree modulation module to obtain a plurality of exposure modulation parameters;

[0080] A fusion processing unit 34, configured to perform fusion processing on the local features, the global features, and the exposure modulation parameters to obtain an affine grid, and obtain an affine matrix through an upsampling operation on the affine grid;

[0081] A mapping operation unit 35, configured to input the image to be processed and the affine matrix into the matrix mapping module to perform a mapping operation to obtain a generated image after exposure adjustment.

[0082] Each unit module of the apparatus 30 can respectively execute the corresponding steps in the above method embodiments for adjusting image exposure parameters, so the unit modules will not be elaborated here. For details, please refer to the description of the above corresponding steps.

[0083] An embodiment of the present invention further provides an image exposure parameter adjustment device, which includes the above-mentioned image exposure parameter adjustment apparatus. The image exposure parameter adjustment apparatus can adopt the Figure 3 structure of the embodiment, and correspondingly, can execute the Figure 1 technical solution of the method embodiment shown. The implementation principle and technical effects are similar. For details, please refer to the relevant records in the above embodiments, and will not be elaborated here.

[0084] The device includes: a device with a photographing function such as a mobile phone, a digital camera, or a tablet computer, or a device with an image processing function, or a device with an image display function. The device may include components such as a memory, a processor, an input unit, a display unit, and a power supply.

[0085] Among them, the memory can be used to store software programs and modules. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory. The memory mainly includes a program storage area and a data storage area. Among them, the program storage area can store an operating system, application programs required for at least one function (such as an image playback function, etc.); the data storage area can store data created according to the use of the device, etc. In addition, the memory can include high-speed random access memory, and can also include non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other non-volatile solid-state storage devices. Correspondingly, the memory can also include a memory controller to provide access to the memory for the processor and the input unit.

[0086] The input unit can be used to receive input digital or character or image information, and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function controls. Specifically, in addition to including a camera, the input unit of this embodiment can also include a touch-sensitive surface (such as a touch display screen) and other input devices.

[0087] The display unit can be used to display information input by the user or information provided to the user and various graphical user interfaces of the device. These graphical user interfaces can be composed of graphics, text, icons, videos, and any combination thereof. The display unit can include a display panel. Optionally, the display panel can be configured in the form of an LCD (Liquid Crystal Display), an OLED (Organic Light-Emitting Diode), etc. Further, the touch-sensitive surface can cover the display panel. When the touch-sensitive surface detects a touch operation on or near it, it is transmitted to the processor to determine the type of touch event. Subsequently, the processor provides a corresponding visual output on the display panel according to the type of touch event.

[0088] An embodiment of the present invention also provides a computer-readable storage medium. The computer-readable storage medium can be the computer-readable storage medium included in the memory in the above embodiment; it can also exist alone and be a computer-readable storage medium not assembled into the device. At least one instruction is stored in the computer-readable storage medium, and the instruction is loaded and executed by the processor to implement Figure 1 the image exposure parameter adjustment method shown. The computer-readable storage medium can be a read-only memory, a magnetic disk, an optical disc, etc.

[0089] It should be noted that the various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. For the same or similar parts among the embodiments, reference can be made to each other. For the apparatus embodiments, device embodiments, and storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple. For the relevant parts, reference can be made to the corresponding descriptions in the method embodiments.

[0090] Also, in this document, the term "comprise", "include" or any other variation thereof is intended to cover non-exclusive inclusion, such that a process, method, article or device that comprises a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the phrase "comprising a..." does not exclude the presence of additional identical elements in the process, method, article or device that comprises the element.

[0091] The above description shows and describes the preferred embodiments of the present invention. It should be understood that the present invention is not limited to the forms disclosed herein, and should not be regarded as excluding other embodiments, but can be used in various other combinations, modifications and environments, and can be changed within the scope of the inventive concept herein through the above teachings or the techniques or knowledge in related fields. Any changes and variations made by those skilled in the art without departing from the spirit and scope of the present invention shall fall within the protection scope of the appended claims of the present invention.

Claims

1. An image exposure parameter adjustment method, characterized in that, The method includes: Obtain an image to be processed, and input the image to be processed into a trained exposure adjustment model to adjust exposure parameters. Among them, the network structure of the exposure adjustment model includes a feature extraction module, a degree modulation module, and a matrix mapping module. The feature extraction module is constructed based on a convolutional neural network and a convolutional neural network with an attention mechanism; Input the image to be processed into the feature extraction module, and respectively extract the local feature and the global feature of the image according to the convolutional neural network and the convolutional neural network with an attention mechanism in the feature extraction module. Further, the step of extracting the local feature of the image according to the convolutional neural network in the feature extraction module includes: Perform different non-linear transformation combinations on the image to be processed according to the convolutional neural network in the feature extraction module to obtain local features; The step of extracting the global feature of the image according to the convolutional neural network with an attention mechanism in the feature extraction module includes: Dynamically combine and extract the extracted local features in an adaptive weighting manner according to the convolutional neural network with an attention mechanism in the feature extraction module to obtain global features; Input preset adjustment parameters into the degree modulation module to obtain a plurality of exposure modulation parameters; Fuse the local feature, the global feature, and the exposure modulation parameters to obtain an affine grid, and obtain an affine matrix by upsampling the affine grid; Input the image to be processed and the affine matrix into the matrix mapping module for mapping operation to obtain a generated image after exposure adjustment.

2. The image exposure parameter adjustment method according to claim 1, characterized in that, The step of dynamically combining and extracting the extracted local features in an adaptive weighting manner according to the convolutional neural network with an attention mechanism in the feature extraction module to obtain global features includes: According to the formula X feature = W l * [softmax(X q ⊙ X k ) ⊙ X v to perform dynamic combination of local features of the image, where ⊙ represents matrix operation, X q , X k , X v represent local feature matrices formed by combining different non-linear transformations of the image X, W l represents the weight parameter of the perceptron, and X feature represents the global feature dynamically extracted by the convolutional neural network of the attention mechanism for the image.

3. The method for adjusting image exposure parameters according to claim 1, wherein The degree modulation module is constructed based on a multi-layer perceptron; the step of inputting preset adjustment parameters into the degree modulation module to obtain a plurality of exposure modulation parameters includes: Input preset adjustment parameters into the multi-layer perceptron, output a plurality of scaling parameters and displacement parameters, and modulate the exposure degree through the plurality of scaling parameters and the displacement parameters to obtain the plurality of exposure modulation parameters.

4. The method for adjusting image exposure parameters according to claim 3, characterized in that, The step of modulating the exposure degree through the plurality of scaling parameters and the displacement parameters to obtain the plurality of exposure modulation parameters includes: Modulate the exposure degree according to the formula C′ = α * C + β, where α represents a scaling parameter, β represents a displacement parameter, C represents a preset exposure adjustment parameter, and C′ represents a modulated exposure modulation parameter.

5. The method for adjusting image exposure parameters according to claim 1, wherein The training process of the exposure adjustment model includes: Optimize the exposure adjustment model using the preset loss function \(L = L_1+L\) 感知 +L fft where \(L_1\) represents the \(L_1\) loss, \(L\) 感知 represents the perceptual loss, and \(L\) fft represents the FFT loss. In the formula, \(x\) represents the target image, \(\hat{x}\) represents the generated image, and \(F\) represents the fast Fourier transform operation on the image.

6. An image exposure parameter adjustment device, characterized in that The device includes: An acquisition unit, configured to obtain an image to be processed, and input the image to be processed into a trained exposure adjustment model to adjust exposure parameters. Among them, the network structure of the exposure adjustment model includes a feature extraction module, a degree modulation module, and a matrix mapping module. The feature extraction module is constructed based on a convolutional neural network and a convolutional neural network with an attention mechanism; A feature extraction unit, configured to input the to-be-processed image into the feature extraction module, and extract the local feature and the global feature of the image according to the convolutional neural network and the convolutional neural network with an attention mechanism in the feature extraction module respectively; further, the extracting the local feature of the image according to the convolutional neural network in the feature extraction module includes: Performing different non-linear transformation combinations on the to-be-processed image according to the convolutional neural network in the feature extraction module to obtain local features; The extracting the global feature of the image according to the convolutional neural network with an attention mechanism in the feature extraction module includes: Dynamically combining and extracting the extracted local features in an adaptive weighting manner according to the convolutional neural network with an attention mechanism in the feature extraction module to obtain global features; A parameter modulation unit, configured to input preset adjustment parameters into the degree modulation module to obtain a plurality of exposure modulation parameters; A fusion processing unit, configured to perform fusion processing on the local feature, the global feature, and the exposure modulation parameters to obtain an affine grid, and obtain an affine matrix by performing an upsampling operation on the affine grid; A mapping operation unit, configured to input the to-be-processed image and the affine matrix into the matrix mapping module for mapping operation to obtain a generated image after exposure adjustment.

7. An image exposure parameter adjustment device, characterized in that, Comprising a processor, a memory, and a computer program stored in the memory, the computer program is executed by the processor to implement the steps of an image exposure parameter adjustment method according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, and the computer program is executed by a processor to implement the steps of an image exposure parameter adjustment method according to any one of claims 1 to 5.

Images