Image Color Enhancement Method, Device and Electronic Device Based on Lookup Table Estimation

Abstract

The present application discloses an image color enhancement method, device and electronic device based on lookup table estimation, which relates to the field of video color enhancement. The method includes: using the feature extraction module of a preset color enhancement model to perform feature extraction on the image to be color enhanced, so as to obtain a pixel intensity feature image and a color feature image; using the lookup table weight estimation module of the preset color enhancement model to perform weight estimation on the pixel intensity feature image and the color feature image respectively, so as to obtain a pixel intensity lookup table weight corresponding to the pixel intensity feature image and an image color lookup table weight corresponding to the color feature image; performing calculation processing based on the pixel intensity lookup table weight, the image color lookup table weight and a preset initial 3D lookup table to obtain a target 3D lookup table; using the target 3D lookup table to perform color enhancement processing on the image to be color enhanced to obtain a target color enhanced image. The method of the present application can make the color enhancement result more stable and natural.

Description

Technical Field

[0001] The present invention relates to the field of video color enhancement, and particularly to an image color enhancement method, device and electronic device based on lookup table estimation. Background Art

[0002] Video color enhancement technology is a key technology in video post-processing, aiming to improve the visual effect of the video image by adjusting parameters such as color saturation, contrast, and brightness of the video image, so as to present a more rich and realistic color performance. Due to the differences in the optical characteristics and sensor performance of different types of imaging devices, problems such as insufficient saturation, color distortion, and color cast may occur during video shooting. These problems are particularly obvious when played on high-definition display devices, resulting in the overall graying of the video image and the dullness of colors, seriously affecting the final visual experience.

[0003] Traditional methods usually rely on manually designed rules or parameters and are difficult to adapt to diverse video content and complex scene requirements. Compared with traditional methods, color enhancement methods based on deep learning can more accurately understand the content of video images, learn more complex color mapping relationships, and achieve automated and efficient processing while ensuring robustness. However, existing video color enhancement methods based on deep learning still face many challenges when dealing with complex real-world scenes. Factors such as lighting changes, object occlusion, and differences in multi-device display characteristics can lead to poor adaptability of the model and it is difficult to maintain a consistent enhancement effect in all scenarios. In addition, deep learning methods usually require a large amount of computing resources and training data, and the computational complexity is relatively high, which limits their deployment in real-time application scenarios. Summary of the Invention

[0004] In view of this, the present invention provides an image color enhancement method, device, storage medium and electronic device based on lookup table estimation, mainly aiming to solve the problems of color cast and uneven color transition in the current image color enhancement process.

[0005] To solve the above problems, the present application provides an image color enhancement method based on lookup table estimation, including:

[0006] Using the feature extraction module of a preset color enhancement model to perform feature extraction on the image to be color enhanced, obtaining the pixel intensity feature image and color feature image of the image to be color enhanced;

[0007] Using the lookup table weight estimation module of the preset color enhancement model to perform weight estimation on the pixel intensity feature image and the color feature image respectively, obtaining the pixel intensity lookup table weight corresponding to the pixel intensity feature image and the image color lookup table weight corresponding to the color feature image;

[0008] Based on the pixel intensity lookup table weights, the image color lookup table weights, and a preset initial 3D lookup table, perform calculation processing to obtain a target 3D lookup table;

[0009] Use the target 3D lookup table to perform color enhancement processing on the image to be color-enhanced, and obtain a target color-enhanced image.

[0010] Optionally, the feature extraction module includes a first-branch feature extraction module and a second-branch feature extraction module. The feature extraction module using a preset color enhancement model extracts features from the image to be color-enhanced to obtain the pixel intensity features and color features of the image to be color-enhanced, specifically including:

[0011] Use the first-branch feature extraction module of the preset color enhancement model to extract features from the image to be color-enhanced to obtain a pixel intensity feature image;

[0012] Use the second-branch feature extraction module of the preset color enhancement model to extract features from the image to be color-enhanced to obtain a color feature image;

[0013] Among them, the first-branch feature extraction module includes a downsampling processing layer and multiple combined feature extraction modules, and each combined feature extraction module includes a feature refinement layer and an adaptive sparse self-attention layer;

[0014] The second-branch feature extraction module includes a feature refinement layer and multiple hybrid-scale convolutional layers.

[0015] Optionally, using the first-branch feature extraction module of the preset color enhancement model to extract features from the image to be color-enhanced to obtain a pixel intensity feature image specifically includes:

[0016] Use the downsampling processing layer of the first-branch feature extraction module to perform downsampling processing on the image to be color-enhanced to obtain a first image with a predetermined sampling multiple;

[0017] Use the feature refinement layer of the first combined feature extraction module to refine the features of the first image to obtain a first refined feature image;

[0018] Use the adaptive sparse self-attention layer of the first combined feature extraction module to capture global context-aware information from the first refined feature image to obtain an initial pixel intensity feature image;

[0019] Based on the initial pixel intensity feature image, successively use non-first combined feature extraction modules to perform image feature extraction processing to obtain the pixel intensity feature image.

[0020] Optionally, the feature refinement layer of the first combined feature extraction module is used to refine the features of the first image to obtain a first refined feature image, specifically including:

[0021] Perform normalization processing on the first image to obtain a first feature image;

[0022] Based on the first feature image, use a partial convolution network to perform key information enhancement processing to obtain a second feature image;

[0023] Perform linear transformation processing on the second feature image to obtain a third feature image;

[0024] Based on the third feature image, use a depthwise separable convolution network to perform local feature extraction to obtain a fourth feature image;

[0025] Based on the third feature image, use a partial convolution network to perform key information enhancement processing to obtain a fifth feature image;

[0026] Perform multiplication operation processing on the fourth feature image and the fifth feature image to obtain an initial refined feature image;

[0027] Use a preset GELU activation function to optimize the initial refined feature image to obtain the first refined feature image.

[0028] Optionally, the adaptive sparse self-attention layer of the first combined feature extraction module is used to capture global context-aware information of the first refined feature image to obtain an initial pixel intensity feature image, specifically including:

[0029] Perform normalization processing on the first refined feature image to obtain a first pixel feature image;

[0030] Perform projection processing on the first pixel feature image to obtain the query matrix, key matrix, and value matrix of the first pixel feature image;

[0031] Based on the query matrix, the key matrix, a learnable bias matrix, and a predetermined scaling factor, perform calculation processing using a first activation function to obtain a dense self-attention score;

[0032] Based on the query matrix, the key matrix, a learnable bias matrix, and a predetermined scaling factor, perform calculation processing using a second activation function to obtain a sparse self-attention score;

[0033] Based on the value matrix, the dense self-attention score, the sparse self-attention score, a first learnable weight matrix, and a second learnable weight matrix, perform calculation processing to obtain an attention matrix;

[0034] Performing an addition operation on the basis of the attention matrix and the first refined feature image to obtain the initial pixel intensity feature image.

[0035] Optionally, the second branch feature extraction module of the preset color enhancement model is used to extract features from the image to be color-enhanced to obtain a color feature image, which specifically includes:

[0036] Using the feature refinement layer of the second branch feature extraction module to refine the features of the color feature image to obtain a second refined feature image;

[0037] Performing a normalization process on the second refined feature image to obtain a normalized second refined feature image;

[0038] Separately using deep convolutional networks with convolution kernels of 3×3 and 5×5 to perform local feature extraction on the second refined feature image to obtain a first color local feature image and a second color local feature image;

[0039] Performing feature fusion on the first color local feature image and the second color local feature image to obtain a color fusion image;

[0040] Separately using deep convolutional networks with convolution kernels of 3×3 and 5×5 to perform local feature extraction on the color fusion image to obtain a third color local feature image and a fourth color local feature image;

[0041] Performing image stitching on the basis of the third color local feature image and the fourth color local feature image to obtain the color feature image.

[0042] Optionally, the lookup table weight estimation module of the preset color enhancement model is used to perform weight estimation on the pixel intensity feature image and the color feature image respectively to obtain a pixel intensity lookup table weight corresponding to the pixel intensity feature image and an image color lookup table weight corresponding to the color feature image, which specifically includes:

[0043] Using the global branch of the lookup table weight estimation module of the preset color enhancement model to extract features from the pixel intensity feature image and the color feature image respectively to obtain a global pixel branch feature image corresponding to the pixel intensity feature image and a global color branch feature image corresponding to the color feature image;

[0044] Using the local branch of the lookup table weight estimation module of the preset color enhancement model to extract features from the pixel intensity feature image and the color feature image respectively to obtain a local pixel branch feature image corresponding to the pixel intensity feature image and a local color branch feature image corresponding to the color feature image;

[0045] Perform feature fusion on the global pixel branch feature image and the local pixel branch feature image to obtain a pixel fusion feature image;

[0046] Perform feature fusion on the global color branch feature image and the local color branch feature image to obtain a color fusion feature image;

[0047] Perform non-linear transformation processing on the pixel fusion feature image and the color fusion feature image respectively to obtain a pixel intensity lookup table weight corresponding to the pixel intensity feature image and an image color lookup table weight corresponding to the color feature image.

[0048] To solve the above problems, the present application provides an image color enhancement device based on lookup table estimation, including:

[0049] A feature extraction module, configured to use the feature extraction module of a preset color enhancement model to perform feature extraction on an image to be color enhanced, and obtain a pixel intensity feature image and a color feature image of the image to be color enhanced;

[0050] A weight estimation module, configured to use the lookup table weight estimation module of the preset color enhancement model to perform weight estimation on the pixel intensity feature image and the color feature image respectively, and obtain a pixel intensity lookup table weight corresponding to the pixel intensity feature image and an image color lookup table weight corresponding to the color feature image;

[0051] A calculation module, configured to perform calculation processing based on the pixel intensity lookup table weight, the image color lookup table weight, and a preset initial 3D lookup table to obtain a target 3D lookup table;

[0052] A color enhancement module, configured to use the target 3D lookup table to perform color enhancement processing on the image to be color enhanced to obtain a target color enhanced image.

[0053] To solve the above problems, the present application provides a storage medium storing a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned image color enhancement method based on lookup table estimation are implemented.

[0054] To solve the above problems, the present application provides an electronic device, including at least a memory and a processor, where the memory stores a computer program, and when the processor executes the computer program on the memory, the steps of the above-mentioned image color enhancement method based on lookup table estimation are implemented.

[0055] Advantages of the present application: Through the adaptive lookup table weight estimation module, two lookup table weights are generated respectively based on pixel intensity features and color features, and multiplied by the initialized 3D lookup table to adaptively adjust the color mapping relationship. It can dynamically adjust color enhancement parameters according to the content of the input image, so as to adapt to color differences in different scenes, different lighting conditions, and different device imaging, making the color enhancement results more stable and natural, while avoiding problems such as color deviation and uneven color transition that may occur in existing methods.

[0056] The above description is only an overview of the technical solution of the present invention. In order to understand the technical means of the present invention more clearly, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features and advantages of the present invention more obvious and understandable, the following specifically describes the embodiments of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] By reading the following detailed description of the preferred embodiments, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of showing the preferred embodiments and are not considered to be a limitation of the present invention. And throughout the drawings, the same reference numerals are used to represent the same components. In the drawings:

[0058] Figure 1 A flowchart showing a method for image color enhancement based on lookup table estimation provided by an embodiment of the present application;

[0059] Figure 2 A flowchart showing a method for image color enhancement based on lookup table estimation provided by another embodiment of the present application;

[0060] Figure 3 A schematic structural diagram showing a feature refinement layer of a preset color enhancement model provided by an embodiment of the present application;

[0061] Figure 4 A schematic structural diagram showing an adaptive sparse self-attention layer of a preset color enhancement model provided by an embodiment of the present application;

[0062] Figure 5 A schematic structural diagram showing a model structure of a preset color enhancement model provided by an embodiment of the present application;

[0063] Figure 6 A schematic structural diagram showing a hybrid scale convolution layer of a preset color enhancement model provided by an embodiment of the present application;

[0064] Figure 7 A schematic structural diagram showing a lookup table estimation module of a preset color enhancement model provided by an embodiment of the present application;

[0065] Figure 8 A structural block diagram of an image color enhancement device based on lookup table estimation provided by another embodiment of the present application is shown. DETAILED DESCRIPTION

[0066] Various aspects and features of the present application are described herein with reference to the accompanying drawings.

[0067] It should be understood that various modifications may be made to the embodiments of the present application. Therefore, the above description should not be considered as limiting, but merely as an example of an embodiment. Other modifications within the scope and spirit of the present application will occur to those skilled in the art.

[0068] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments of the present application and, together with the general description of the present application given above and the detailed description of the embodiments given below, serve to explain the principles of the present application.

[0069] These and other characteristics of the present application will become apparent from the following description of a preferred form of embodiment given as a non-limiting example with reference to the accompanying drawings.

[0070] It should also be understood that although the present application has been described with reference to certain specific examples, those skilled in the art will readily be able to implement many other equivalent forms of the present application.

[0071] The above and other aspects, features and advantages of the present application will become more apparent from the following detailed description when taken in conjunction with the accompanying drawings.

[0072] Specific embodiments of the present application will be described hereinafter with reference to the accompanying drawings; however, it should be understood that the embodiments described are merely examples of the present application and may be implemented in a variety of ways. Well-known and / or repetitive functions and structures are not described in detail to avoid obscuring the present application with unnecessary or redundant detail. Therefore, the specific structural and functional details described herein are not intended to be limiting, but rather serve merely as a representative basis to teach those skilled in the art to variously employ the present application with substantially any suitable detailed structure.

[0073] This specification may use the phrases "in one embodiment," "in another embodiment," "in yet another embodiment," or "in other embodiments," which may all refer to one or more of the same or different embodiments according to the present application.

[0074] The embodiment of the present application provides an image color enhancement method based on lookup table estimation, such as Figure 1 Shown, including:

[0075] Step S101: Use the feature extraction module of the preset color enhancement model to extract features from the image to be color-enhanced, and obtain the pixel intensity feature image and color feature image of the image to be color-enhanced;

[0076] In the specific implementation process, the preset color enhancement model includes a feature extraction module and a lookup table weight estimation module; the feature extraction module includes a first-branch feature extraction module and a second-branch feature extraction module. Use the first-branch feature extraction module of the preset color enhancement model to extract features from the image to be color-enhanced to obtain a pixel intensity feature image; use the second-branch feature extraction module of the preset color enhancement model to extract features from the image to be color-enhanced to obtain a color feature image; the preset color enhancement model is divided into two branches, which respectively process image information of different scales. The first-branch feature extraction module first downsamples the input image to be color-enhanced by a predetermined multiple to obtain the downsampled image. Then, the downsampled image first passes through a feature refinement module to obtain refined image features, and the refined image features capture the global context-aware information of the image through an adaptive sparse self-attention module to obtain a pixel intensity feature image. The second-branch feature extraction module directly uses the image to be color-enhanced with the original size as the input, obtains refined image features through a feature refinement module, and then the refined image features extract the color features of the image through a mixed-scale convolution module to obtain a color feature image.

[0077] Step S102: Use the lookup table weight estimation module of the preset color enhancement model to perform weight estimation on the pixel intensity feature image and the color feature image respectively, and obtain the pixel intensity lookup table weight corresponding to the pixel intensity feature image and the image color lookup table weight corresponding to the color feature image;

[0078] In the specific implementation process, the global branch of the lookup table weight estimation module of the preset color enhancement model is used to extract features from the pixel intensity feature image and the color feature image respectively, obtaining a global pixel branch feature image corresponding to the pixel intensity feature image and a global color branch feature image corresponding to the color feature image; the local branch of the lookup table weight estimation module of the preset color enhancement model is used to extract features from the pixel intensity feature image and the color feature image respectively, obtaining a local pixel branch feature image corresponding to the pixel intensity feature image and a local color branch feature image corresponding to the color feature image; the global pixel branch feature image and the local pixel branch feature image are subjected to feature fusion to obtain a pixel fusion feature image; the global color branch feature image and the local color branch feature image are subjected to feature fusion to obtain a color fusion feature image; the pixel fusion feature image and the color fusion feature image are respectively subjected to non-linear transformation processing to obtain a pixel intensity lookup table weight corresponding to the pixel intensity feature image and an image color lookup table weight corresponding to the color feature image.

[0079] Step S103: Based on the pixel intensity lookup table weight, the image color lookup table weight, and a preset initial 3D lookup table, perform calculation processing to obtain a target 3D lookup table;

[0080] In the specific implementation process, multiplication operation processing is performed based on the pixel intensity lookup table weight, the image color lookup table weight, and a preset initial 3D lookup table to obtain a target 3D lookup table. The initialized 3D lookup table refers to an initially unit mapping lookup table, where the input color coordinates (r, g, b) remain unchanged after being mapped by the lookup table. For example: for an N×N×N×3 lookup table, uniform sampling coordinates in the input color range [0, 255] can be constructed first, and then each entry of the lookup table is set to the input coordinate itself to obtain a unit mapping lookup table.

[0081] Step S104: Use the target 3D lookup table to perform color enhancement processing on the image to be color enhanced, obtaining a target color enhanced image.

[0082] In the specific implementation process, the final target 3D lookup table is used to perform color mapping transformation on the image to be color enhanced, obtaining a target color enhanced image after color enhancement.

[0083] Through the adaptive lookup table weight estimation module of the present application, two lookup table weights are generated respectively based on pixel intensity features and color features, and multiplied by the initialized 3D lookup table to adaptively adjust the color mapping relationship. It can dynamically adjust the color enhancement parameters according to the content of the input image, so as to adapt to the color differences in different scenes, different lighting conditions and different device imaging, making the color enhancement result more stable and natural, and at the same time avoiding problems such as color deviation and uneven color transition that may occur in the existing methods.

[0084] Another embodiment of the present application provides another image color enhancement method based on lookup table estimation, as Figure 2 shown, including:

[0085] Step S201: Use the first branch feature extraction module of the preset color enhancement model to extract features from the image to be color enhanced, and obtain a pixel intensity feature image;

[0086] In the specific implementation process of this step, the preset color enhancement model includes a feature extraction module and a lookup table weight estimation module; the feature extraction module includes a first branch feature extraction module and a second branch feature extraction module, and the first branch feature extraction module includes a downsampling processing layer and multiple combined feature extraction modules, and each combined feature extraction module includes a feature refinement layer and an adaptive sparse self-attention layer; specifically, use the downsampling processing layer of the first branch feature extraction module to perform downsampling processing on the image to be color enhanced to obtain a first image with a predetermined sampling multiple; the predetermined multiple can be four times, and the image to be color enhanced is downsampled four times to obtain the first image after being downsampled four times. Use the feature refinement layer of the first combined feature extraction module to refine the features of the first image to obtain a first refined feature image; the structural schematic diagram of the feature refinement layer is as Figure 3As shown in the figure, the feature refinement layer includes a partial convolution network (P convolution layer), depthwise separable convolution, a gating mechanism, and a GELU activation function; the first image is normalized to obtain a first feature image; based on the first feature image, key information is enhanced using a partial convolution network to obtain a second feature image; P convolution, also known as partial convolution, is mainly used to enhance key information while reducing the propagation of invalid information. The second feature image is linearly transformed to obtain a third feature image; based on the third feature image, local features are extracted using a depthwise separable convolution network to obtain a fourth feature image; depthwise separable convolution performs efficient local feature extraction in the channel dimension, reducing the computational load while enhancing local information. Based on the third feature image, key information is enhanced using a partial convolution network to obtain a fifth feature image; key information is enhanced again while reducing the propagation of invalid information. Based on the fourth feature image and the fifth feature image, a multiplication operation is performed to obtain an initial refined feature image; the gating mechanism is used to weight and adjust the features passing through the depthwise separable convolution branch in the feature refinement module for another branch of features to enhance key information and weaken redundant features. The initial refined feature image is optimized using a preset GELU activation function to obtain the first refined feature image. The GELU activation function is used to optimize the final feature distribution, thereby improving the feature representation ability of the model while maintaining computational efficiency and providing a better feature representation for subsequent color enhancement tasks. The adaptive sparse self-attention layer of the first combined feature extraction module is used to capture global context-aware information from the first refined feature image to obtain an initial pixel intensity feature image; the structural schematic diagram of the adaptive sparse self-attention layer is as shown in Figure 4 As shown, specifically, the first refined feature image is normalized to obtain a first pixel feature image; the first pixel feature image is projected to obtain a query matrix of the first pixel feature image, a key matrix and a value matrix V ; based on the query matrix, the key matrix, a learnable bias matrix, and a predetermined scaling factor, a first activation function is used for calculation to obtain a dense self-attention score ; the mathematical expression of the dense self-attention score can be shown by the following formula (1):

[0087] (1)

[0088] where, represents the query matrix, represents the key matrix, represents the learnable bias matrix, represents the scaling factor, is the first activation function. Based on the query matrix, the key matrix, the learnable bias matrix, and a predetermined scaling factor, a second activation function is used for calculation processing to obtain the sparse self-attention scores ; The mathematical expression of the dense self-attention scores can be shown by the following formula (2):

[0089] (2)

[0090] Wherein, represents the query matrix, represents the key matrix, represents the learnable bias matrix, represents the scaling factor, is the second activation function. The learnable bias matrix adopts the method of random initialization with a normal distribution. The mean of the normal distribution is 0, and the standard deviation is 0.01. The initialized values are between [-0.02, 0.02]. Based on the value matrix, the dense self-attention scores, the sparse self-attention scores, the first learnable weight matrix and the second learnable weight matrix calculation processing is performed to obtain the attention matrix ; The calculation mathematical formula of the attention matrix can be shown by the following formula (3):

[0091] (3)

[0092] Wherein, represents the attention matrix calculated in the adaptive sparse self-attention module, represents the value matrix. Here, the two learnable weight matrices and both adopt the method of random initialization with a normal distribution. The mean of the normal distribution is 0, and the standard deviation is 0.01. The initialized values are between [-0.02, 0.02]. Based on the addition operation between the attention matrix and the first refined feature image, the initial pixel intensity feature image is obtained. Based on the initial pixel intensity feature image, the non-first combined feature extraction module is used one by one for image feature extraction processing to obtain the pixel intensity feature image. As Figure 5 shown is the schematic diagram of the model structure of the preset color enhancement model of the present application. The number of combined feature extraction modules of the present application is 2-4 in series, which can have high learning ability while ensuring low computational complexity.

[0093] Step S202: Use the second branch feature extraction module of the preset color enhancement model to perform feature extraction on the image to be color enhanced to obtain a color feature image;

[0094] In the specific implementation process of this step, the second branch feature extraction module includes a feature refinement layer and multiple mixed-scale convolutional layers. The feature refinement layer of the second branch feature extraction module is used to refine the features of the color feature image to obtain a second refined feature image. The number of the mixed-scale convolutional layers is 2-4 in series, which can have high learning ability while ensuring relatively low computational complexity. The second refined feature image is normalized to obtain a normalized second refined feature image. The structural schematic diagram of the mixed-scale convolutional layer is as follows Figure 6 shown, the mixed-scale convolutional module first normalizes the input features, and then expands the channel dimension through two 1×1 depthwise separable convolution branches to form two parallel paths. The depth convolutional networks with convolution kernels of 3×3 and 5×5 are respectively used to extract local features of the second refined feature image, and calculations are performed through the ReLU activation function to obtain a first color local feature image and a second color local feature image. Feature fusion is performed on the first color local feature image and the second color local feature image to obtain a color fusion image. The depth convolutional networks with convolution kernels of 3×3 and 5×5 are respectively used to extract local features of the color fusion image, and calculations are performed again through the ReLU activation function to obtain a third color local feature image and a fourth color local feature image. Image stitching processing is performed based on the third color local feature image and the fourth color local feature image to obtain the color feature image. The stitching processing process is to perform pixel addition along the channel dimension of the third color local feature image and the fourth color local feature image. Then, channel compression is performed through a 1×1 ordinary convolution to obtain output features, so as to reduce the computational complexity and adjust the feature distribution. Finally, the input features and the output features are added through a residual connection to obtain the output of the mixed-scale convolutional layer. The output of the previous mixed-scale convolutional layer is used as the input of the next mixed-scale convolutional layer until the output of the last mixed-scale convolutional layer obtains the color feature image.

[0095] Step S203: The global branches of the lookup table weight estimation module of the preset color enhancement model are respectively used to extract features from the pixel intensity feature image and the color feature image to obtain a global pixel branch feature image corresponding to the pixel intensity feature image and a global color branch feature image corresponding to the color feature image;

[0096] In the specific implementation process of this step, the structural schematic diagram of the lookup table estimation module is as follows Figure 7As shown, the lookup table estimation module includes a global branch and a local branch; the global branch includes an average pooling layer, a 1×1 depth convolution layer, and a Reshape layer; specifically, the average pooling layer of the global branch is used to perform average pooling on the pixel intensity feature image to obtain a first pooled image; based on the first pooled image, a convolution operation is performed using the 1×1 depth convolution layer of the global branch to obtain a first depth convolution image; based on the first depth convolution image, feature reshaping is performed using the Reshape layer of the global branch to obtain the global pixel branch feature; the average pooling layer of the global branch is used to perform average pooling on the color feature image to obtain a second pooled image; based on the second pooled image, a convolution operation is performed using the 1×1 depth convolution layer of the global branch to obtain a second depth convolution image; based on the second depth convolution image, feature reshaping is performed using the Reshape layer of the global branch to obtain the global color branch feature image.

[0097] Step S204: The local branch of the lookup table weight estimation module of the preset color enhancement model is used to extract features from the pixel intensity feature image and the color feature image respectively, to obtain a local pixel branch feature image corresponding to the pixel intensity feature image and a local color branch feature image corresponding to the color feature image;

[0098] In the specific implementation process of this step, the local branch includes a 1×1 depth convolution layer and a Reshape layer; a convolution operation is performed on the pixel intensity feature image using the 1×1 depth convolution layer of the local branch to obtain a third depth convolution image; based on the third depth convolution image, feature reshaping is performed using the Reshape layer of the local branch to obtain the local pixel branch feature image; a convolution operation is performed on the color feature image using the 1×1 depth convolution layer of the local branch to obtain a fourth depth convolution image; based on the fourth depth convolution image, feature reshaping is performed using the Reshape layer of the local branch to obtain the local color branch feature image;

[0099] Step S205: Feature fusion is performed on the global pixel branch feature image and the local pixel branch feature image to obtain a pixel fusion feature image;

[0100] In the specific implementation process of this step, an element-wise addition operation is performed on the global pixel branch feature image and the local pixel branch feature image to obtain a pixel fusion feature image.

[0101] Step S206: Feature fusion is performed on the global color branch feature image and the local color branch feature image to obtain a color fusion feature image;

[0102] In the specific implementation process of this step, an element-wise addition operation is performed on the global color branch feature image and the local color branch feature image to obtain a color fusion feature image.

[0103] Step S207: Perform non-linear transformation processing on the pixel fusion feature image and the color fusion feature image respectively to obtain a pixel intensity lookup table weight corresponding to the pixel intensity feature image and an image color lookup table weight corresponding to the color feature image;

[0104] In the specific implementation process of this step, based on the pixel fusion feature image, non-linear transformation processing is performed using the fully connected layer (FC) and GELU activation function of the lookup table estimation module to obtain a pixel intensity lookup table weight corresponding to the pixel intensity feature image; based on the color fusion feature image, non-linear transformation processing is performed using the fully connected layer (FC) and GELU activation function of the lookup table estimation module to obtain an image color lookup table weight corresponding to the color feature image. The lookup table estimation module has a compact structure and high computational efficiency, reducing the computational overhead while ensuring the lookup table estimation ability, and is suitable for lightweight tasks.

[0105] Step S208: Perform calculation processing based on the pixel intensity lookup table weight, the image color lookup table weight, and a preset initial 3D lookup table to obtain a target 3D lookup table;

[0106] In the specific implementation process of this step, the pixel intensity lookup table weight and the image color lookup table weight are respectively multiplied by the initialized preset initial 3D lookup table to obtain the final target 3D lookup table.

[0107] Step S209: Perform color enhancement processing on the image to be color-enhanced using the target 3D lookup table to obtain a target color-enhanced image.

[0108] In the specific implementation process of this step, the image to be color-enhanced is subjected to color mapping transformation using the final target 3D lookup table to obtain the target color-enhanced image after color enhancement.

[0109] In this application, the first-branch feature extraction module of the preset color enhancement model is used to extract features from the image to be color-enhanced, obtaining a pixel intensity feature image; the second-branch feature extraction module of the preset color enhancement model is used to extract features from the image to be color-enhanced, obtaining a color feature image; a dual-branch structure is adopted to process pixel intensity features and color features respectively, realizing efficient modeling of different-scale information. The first-branch feature extraction module can capture the global context information of the image through the adaptive sparse self-attention mechanism, enhancing the accuracy of pixel intensity estimation; the second-branch feature extraction module combines a hybrid-scale convolutional layer to effectively extract the local color information of the image, making the color enhancement more consistent and hierarchical. Through this method of dual-branch collaborative modeling, this application can not only restore the overall color style of the image but also maintain the color details of local regions, improving the authenticity and naturalness of color restoration. This application designs an adaptive lookup table weight estimation module to generate two lookup table weights based on pixel intensity features and color features respectively, and multiply them with the initialized 3D lookup table to adaptively adjust the color mapping relationship. This method can dynamically adjust color enhancement parameters according to the content of the input image, thereby adapting to color differences in different scenarios, different lighting conditions, and different device imaging, making the color enhancement results more stable and natural, and avoiding problems such as color deviation and uneven color transition that may occur in existing methods.

[0110] Another embodiment of this application provides an image color enhancement device based on lookup table estimation, as Figure 8 shown, including:

[0111] A feature extraction module 1, configured to use the feature extraction module of the preset color enhancement model to extract features from the image to be color-enhanced, obtaining the pixel intensity feature image and the color feature image of the image to be color-enhanced;

[0112] A weight estimation module 2, configured to use the lookup table weight estimation module of the preset color enhancement model to perform weight estimation on the pixel intensity feature image and the color feature image respectively, obtaining a pixel intensity lookup table weight corresponding to the pixel intensity feature image and an image color lookup table weight corresponding to the color feature image;

[0113] A calculation module 3, configured to perform calculation processing based on the pixel intensity lookup table weight, the image color lookup table weight, and the preset initial 3D lookup table, obtaining a target 3D lookup table;

[0114] A color enhancement module 4, configured to use the target 3D lookup table to perform color enhancement processing on the image to be color-enhanced, obtaining a target color-enhanced image.

[0115] In the specific implementation process, the feature extraction module 1 is specifically used for: the feature extraction module includes a first-branch feature extraction module and a second-branch feature extraction module. The feature extraction module using a preset color enhancement model extracts features from the image to be color-enhanced to obtain the pixel intensity features and color features of the image to be color-enhanced, specifically including: the first-branch feature extraction module using a preset color enhancement model extracts features from the image to be color-enhanced to obtain a pixel intensity feature image; the second-branch feature extraction module using a preset color enhancement model extracts features from the image to be color-enhanced to obtain a color feature image; wherein, the first-branch feature extraction module includes a downsampling processing layer and multiple combined feature extraction modules, and each combined feature extraction module includes a feature refinement layer and an adaptive sparse self-attention layer; the second-branch feature extraction module includes a feature refinement layer and multiple mixed-scale convolutional layers.

[0116] In the specific implementation process, the feature extraction module 1 is further used for: using the downsampling processing layer of the first-branch feature extraction module to perform downsampling processing on the image to be color-enhanced to obtain a first image with a predetermined sampling multiple; using the feature refinement layer of the first combined feature extraction module to perform feature refinement on the first image to obtain a first refined feature image; using the adaptive sparse self-attention layer of the first combined feature extraction module to capture global context awareness information from the first refined feature image to obtain an initial pixel intensity feature image; based on the initial pixel intensity feature image, successively using non-first combined feature extraction modules to perform image feature extraction processing to obtain the pixel intensity feature image.

[0117] In the specific implementation process, the feature extraction module 1 is further used for: performing normalization processing on the first image to obtain a first feature image; based on the first feature image, using a partial convolution network to perform key information enhancement processing to obtain a second feature image; performing linear transformation processing on the second feature image to obtain a third feature image; based on the third feature image, using a depthwise separable convolution network to perform local feature extraction to obtain a fourth feature image; based on the third feature image, using a partial convolution network to perform key information enhancement processing to obtain a fifth feature image; performing multiplication operation processing on the fourth feature image and the fifth feature image to obtain an initial refined feature image; using a preset GELU activation function to optimize the initial refined feature image to obtain the first refined feature image.

[0118] In the specific implementation process, the feature extraction module 1 is further configured to: perform normalization processing on the first refined feature image to obtain a first pixel feature image; perform projection processing on the first pixel feature image to obtain a query matrix, a key matrix, and a value matrix of the first pixel feature image; perform calculation processing using a first activation function based on the query matrix, the key matrix, a learnable bias matrix, and a predetermined scaling factor to obtain a dense self-attention score; perform calculation processing using a second activation function based on the query matrix, the key matrix, a learnable bias matrix, and a predetermined scaling factor to obtain a sparse self-attention score; perform calculation processing based on the value matrix, the dense self-attention score, the sparse self-attention score, a first learnable weight matrix, and a second learnable weight matrix to obtain an attention matrix; perform addition operation processing based on the attention matrix and the first refined feature image to obtain the initial pixel intensity feature image.

[0119] In the specific implementation process, the feature extraction module 1 is further configured to: use the feature refinement layer of the second branch feature extraction module to refine the color feature image to obtain a second refined feature image; perform normalization processing on the second refined feature image to obtain a normalized second refined feature image; respectively use deep convolutional networks with convolution kernels of 3×3 and 5×5 to perform local feature extraction on the second refined feature image to obtain a first color local feature image and a second color local feature image; perform feature fusion on the first color local feature image and the second color local feature image to obtain a color fusion image; respectively use deep convolutional networks with convolution kernels of 3×3 and 5×5 to perform local feature extraction on the color fusion image to obtain a third color local feature image and a fourth color local feature image; perform image stitching processing based on the third color local feature image and the fourth color local feature image to obtain the color feature image.

[0120] In the specific implementation process, the weight estimation module 2 is specifically configured to: respectively perform feature extraction on the pixel intensity feature image and the color feature image by using the global branch of the lookup table weight estimation module of the preset color enhancement model, so as to obtain a global pixel branch feature image corresponding to the pixel intensity feature image and a global color branch feature image corresponding to the color feature image; respectively perform feature extraction on the pixel intensity feature image and the color feature image by using the local branch of the lookup table weight estimation module of the preset color enhancement model, so as to obtain a local pixel branch feature image corresponding to the pixel intensity feature image and a local color branch feature image corresponding to the color feature image; perform feature fusion on the global pixel branch feature image and the local pixel branch feature image to obtain a pixel fusion feature image; perform feature fusion on the global color branch feature image and the local color branch feature image to obtain a color fusion feature image; respectively perform non-linear transformation processing on the pixel fusion feature image and the color fusion feature image to obtain a pixel intensity lookup table weight corresponding to the pixel intensity feature image and an image color lookup table weight corresponding to the color feature image.

[0121] In this application, through the adaptive lookup table weight estimation module, two lookup table weights are respectively generated based on the pixel intensity feature and the color feature, and multiplied by the initialized 3D lookup table to adaptively adjust the color mapping relationship. It can dynamically adjust the color enhancement parameters according to the content of the input image, so as to adapt to the color differences in different scenes, different lighting conditions, and different device imaging, making the color enhancement result more stable and natural, and at the same time avoiding problems such as color deviation and uneven color transition that may occur in the existing methods.

[0122] Another embodiment of this application provides a storage medium, which stores a computer program. When the computer program is executed by a processor, the following method steps are implemented:

[0123] Step 1: Use the feature extraction module of the preset color enhancement model to perform feature extraction on the image to be color-enhanced, so as to obtain the pixel intensity feature image and the color feature image of the image to be color-enhanced;

[0124] Step 2: Use the lookup table weight estimation module of the preset color enhancement model to respectively perform weight estimation on the pixel intensity feature image and the color feature image, so as to obtain a pixel intensity lookup table weight corresponding to the pixel intensity feature image and an image color lookup table weight corresponding to the color feature image;

[0125] Step 3: Perform calculation processing based on the pixel intensity lookup table weight, the image color lookup table weight, and the preset initial 3D lookup table to obtain a target 3D lookup table;

[0126] Step 4: Use the target 3D lookup table to perform color enhancement processing on the image to be color enhanced, and obtain a target color enhanced image.

[0127] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the various embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0128] Those skilled in the art can clearly understand that for the convenience and brevity of description, only the above-mentioned division of each functional unit and module is used as an example. In actual applications, the above functions can be allocated to different functional units and modules according to needs, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0129] For the specific implementation process of the above method steps, reference can be made to the embodiments of any of the above image color enhancement methods based on lookup table estimation. This embodiment will not be repeated here.

[0130] Through the adaptive lookup table weight estimation module, the present application generates two lookup table weights based on pixel intensity features and color features respectively, and multiplies them with the initialized 3D lookup table to adaptively adjust the color mapping relationship. It can dynamically adjust the color enhancement parameters according to the content of the input image, so as to adapt to the color differences of different scenes, different lighting conditions, and different device imaging, making the color enhancement results more stable and natural, and at the same time avoiding problems such as color deviation and uneven color transition that may occur in existing methods.

[0131] Another embodiment of the present application provides an electronic device, which may be a server. The electronic device includes a processor, a memory, a network interface, and a database connected through a system bus. Among them, the processor of the electronic device is used to provide computing and control capabilities. The memory of the electronic device includes a non-volatile and / or volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the electronic device is used to communicate with an external client through a network connection. When the electronic device program is executed by the processor, it realizes the functions or steps of the server side of an image color enhancement method based on lookup table estimation.

[0132] In one embodiment, an electronic device is provided, which may be a client. The electronic device includes a processor, a memory, a network interface, a display screen, and an input device connected through a system bus. Among them, the processor of the electronic device is used to provide computing and control capabilities. The memory of the electronic device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the electronic device is used to communicate with an external server through a network connection. When the electronic device program is executed by the processor, it realizes the functions or steps of the client side of an image color enhancement method based on lookup table estimation.

[0133] Another embodiment of the present application provides an electronic device, at least including a memory and a processor. A computer program is stored on the memory. When the processor executes the computer program on the memory, the following method steps are implemented:

[0134] Step 1: Use the feature extraction module of the preset color enhancement model to perform feature extraction on the image to be color enhanced, and obtain the pixel intensity feature image and the color feature image of the image to be color enhanced;

[0135] Step 2: Use the lookup table weight estimation module of the preset color enhancement model to perform weight estimation on the pixel intensity feature image and the color feature image respectively, and obtain the pixel intensity lookup table weight corresponding to the pixel intensity feature image and the image color lookup table weight corresponding to the color feature image;

[0136] Step 3: Perform calculation processing based on the pixel intensity lookup table weight, the image color lookup table weight, and a preset initial 3D lookup table to obtain a target 3D lookup table;

[0137] Step 4: Use the target 3D lookup table to perform color enhancement processing on the image to be color enhanced to obtain a target color enhanced image.

[0138] For the specific implementation process of the above method steps, reference can be made to the embodiments of any of the above image color enhancement methods based on lookup table estimation. These embodiments will not be repeated here.

[0139] In this application, an adaptive lookup table weight estimation module generates two lookup table weights respectively based on pixel intensity features and color features, and multiplies them with the initialized 3D lookup table to adaptively adjust the color mapping relationship. It can dynamically adjust the color enhancement parameters according to the content of the input image, so as to adapt to the color differences in different scenes, different lighting conditions, and different device imaging, making the color enhancement results more stable and natural, while avoiding problems such as color deviation and uneven color transition that may occur in existing methods.

[0140] The above embodiments are only exemplary embodiments of this application and are not used to limit this application. Those skilled in the art can make various modifications or equivalent replacements to this application within the essence and protection scope of this application, and such modifications or equivalent replacements should also be regarded as falling within the protection scope of this application.

Claims

1. An image color enhancement method based on lookup table estimation, characterized in that, Including: A feature extraction module using a preset color enhancement model extracts features from the image to be color-enhanced, obtaining a pixel intensity feature image and a color feature image of the image to be color-enhanced. The feature extraction module includes a first-branch feature extraction module and a second-branch feature extraction module; Specifically including: The first-branch feature extraction module of the preset color enhancement model extracts features from the image to be color-enhanced, obtaining a pixel intensity feature image; The second-branch feature extraction module of the preset color enhancement model extracts features from the image to be color-enhanced, obtaining a color feature image; Among them, the first-branch feature extraction module includes a downsampling processing layer and multiple combined feature extraction modules. Each combined feature extraction module includes a feature refinement layer and an adaptive sparse self-attention layer for capturing global context awareness information; The second-branch feature extraction module includes a feature refinement layer and multiple mixed-scale convolutional layers for capturing local color information; The lookup table weight estimation module of the preset color enhancement model respectively performs weight estimation on the pixel intensity feature image and the color feature image, obtaining a pixel intensity lookup table weight corresponding to the pixel intensity feature image and an image color lookup table weight corresponding to the color feature image; Specifically including: The global branch of the lookup table weight estimation module of the preset color enhancement model respectively extracts features from the pixel intensity feature image and the color feature image, obtaining a global pixel branch feature image corresponding to the pixel intensity feature image and a global color branch feature image corresponding to the color feature image; The local branch of the lookup table weight estimation module of the preset color enhancement model respectively extracts features from the pixel intensity feature image and the color feature image, obtaining a local pixel branch feature image corresponding to the pixel intensity feature image and a local color branch feature image corresponding to the color feature image; Feature fusion is performed on the global pixel branch feature image and the local pixel branch feature image, obtaining a pixel fusion feature image; Feature fusion is performed on the global color branch feature image and the local color branch feature image, obtaining a color fusion feature image; Nonlinear transformation processing is respectively performed on the pixel fusion feature image and the color fusion feature image, obtaining a pixel intensity lookup table weight corresponding to the pixel intensity feature image and an image color lookup table weight corresponding to the color feature image; Based on the pixel intensity lookup table weight, the image color lookup table weight, and a preset initial 3D lookup table, calculation processing is performed to obtain a target 3D lookup table; The target 3D lookup table is used to perform color enhancement processing on the image to be color-enhanced, obtaining a target color-enhanced image.

2. The method according to claim 1, characterized in that, The first-branch feature extraction module of the preset color enhancement model extracts features from the image to be color-enhanced, obtaining a pixel intensity feature image, specifically including: The downsampling processing layer of the first branch feature extraction module is used to perform downsampling processing on the image to be color enhanced, and a first image with a predetermined sampling multiple is obtained; The feature refinement layer of the first combined feature extraction module is used to refine the features of the first image to obtain a first refined feature image; The adaptive sparse self-attention layer of the first combined feature extraction module is used to capture global context-aware information of the first refined feature image to obtain an initial pixel intensity feature image; Based on the initial pixel intensity feature image, the non-first combined feature extraction modules are successively used for image feature extraction processing to obtain the pixel intensity feature image.

3. The method according to claim 2, wherein The step of using the feature refinement layer of the first combined feature extraction module to refine the features of the first image to obtain a first refined feature image specifically includes: Performing normalization processing on the first image to obtain a first feature image; Based on the first feature image, a partial convolution network is used for key information enhancement processing to obtain a second feature image; Performing linear transformation processing on the second feature image to obtain a third feature image; Based on the third feature image, a depthwise separable convolution network is used for local feature extraction to obtain a fourth feature image; Based on the third feature image, a partial convolution network is used for key information enhancement processing to obtain a fifth feature image; Performing multiplication operation processing on the fourth feature image and the fifth feature image to obtain an initial refined feature image; Using a preset GELU activation function to optimize the initial refined feature image to obtain the first refined feature image.

4. The method according to claim 2, wherein The step of using the adaptive sparse self-attention layer of the first combined feature extraction module to capture global context-aware information of the first refined feature image to obtain an initial pixel intensity feature image specifically includes: Performing normalization processing on the first refined feature image to obtain a first pixel feature image; Performing projection processing on the first pixel feature image to obtain a query matrix, a key matrix, and a value matrix of the first pixel feature image; Based on the query matrix, the key matrix, a learnable bias matrix, and a predetermined scaling factor, a first activation function is used for calculation processing to obtain a dense self-attention score; Based on the query matrix, the key matrix, a learnable bias matrix, and a predetermined scaling factor, a second activation function is used for calculation processing to obtain a sparse self-attention score; Based on the value matrix, the dense self-attention score, the sparse self-attention score, a first learnable weight matrix, and a second learnable weight matrix, calculation processing is performed to obtain an attention matrix; Based on the attention matrix and the first refined feature image, addition operation processing is performed to obtain the initial pixel intensity feature image.

5. The method according to claim 1, characterized in that, The step of using the second branch feature extraction module of the preset color enhancement model to perform feature extraction on the image to be color enhanced to obtain a color feature image specifically includes: Using the feature refinement layer of the second branch feature extraction module to refine the features of the color feature image to obtain a second refined feature image; Normalize the second refined feature image to obtain the normalized second refined feature image; Use deep convolutional networks with convolution kernels of 3×3 and 5×5 respectively to perform local feature extraction on the second refined feature image to obtain a first color local feature image and a second color local feature image; Perform feature fusion on the first color local feature image and the second color local feature image to obtain a color fusion image; Use deep convolutional networks with convolution kernels of 3×3 and 5×5 respectively to perform local feature extraction on the color fusion image to obtain a third color local feature image and a fourth color local feature image; Perform image stitching processing based on the third color local feature image and the fourth color local feature image to obtain the color feature image.

6. An image color enhancement device based on lookup table estimation, characterized in that, It includes: A feature extraction module, configured to use the feature extraction module of a preset color enhancement model to perform feature extraction on the image to be color enhanced, to obtain a pixel intensity feature image and a color feature image of the image to be color enhanced; The feature extraction module includes a first branch feature extraction module and a second branch feature extraction module, and specifically is configured to use the first branch feature extraction module of a preset color enhancement model to perform feature extraction on the image to be color enhanced to obtain a pixel intensity feature image; Use the second branch feature extraction module of a preset color enhancement model to perform feature extraction on the image to be color enhanced to obtain a color feature image; wherein, the first branch feature extraction module includes a downsampling processing layer and multiple combined feature extraction modules, and each combined feature extraction module includes a feature refinement layer and an adaptive sparse self-attention layer for capturing global context information; the second branch feature extraction module includes a feature refinement layer and multiple mixed-scale convolutional layers for capturing local color information; A weight estimation module, configured to perform weight estimation on the pixel intensity feature image and the color feature image respectively by using the lookup table weight estimation module of the preset color enhancement model, so as to obtain a pixel intensity lookup table weight corresponding to the pixel intensity feature image and an image color lookup table weight corresponding to the color feature image; specifically, it is configured to use the global branch of the lookup table weight estimation module of the preset color enhancement model to perform feature extraction on the pixel intensity feature image and the color feature image respectively, so as to obtain a global pixel branch feature image corresponding to the pixel intensity feature image and a global color branch feature image corresponding to the color feature image; use the local branch of the lookup table weight estimation module of the preset color enhancement model to perform feature extraction on the pixel intensity feature image and the color feature image respectively, so as to obtain a local pixel branch feature image corresponding to the pixel intensity feature image and a local color branch feature image corresponding to the color feature image; perform feature fusion on the global pixel branch feature image and the local pixel branch feature image to obtain a pixel fusion feature image; perform feature fusion on the global color branch feature image and the local color branch feature image to obtain a color fusion feature image; perform non-linear transformation processing on the pixel fusion feature image and the color fusion feature image respectively to obtain a pixel intensity lookup table weight corresponding to the pixel intensity feature image and an image color lookup table weight corresponding to the color feature image; A calculation module, configured to perform calculation processing based on the pixel intensity lookup table weight, the image color lookup table weight, and a preset initial 3D lookup table to obtain a target 3D lookup table; A color enhancement module, configured to perform color enhancement processing on the image to be color enhanced by using the target 3D lookup table to obtain a target color enhanced image.

7. A storage medium, characterized in that, The storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the image color enhancement method based on lookup table estimation according to any one of claims 1-5 above are implemented.

8. An electronic device, characterized in that, It at least includes a memory and a processor, the memory stores a computer program, and when the processor executes the computer program on the memory, the steps of the image color enhancement method based on lookup table estimation according to any one of claims 1-5 above are implemented.

Images