Adaptive color management method and device based on scene recognition

Abstract

The invention relates to an adaptive color management method and device based on scene recognition, and the method comprises the steps: obtaining image data, and calling a deep learning model to process the image data, so as to extract key visual features in the image data; performing weighting and fusion processing on various key visual features through an attention mechanism model to generate a fusion feature vector; and classifying scene type mapping based on the fusion feature vector to obtain a scene classification result, and training a diffusion model based on the scene classification result and a color mapping table of multiple types of scenes to fit scene-color distribution features. And taking the scene type identification result of the real-time scene as the input of the trained diffusion model to generate a dynamic color mapping table, and adjusting the color mapping table based on the dynamic color mapping table to obtain an adjusted color mapping table. And applying the adjusted color mapping table to the image data to adjust the visual attribute of the image data.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and particularly to an adaptive color management method and device based on scene recognition. Background Art

[0002] With the rapid development of digital media technology, the quality and visual effects of image content have increasingly become key factors in user experience, especially in the fields of digital photography, film production, video games, and virtual reality.

[0003] In traditional color management systems, color adjustment mainly relies on pre-set Look-Up Tables (LUTs) or manual adjustment to achieve specific visual effects. Although this method can provide stable and predictable output to a certain extent, it generally lacks sufficient flexibility and is difficult to automatically adapt to diverse scene conditions. For example, under changing natural lighting or different indoor lighting conditions, these static methods cannot be adjusted immediately, making it difficult to ensure natural and accurate color representation.

[0004] With the progress of artificial intelligence and machine learning technologies, automated and intelligent image processing technologies have gradually become a research and application hotspot. In particular, using deep learning models for image scene recognition and color adjustment has shown superior performance compared to traditional methods. These technologies can dynamically adjust color strategies based on image content and context information, thus providing a more user-friendly and personalized visual experience in a changing visual environment. Even so, the demand for high-quality image processing technologies is increasing day by day, especially today when content creation and consumption are increasingly inclined towards diversification and personalization. Users and content producers need more intelligent and automated image processing tools to reduce production costs and time consumption while ensuring visual effects. In addition, with the popularization of high-definition content such as 4K and 8K, the requirements for color processing technologies have also increased, which further promotes the exploration and development of advanced image processing technologies. However, existing automatic color adjustment technologies often rely on a large amount of data input and complex parameter adjustment, which to a certain extent limits the flexibility and convenience of color adjustment. Secondly, many models are still difficult to meet high-standard commercial requirements in terms of real-time performance, especially in application scenarios such as real-time video processing and online content editing. In addition, these technologies often require high operation and maintenance costs, which is particularly obvious when processing large-scale or ultra-high-resolution images. Summary of the Invention

[0005] Based on this, it is necessary to provide an adaptive color management method and device based on scene recognition with high flexibility, high real-time performance, and low operation and maintenance costs for the above technical problems.

[0006] The present invention provides an adaptive color management method based on scene recognition, and the method includes:

[0007] Obtain image data, and call a deep learning model to process the image data to extract key visual features in the image data, where the key visual features include color distribution features, texture features, and edge features;

[0008] Perform weighted sum and fusion processing on various key visual features through an attention mechanism model to generate a fusion feature vector, where the fusion feature vector contains the scene information of the image data and is used for scene type mapping;

[0009] Classify the scene type mapping based on the fusion feature vector to obtain a scene classification result, and train a Diffusion model based on the scene classification result and a color mapping table corresponding to multiple types of scenes to fit the scene-color distribution characteristics;

[0010] Use the scene type recognition result of the real-time scene as the input of the trained diffusion model to generate a dynamic color mapping table, and adjust the color mapping table based on the dynamic color mapping table to obtain an adjusted color mapping table;

[0011] Apply the adjusted color mapping table to the image data to adjust the visual attributes of the image data.

[0012] In one embodiment, the obtaining of the image data and the calling of the deep learning model to process the image data to extract the key visual features in the image data includes:

[0013] Obtain the number of pixel points in the image data and the color value of each pixel point, and call the Kronecker function to count the occurrence frequency of pixel points with the same color value in the image data to obtain a color histogram for characterizing the color distribution feature;

[0014] Obtain the number of pixel pairs and the gray value of each pixel pair in the image data at any distance and direction, and generate the texture feature of the image data through a gray-level co-occurrence matrix according to the number of pixel pairs and the gray value of each pixel pair.

[0015] In one embodiment, the obtaining of the image data and the calling of the deep learning model to process the image data to extract the key visual features in the image data further includes:

[0016] Obtain the pixel intensity at any position in the image data, and calculate the gradient of the pixel intensity to determine the position of the object boundary in the image data to obtain the edge feature;

[0017] Obtain the weight coefficients of the color distribution feature, texture feature, and edge feature respectively, and calculate the integrated feature vector of the color distribution feature, texture feature, and edge feature based on the weight coefficients;

[0018] Among them, the weight coefficients are obtained by the YOLOv8 network through extraction and setting of the color distribution feature, texture feature, and edge feature, and are used to adjust the contribution ratio of different features in the integrated feature vector.

[0019] In one embodiment, the weighted sum and fusion processing of various key visual features by the attention mechanism model to generate a fusion feature vector includes:

[0020] Extract the feature vector from the similar pixel region in the image data, and call the attention mechanism model to calculate the weight assignment of the feature vector, where the weight assignment is used to adjust the contribution ratio of each feature vector in the fusion feature vector;

[0021] Train each feature vector and the corresponding weight assignment of the feature vector through label data, and perform weighted sum and fusion processing on all feature vectors in the image data to solidify the fusion feature vector matrix;

[0022] Among them, the similar pixel region is a region composed of pixel points with a similarity exceeding a first threshold after the image data is processed by an image segmentation model, and the feature vector includes the color distribution feature, texture feature, and edge feature of the similar pixel region.

[0023] In one embodiment, the classification of the scene type mapping based on the fusion feature vector to obtain a scene classification result, and the training of the Diffusion model based on the scene classification result and the color mapping table corresponding to multiple types of scenes to fit the scene-color distribution feature includes:

[0024] Take the fusion feature vector matrix as the input of the meta-learning model and the multi-class classifier to output the classification output probability for characterizing the first type of scene in the image data, and obtain the preset color distribution information and adjustment factor of the first type of scene according to the existing color mapping table under the first type of scene;

[0025] Train the Diffusion model to generate the dynamic color mapping table according to the classification output probability, preset color distribution information, adjustment factor, illumination intensity variable of the scene, and color mapping table label corresponding to each type of scene in the image data to obtain the trained diffusion model;

[0026] Among them, the first type of scenario is any scenario with a determined scenario type in the image data, the classification output probability is the probability that the first type of scenario is the determined scenario type, and the adjustment factor is used to adjust the contribution degree of the scenario classification result to the color mapping table.

[0027] In one embodiment, the method of using the scenario type recognition result of the real-time scenario as the input of the trained diffusion model to generate a dynamic color mapping table and adjusting the color mapping table based on the dynamic color mapping table to obtain an adjusted color mapping table includes:

[0028] Call the trained diffusion model to process the real-time data of the current scenario to generate the dynamic color mapping table, and adjust the color mapping table in combination with the learning rate parameter to obtain the adjusted color mapping table;

[0029] Among them, the learning rate parameter is used to control the speed and amplitude of the adjustment of the color mapping table.

[0030] In one embodiment, the method of applying the adjusted color mapping table to the image data to adjust the visual attributes of the image data includes:

[0031] Call a transformation function to calculate the adjusted color mapping table and the color distribution of the image data, so as to map the color distribution in the image data to the color space in the adjusted color mapping table, and obtain the image data with adjusted colors.

[0032] The present invention also provides an adaptive color management device based on scenario recognition, and the device includes:

[0033] A feature extraction module, configured to obtain image data and call a deep learning model to process the image data to extract key visual features in the image data, where the key visual features include color distribution features, texture features, and edge features;

[0034] A feature fusion module, configured to perform weighted sum and fusion processing on various key visual features through an attention mechanism model to generate a fusion feature vector, where the fusion feature vector contains the scenario information of the image data and is used for scenario type mapping;

[0035] A color mapping table generation module, configured to classify the scenario type mapping based on the fusion feature vector to obtain a scenario classification result, and train a Diffusion model based on the scenario classification result and color mapping tables corresponding to multiple types of scenarios to fit the scenario-color distribution features;

[0036] A color mapping table adjustment module, configured to use the scene type recognition result of the real-time scene as the input of a trained diffusion model to generate a dynamic color mapping table, and adjust the color mapping table based on the dynamic color mapping table to obtain an adjusted color mapping table;

[0037] An image color adjustment module, configured to apply the adjusted color mapping table to the image data to adjust the visual attributes of the image data.

[0038] The present invention also provides an electronic device, including a memory and a processor, where the memory stores a computer program, and when the processor executes the computer program, the method for adaptive color management based on scene recognition as described in any one of the above is implemented.

[0039] The present invention also provides a computer storage medium, storing a computer program, and when the computer program is executed by a processor, the method for adaptive color management based on scene recognition as described in any one of the above is implemented.

[0040] The present invention also provides a computer program product, including a computer program, and when the computer program is executed by a processor, the method for adaptive color management based on scene recognition as described in any one of the above is implemented.

[0041] The above method and device for adaptive color management based on scene recognition obtain an original image, and call a deep learning model to process the original image to extract key visual features in the image, including color distribution features, texture features, and edge features. Then, through an attention mechanism model, various key visual features are weighted and fused to generate a fused feature vector. The generated fused feature vector contains the scene information of the original image and can be used for scene type mapping. Subsequently, based on the fused feature vector, the scene type mapping is classified to obtain corresponding scene classification results. Then, the scene type recognition result of the real-time scene is used as the input of a trained diffusion model to generate a dynamic color mapping table, and the color mapping table is adjusted based on the dynamic color mapping table to obtain an adjusted color mapping table. Finally, the adjusted color mapping table is applied to the original image, and the visual attributes of the original image can be adjusted to obtain a new image with new visual attributes. By introducing scene recognition technology and a dynamic LUT adjustment mechanism, through real-time analysis of the visual features and scene data of the image, the method can dynamically generate and adjust the LUT, realize scene-adaptive color management, improve the accuracy and flexibility of color adjustment, and without relying on a large amount of data input and complex parameter adjustment, the operation and maintenance cost is low, greatly expanding the application scope and user group, and meeting the wider market demand. Description of the Drawings

[0042] To more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0043] Figure 1 One of the flow diagrams of the adaptive color management method based on scene recognition provided by the present invention;

[0044] Figure 2 The overall flow diagram of image color management of the adaptive color management method based on scene recognition in the specific embodiment provided by the present invention;

[0045] Figure 3 Another flow diagram of the adaptive color management method based on scene recognition provided by the present invention;

[0046] Figure 4 Another flow diagram of the adaptive color management method based on scene recognition provided by the present invention;

[0047] Figure 5 Another flow diagram of the adaptive color management method based on scene recognition provided by the present invention;

[0048] Figure 6 Another flow diagram of the adaptive color management method based on scene recognition provided by the present invention;

[0049] Figure 7 Another flow diagram of the adaptive color management method based on scene recognition provided by the present invention;

[0050] Figure 8 The structural diagram of the adaptive color management device based on scene recognition provided by the present invention;

[0051] Figure 9 The internal structure diagram of the electronic device provided by the present invention. Detailed Description of the Invention

[0052] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the 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.

[0053] The following will be combined with Figures 1 to 9Describe the adaptive color management method and device based on scene recognition of the present invention.

[0054] As Figure 1 shown, in one embodiment, an adaptive color management method based on scene recognition includes the following steps:

[0055] Step S110, obtain image data, and call a deep learning model to process the image data to extract key visual features in the image data, where the key visual features include color distribution features, texture features, and edge features.

[0056] Specifically, the server obtains the original image data and calls a deep learning model (such as UNet or YOLOv8) to process the image data to extract the color distribution features, texture features, and edge features in the image data, that is, the key visual features.

[0057] Combined with Figure 2 shown, in a specific embodiment, in the application of advanced image processing and scene understanding, effective visual feature extraction is a crucial step. In the process of extracting visual features, it usually involves using advanced deep learning models (such as UNet or YOLOv8) to process the input image data and extract the key visual features that describe the image from the image data. These features include color histograms, textures, and edge information, which can help the system better understand the image content and context.

[0058] Among them, the color histogram is a basic method for describing the color distribution of an image, and its algorithm formula is:

[0059] H(c) = 1 / N * sum(i = 1 to N) delta(c - c_i).

[0060] In the formula, H(c) is the color histogram, N is the total number of pixels in the image, c_i is the color value of the i-th pixel, and delta is the Kronecker function, which is used to count the frequency of the color c appearing in the image. This feature can help identify the dominant colors and the pattern of color distribution in the image, and is used for the classification and analysis of the scene.

[0061] Texture features are described by analyzing the spatial layout between pixels, and they reflect the structure and appearance of the image surface. Texture features can be calculated by the gray-level co-occurrence matrix (GLCM) or similar methods, and the algorithm formula is:

[0062] G(p,q) = 1 / M * sum(j = 1 to M) (I(p_j) - mean(I)) * (I(q_j) - mean(I)).

[0063] In the formula, G(p,q) is the gray-level co-occurrence matrix, M is the total number of pixel pairs considered at a specific distance and direction, I(p_j) and I(q_j) are the gray values of the pixel pairs respectively, and this metric reflects the regularity and complexity of the texture within the image.

[0064] Edge Features is a method used in image analysis to identify the boundaries of objects in an image, and its algorithm formula is:

[0065] S(x,y) = sqrt((I(x + 1,y) - I(x - 1,y))^2 + (I(x,y + 1) - I(x,y - 1))^2).

[0066] In the formula, S(x,y) is the edge feature, and I(x,y) represents the pixel intensity at position (x,y). This algorithm formula determines the edge position by calculating the gradient of the pixel intensity and is used to capture the shape and contour information in the image.

[0067] Feature integration, which integrates all the extracted feature vectors, can be expressed as:

[0068] F = w1 * H(c) + w2 * G(p,q) + w3 * S(x,y).

[0069] Among them, F is the integrated feature vector, and w1, w2, and w3 are weight coefficients used to adjust the contribution ratio of different features in the final feature vector. This weighted combination enables the system to adjust the importance of features according to the requirements of different scenarios to optimize performance.

[0070] In practical applications, it includes automatic image classification, scene understanding, and content-based image retrieval, etc. For example, in an autonomous driving system, this feature extraction technology can help the vehicle identify and understand its surrounding environment, such as road signs, traffic lights, and other vehicles, so as to make corresponding driving decisions. In medical imaging, it can help identify and classify different types of tissues and lesions, improving the accuracy and efficiency of diagnosis.

[0071] In this embodiment, YOLO (You Only Look Once) series models such as YOLOv5 or YOLOv8 can be used as deep learning models to extract key visual features in images. During the process of feature vector extraction, image preprocessing is first carried out. Before the image is input into the YOLO model, the image data first undergoes a series of preprocessing steps, including scaling to a fixed size (usually the size used during model training, such as 416x416 pixels), and normalization processing to ensure the unity of the input data and the efficiency of model processing. Subsequently, the YOLO model is used to extract features. The YOLO model performs a series of convolution, batch normalization, and activation operations on the input image through its convolutional layer structure, thereby extracting image features at multiple scales. These operations include using convolution kernels of different sizes to capture features from fine-grained to coarse-grained, enabling the model to understand the image content at both global and local scales. Among them, YOLOv5 and YOLOv8 particularly emphasize speed and accuracy during the feature extraction process, and improve the efficiency and effect of feature extraction through techniques such as depthwise separable convolution and residual connections.

[0072] In this embodiment, the feature maps obtained from the YOLO model can be directly used as the basis for scene classification, or further processed to generate more refined fused feature vectors. For example, through methods such as global average pooling or max pooling, the spatial information of each feature map can be compressed into a vector of a fixed length. These vectors are further weighted and fused through an attention mechanism model to obtain the final fused feature vector. The attention mechanism dynamically adjusts the contribution ratio of each feature vector by learning the importance of each feature to maximize the accuracy of scene recognition.

[0073] For example, in an autonomous driving system, the YOLO model can quickly identify objects such as vehicles, pedestrians, and road signs in images and provide the location and category information of these objects to help the driving system make decisions. In medical imaging, YOLO can help identify and classify different types of tissues and lesions in images, such as tumor detection, thereby providing more accurate diagnostic information. In this way, the YOLO model not only provides fast and accurate feature extraction capabilities, but also provides strong technical support for the dynamic color management system through the combination with the attention mechanism.

[0074] Step S120, perform weighted and fusion processing on various key visual features through an attention mechanism model to generate a fused feature vector, where the fused feature vector contains the scene information of the image data and is used for scene type mapping.

[0075] Specifically, the server performs weighted sum and feature fusion processing on various types of key visual features extracted in step S110 through a self-attention mechanism model to generate a comprehensive feature vector containing the scene information of the image data and used for scene type mapping, that is, a fusion feature vector.

[0076] As shown in Figure 2 In a specific embodiment, in the adaptive color management method based on scene recognition provided by the present invention, feature fusion is a key step in image processing and computer vision, which is used to combine multiple features extracted from different sources or algorithms in order to more comprehensively understand the image content. In many advanced applications, such as automatic scene recognition, face recognition, intelligent video surveillance, etc., the effectiveness of feature fusion directly affects the performance of the system and the accuracy of scene recognition.

[0077] In this embodiment, feature fusion can be achieved by weighted summation. Each type of feature is assigned a weight to reflect its importance for the final task (such as a classification task or an identification task). The algorithm formula is:

[0078] V = sum(k = 1 to K) alpha_k * F_k.

[0079] Where, V represents the fused feature vector, F_k is the k-th feature vector extracted from the image, and alpha_k is the weight calculated for this feature through the attention mechanism.

[0080] It should be noted that F_k represents the k-th feature vector, such as a color histogram, a texture descriptor, an edge feature, etc. These feature vectors are obtained by visual feature extraction. alpha_k is a weight value determined through the attention mechanism model and is used to adjust the contribution degree of each feature in the fusion vector. This weight not only reflects the importance of the feature but also can be dynamically adjusted according to the current task or dataset to optimize the performance.

[0081] Furthermore, it should be noted that the role of the attention mechanism model in feature fusion is to automatically adjust its weight according to the relevance and information volume of the features. This mechanism allows the model to focus on the information that is most useful for the current task, thereby improving the processing efficiency and the accuracy of the result. For example, when processing an image with a complex background, texture and edge features may be more informative than a color histogram, so they will be assigned higher weights.

[0082] In practical applications, feature fusion technology is widely used to improve the adaptability and robustness of the system to complex environments. For example, in the vision system of autonomous vehicles, different features extracted from the road environment (such as the edge features of road markings, the texture features of vehicles and pedestrians, etc.) need to be effectively fused to ensure correct scene understanding and decision support. Through feature fusion, the system can more accurately detect and identify different objects, such as pedestrians, vehicles, traffic signs, etc., and maintain a high recognition rate even under changing lighting conditions, occlusion, or poor weather conditions. In a security monitoring system, by fusing visual features from multiple cameras, the monitoring system can better identify and track moving objects in the scene and improve the ability to detect abnormal behaviors. For example, in the monitoring of airports or shopping malls, through feature fusion technology, the security monitoring system can more effectively identify security threats such as left luggage and abnormal crowd gatherings, and thus respond promptly to possible security incidents. Therefore, feature fusion not only enhances the comprehensive performance of image analysis but also improves the adaptability and accuracy in a changing environment, and is an indispensable part.

[0083] Step S130: Classify the scene type mapping based on the fused feature vector to obtain a scene classification result, and train the Diffusion model based on the scene classification result and the color mapping tables corresponding to multiple types of scenes to fit the scene-color distribution features.

[0084] Specifically, the server classifies the scene type mapping in the image data based on the fused feature vector obtained by feature fusion in step S120, and trains the Diffusion model based on the scene classification result and the color mapping tables corresponding to multiple types of scenes to fit the scene-color distribution features.

[0085] Combined with Figure 2 As shown, in a specific embodiment, for the adaptive color management method based on scene recognition provided by the present invention, the scene type mapping involves associating the extracted visual features with specific scene types and applying this information to adjust the visual representation of the image. This process is particularly important in applications such as content recognition, augmented reality, and automatic color adjustment.

[0086] In this embodiment, the scene type mapping can be achieved by combining multiple classifiers and a meta-learning model, and the algorithm formula can be expressed as:

[0087] L = beta * sum(r = 1 to R) M_r(V) * P_r.

[0088] In the formula, L represents the finally generated scene-specific look-up table (LUT), V is the fused feature vector, M_r is the classification model output corresponding to the r-th type of scene (usually probability), R is the total number of scene types, P_r is the pre-set color distribution scheme for the r-th type of scene, and beta is a regulation factor used to adjust the influence intensity of the model output on the final color distribution scheme.

[0089] It should be noted that M_r(V) is a function or model that receives the fused feature vector V as input and outputs a value representing the probability that the input image belongs to the r-th type of scene. P_r is the pre-defined color distribution scheme for a specific scene type, and its purpose is to optimize the image performance in that scene. These color distribution schemes can be pre-set based on art guidance, historical data, or prior knowledge. The regulation factor beta is used to adjust the contribution degree of the scene classification result to the final color mapping. By changing the value of beta, the conservativeness or radicalness of color adjustment can be controlled.

[0090] In this embodiment, the core purpose of scene type mapping is to ensure that the image is more visually consistent with its content and context. For example, a scene may have different visual requirements due to its lighting conditions, main colors, or included objects (such as a beach, forest, urban night view, etc.). By using color distribution schemes optimized for different scenes, the visual quality of the image can be improved and the user experience can be enhanced.

[0091] In practical applications, such as in film post-production, different scenes require different hue and brightness levels to convey the correct mood and time. Scene type mapping allows for automatic adjustment of these parameters to ensure that each scene can visually convey the user's intention. For example, a night scene may apply an LUT that makes the colors colder and the contrast higher, while a sunrise scene may use an LUT that enhances the warm tones and increases the brightness. In a smart home system, scene type mapping can be used to automatically adjust the color temperature and brightness of ambient light to adapt to the daylight conditions or household activities at different times. For example, when watching a movie at night, a soft lighting scheme can be selected, while when reading or working, a brighter lighting scheme closer to natural light can be chosen. In this way, scene type mapping not only improves the visual effects of images and the environment but also increases the dynamics and personalization of user interaction, making it more in line with people's lives and feelings.

[0092] Step S140: Use the scene type recognition result of the real-time scene as the input of the trained diffusion model to generate a dynamic color mapping table, and adjust the color mapping table based on the dynamic color mapping table to obtain an adjusted color mapping table.

[0093] Specifically, the server uses the scene type recognition result of the real-time scene as the input of the trained diffusion model, and then generates a real-time dynamic color mapping table, and adjusts the original color mapping table based on this dynamic color mapping table to obtain the adjusted color mapping table.

[0094] Combined with Figure 2 As shown, in a specific embodiment, the adaptive color management method based on scene recognition provided by the present invention, the DLUT (Dynamic Look-Up Table) mapping generation and adjustment aims to dynamically adjust the color mapping table according to real-time scene data to achieve optimal image output. This process not only considers the initial scene classification result, but also includes real-time response to environmental changes, thereby improving the adaptability and flexibility of image processing.

[0095] In this embodiment, the algorithm formula for the generation and adjustment of DLUT is:

[0096] LUT = L + gamma * (D'(C_s) - L).

[0097] Among them, LUT is the adjusted color mapping table, L is the LUT obtained based on the initial scene recognition, usually a pre-defined or color mapping table optimized for specific scenes through machine learning algorithms. D' is a Diffusion model (diffusion model) adjusted according to the real-time scene data C_s, which is used to dynamically generate color mappings according to the specific conditions of the current scene. C_s represents the real-time data of the current scene, which can include variables such as lighting conditions, color distribution, and main elements. Gamma is a learning rate parameter that controls the speed and amplitude of the LUT adjustment to ensure a smooth transition of the color mapping.

[0098] In this embodiment, the core of the DLUT mapping is to perform real-time optimization on the original basis, that is, by introducing the Diffusion model D', it can analyze the specific requirements of the current scene in real time and adjust the LUT accordingly. The advantage of this method lies in its dynamicity to cope with real-time changes such as lighting changes and scene switching. The Diffusion model proposes new color mapping suggestions by analyzing the real-time data C_s, and controls the application speed and degree of these suggestions through the learning rate parameter gamma, so that the LUT can smoothly adapt to scene changes.

[0099] In practical applications, such as dynamic video editing and live broadcasting, the dynamic generation and adjustment of DLUT mapping are crucial. For example, in the live broadcast of a sports event, the lighting conditions may change due to the variation of time (day and night) or weather (sunny or rainy). By using DLUT, the color performance of the video can be adjusted in real time to ensure that the audience always obtains the best visual experience. Additionally, in the field of security monitoring, the monitoring system needs to operate under different ambient light conditions, from dawn to dusk, from indoor to outdoor. DLUT allows the system to adjust the color and contrast of the image in real time according to the current lighting conditions and scene content, improving the clarity and recognizability of the image. Moreover, in mobile photography, DLUT also plays an important role. The camera application of a smartphone can automatically adjust the color settings according to the scene (such as portrait, landscape, night scene) by using a technology similar to DLUT, enabling users to take relatively high-quality photos even in the hands of non-professional users. Therefore, DLUT not only improves the visual quality of the image but also enhances the practicality and flexibility in various application environments, enabling users to obtain the best visual experience in different visual scenarios.

[0100] Step S150, apply the adjusted color mapping table to the image data to adjust the visual attributes of the image data.

[0101] Specifically, the server calls a transformation function to calculate the adjusted color mapping table and the color distribution of the image data, so as to map the color distribution in the image data to the color space in the adjusted color mapping table, and finally obtain the image data with adjusted colors.

[0102] Combined Figure 2 As shown, in a specific embodiment, for the adaptive color management method based on scene recognition provided by the present invention, in the final stage of image processing, color application and output are crucial steps to ensure that the final image meets the expected visual effect. This step involves applying the previously calculated Look-Up Table (LUT) to the original image to adjust the color and other visual attributes of the original image, thereby achieving a specific visual effect.

[0103] In this embodiment, the algorithm formula for color application is:

[0104] I_f = I_o + mu * (T(LUT, I_o) - I_o).

[0105] Among them, \(I_o\) is the original image, that is, the image before color adjustment. \(T(LUT, I_o)\) is the transformation function that applies the LUT. This transformation function takes the LUT and the original image \(I_o\) as inputs and outputs the adjusted image, adjusting the image color by mapping the colors in the original image to a new color space. \(\mu\) is the application intensity, which is a coefficient between 0 and 1 used to control the degree of color adjustment. The larger the value of \(\mu\), the more obvious the color change. \(I_f\) is the final image, that is, the output of the image after color adjustment.

[0106] In this embodiment, by applying the LUT, each color value of the original image \(I_o\) is mapped to a new color value according to the instructions in the LUT. This process not only involves the adjustment of hue and saturation, but also includes the change of brightness and contrast. The role of the transformation function \(T\) is to ensure that these mappings are correctly implemented to reflect the desired visual style or to cope with specific ambient light conditions.

[0107] It should be noted that the generation of the DLUT is not simply to adjust the original LUT, but to create a new DLUT entirely based on the data obtained from the scene classification results. This is driven by the type recognition results of the real-time scene, which are obtained using a trained diffusion model. This model utilizes the correlation data between scene classification and the color mapping table and can predict and generate a new color mapping table that is most suitable for the current scene. Compared with the existing color management methods, the advantage of this method is that each generated DLUT is optimized for the current specific scene, ensuring the maximum relevance and effect of color adjustment. Additionally, the existing color management methods may need to modify or adjust the existing LUT to adapt to the new scene, which may not be efficient or accurate enough in a dynamic or changing environment. In this embodiment, through the real-time generated DLUT, it is directly optimized according to the current scene information, making each color adjustment in the most ideal state. The key lies in that the generation of the DLUT is based on the real-time data analysis of deep learning and scene recognition technologies, which can more flexibly cope with environmental changes and does not rely on the previous color configuration.

[0108] In this embodiment, during the process of applying the DLUT to the original image (\(I_o\)), the transformation function \(T(LUT, I_o)\) used not only simply maps colors, but comprehensively adjusts the color, brightness, contrast, and saturation of the image according to the guidance provided by the DLUT. The parameter \(\mu\) controls the intensity of this adjustment, ensuring that the image can be optimized according to the needs of the specific application scenario. The finally output image (\(I_f\)) can thus show a visual effect that is more suitable for the current scene while maintaining the original details and texture.

[0109] Furthermore, it should be noted that the introduction of the parameter mu allows for fine-tuning of the impact of color adjustment. In some applications, such as art photography or advertising design, more radical color changes may be required to convey specific visual effects or emotions. In other cases, such as medical imaging or security monitoring, more conservative adjustments may be needed to maintain the naturalness and accuracy of the image.

[0110] For example:

[0111] Film and video production: During the post-production process, LUTs are used to adjust the color of films to convey a specific time period or enhance a certain mood. For example, by using warm tones to enhance the atmosphere of a romantic scene, or cold tones to emphasize the severity of a scene.

[0112] Medical imaging: In medical imaging, correct color mapping can help doctors better identify lesion areas. For example, in thermal imaging, different colors can be used to represent areas of different temperatures, helping doctors diagnose inflammation or other diseases.

[0113] Real-time video monitoring: In the field of security monitoring, color adjustment can be used to enhance video images in nighttime or low-light conditions, improving visibility and recognition accuracy.

[0114] The above-mentioned adaptive color management method based on scene recognition obtains the original image and invokes a deep learning model to process the original image to extract key visual features in the image, including color distribution features, texture features, and edge features. Then, through the attention mechanism model, various key visual features are weighted and fused to generate a fused feature vector. The generated fused feature vector contains the scene information of the original image and can be used for scene type mapping. Subsequently, based on the fused feature vector, the scene type mapping is classified to obtain the corresponding scene classification result, and a color mapping table for the corresponding scene of the image is generated based on this scene classification result. Then, the scene type recognition result of the real-time scene is used as the input of the trained diffusion model to generate a dynamic color mapping table, and the color mapping table is adjusted based on this dynamic color mapping table to obtain an adjusted color mapping table. Finally, the adjusted color mapping table is applied to the original image to adjust the visual attributes of the original image and obtain a new image with new visual attributes. This method introduces scene recognition technology and a dynamic LUT adjustment mechanism. By real-time analyzing the visual features and scene data of the image, it can dynamically generate and adjust the LUT, achieve scene-adaptive color management, improve the accuracy and flexibility of color adjustment, and does not require relying on a large amount of data input and complex parameter adjustment. The operation and maintenance cost is relatively low, greatly expanding the application scope and user group, and meeting a wider market demand.

[0115] Such as Figure 3As shown, in one embodiment, the adaptive color management method based on scene recognition provided by the present invention acquires image data and invokes a deep learning model to process the image data to extract key visual features in the image data, specifically including the following steps:

[0116] Step S112: Acquire the number of pixel points in the image data and the color value of each pixel point, and call the Kronecker function to count the occurrence frequency of pixel points with the same color value in the image data to obtain a color histogram for characterizing the color distribution feature.

[0117] Step S114: Acquire the number of pixel pairs in the image data in any distance and direction and the gray value of each pixel pair, and generate the texture feature of the image data according to the number of pixel pairs and the gray value of each pixel pair through a gray-level co-occurrence matrix.

[0118] As Figure 4 As shown, in one embodiment, the adaptive color management method based on scene recognition provided by the present invention acquires image data and invokes a deep learning model to process the image data to extract key visual features in the image data, and specifically further includes the following steps:

[0119] Step S116: Acquire the pixel intensity at any position in the image data and calculate the gradient of the pixel intensity to determine the position of the object boundary in the image data to obtain an edge feature.

[0120] Step S118: Respectively acquire the weight coefficients of the color distribution feature, the texture feature, and the edge feature, and calculate the integrated feature vector of the color distribution feature, the texture feature, and the edge feature based on the weight coefficients.

[0121] Among them, the weight coefficients are obtained by the YOLOv8 network through extraction and setting of the color distribution feature, the texture feature, and the edge feature, and are used to adjust the contribution ratio of different features in the integrated feature vector.

[0122] As Figure 5 As shown, in one embodiment, the adaptive color management method based on scene recognition provided by the present invention performs weighted sum and fusion processing on various key visual features through an attention mechanism model to generate a fusion feature vector, specifically including the following steps:

[0123] Step S122: Extract feature vectors from similar pixel regions in the image data and call the attention mechanism model to calculate the weight assignment of the feature vectors, and the weight assignment is used to adjust the contribution ratio of each feature vector in the fusion feature vector.

[0124] Step S124: Each feature vector and its corresponding weight assignment are obtained through training with label data, and all the feature vectors in the image data are processed by weighted sum fusion to solidify the fused feature vector matrix.

[0125] Among them, the similar pixel region is the region composed of pixel points with a similarity exceeding the first threshold after the image data is processed by the image segmentation model. The feature vectors include the color distribution feature, texture feature, and edge feature of the similar pixel region.

[0126] As Figure 6 shown, in one embodiment, the adaptive color management method based on scene recognition provided by the present invention classifies the scene type mapping based on the fused feature vector to obtain the scene classification result, and trains the Diffusion model based on the scene classification result and the color mapping tables corresponding to multiple types of scenes to fit the scene-color distribution feature. Specifically, it includes the following steps:

[0127] Step S132: The fused feature vector matrix is used as the input of the meta-learning model and the multi-class classifier to output the classification output probability for characterizing the first type of scene in the image data, and according to the existing color mapping table under the first type of scene, the preset color distribution information and the adjustment factor of the first type of scene are obtained.

[0128] Step S134: According to the classification output probability corresponding to each type of scene in the image data, the preset color distribution information, the adjustment factor, the illumination intensity variable of the scene, and the color mapping table label, the Diffusion model is trained to generate a dynamic color mapping table, and the trained diffusion model is obtained.

[0129] Among them, the first type of scene is any determined scene type in the image data, the classification output probability is the probability that the first type of scene is the determined scene type, and the adjustment factor is used to adjust the contribution degree of the scene classification result to the color mapping table.

[0130] As Figure 7 shown, in one embodiment, the adaptive color management method based on scene recognition provided by the present invention uses the scene type recognition result of the real-time scene as the input of the trained diffusion model to generate a dynamic color mapping table, and adjusts the color mapping table based on the dynamic color mapping table to obtain the adjusted color mapping table. Specifically, it includes the following steps:

[0131] Step S142: Call the trained diffusion model to process the real-time data of the current scene to generate a dynamic color mapping table.

[0132] Step S144: Combine the learning rate parameter to adjust the color mapping table to obtain the adjusted color mapping table.

[0133] Among them, the learning rate parameter is used to control the speed and amplitude of the color mapping table adjustment.

[0134] The following describes the adaptive color management device based on scene recognition provided by the present invention. The adaptive color management device based on scene recognition described below can be correspondingly referred to the adaptive color management method based on scene recognition described above.

[0135] As Figure 8 shown, in one embodiment, an adaptive color management device based on scene recognition includes a feature extraction module 810, a feature fusion module 820, a color mapping table generation module 830, a color mapping table adjustment module 840, and an image color adjustment module 850.

[0136] The feature extraction module 810 is used to obtain image data and call a deep learning model to process the image data to extract key visual features in the image data. The key visual features include color distribution features, texture features, and edge features.

[0137] The feature fusion module 820 is used to perform weighted sum and fusion processing on various key visual features through an attention mechanism model to generate a fusion feature vector. The fusion feature vector contains the scene information of the image data and is used for scene type mapping.

[0138] The color mapping table generation module 830 is used to classify the scene type mapping based on the fusion feature vector to obtain a scene classification result, and train the Diffusion model based on the scene classification result and the color mapping tables corresponding to multiple types of scenes to fit the scene-color distribution characteristics.

[0139] The color mapping table adjustment module 840 is used to use the scene type recognition result of the real-time scene as the input of the trained diffusion model to generate a dynamic color mapping table, and adjust the color mapping table based on the dynamic color mapping table to obtain an adjusted color mapping table.

[0140] The image color adjustment module 850 is used to apply the adjusted color mapping table to the image data to adjust the visual attributes of the image data.

[0141] In this embodiment, for the adaptive color management device based on scene recognition provided by the present invention, the feature extraction module is specifically used for:

[0142] Obtain the number of pixel points in the image data and the color value of each pixel point, and call the Kronecker function to count the occurrence frequency of pixel points with the same color value in the image data to obtain a color histogram for characterizing the color distribution feature.

[0143] Obtain the number of pixel pairs in any distance and direction of the image data and the gray value of each pixel pair, and generate the texture feature of the image data according to the number of pixel pairs and the gray value of each pixel pair through the gray-level co-occurrence matrix.

[0144] In this embodiment, for the adaptive color management device based on scene recognition provided by the present invention, the feature extraction module is further specifically configured to:

[0145] Obtain the pixel intensity at any position in the image data, and calculate the gradient of the pixel intensity to determine the position of the object boundary in the image data, so as to obtain the edge feature.

[0146] Respectively obtain the weight coefficients of the color distribution feature, texture feature and edge feature, and calculate the integrated feature vector of the color distribution feature, texture feature and edge feature based on the weight coefficients.

[0147] Among them, the weight coefficients are obtained by the YOLOv8 network through extraction and setting of the color distribution feature, texture feature and edge feature, and are used to adjust the contribution ratio of different features in the integrated feature vector.

[0148] In this embodiment, for the adaptive color management device based on scene recognition provided by the present invention, the feature fusion module is specifically configured to:

[0149] Extract the feature vector from the similar pixel region in the image data, and call the attention mechanism model to calculate the weight assignment of the feature vector. The weight assignment is used to adjust the contribution ratio of each feature vector in the fusion feature vector.

[0150] Train each feature vector and the corresponding weight assignment of the feature vector through the label data, and perform weighted sum fusion processing on all feature vectors in the image data to solidify the fusion feature vector matrix.

[0151] Among them, the similar pixel region is a region composed of pixel points with a similarity exceeding the first threshold after the image data is processed by the image segmentation model. The feature vector includes the color distribution feature, texture feature and edge feature of the similar pixel region.

[0152] In this embodiment, for the adaptive color management device based on scene recognition provided by the present invention, the color mapping table generation module is specifically configured to:

[0153] Take the fusion feature vector matrix as the input of the meta-learning model and the multi-class classifier to output the classification output probability for characterizing the first type of scene in the image data, and obtain the preset color distribution information and adjustment factor of the first type of scene according to the existing color mapping table under the first type of scene.

[0154] Based on the classification output probability corresponding to each type of scene in the image data, the preset color distribution information, the adjustment factor, the illumination intensity variable of the scene, and the color mapping table label, train the Diffusion model to generate a dynamic color mapping table, and obtain a trained diffusion model.

[0155] Among them, the first type of scene is any determined scene type in the image data, the classification output probability is the probability that the first type of scene is the determined scene type, and the adjustment factor is used to adjust the contribution degree of the scene classification result to the color mapping table.

[0156] In this embodiment, for the adaptive color management device based on scene recognition provided by the present invention, the color mapping table adjustment module is specifically used for:

[0157] Call the trained diffusion model to process the real-time data of the current scene to generate a dynamic color mapping table.

[0158] Combine the learning rate parameter to adjust the color mapping table to obtain an adjusted color mapping table.

[0159] Among them, the learning rate parameter is used to control the speed and amplitude of the color mapping table adjustment.

[0160] In this embodiment, for the adaptive color management device based on scene recognition provided by the present invention, the image color adjustment module is specifically used for:

[0161] Call the transformation function to calculate the adjusted color mapping table and the color distribution of the image data, so as to map the color distribution in the image data to the color space in the adjusted color mapping table to obtain color-adjusted image data.

[0162] Figure 9 Illustrates a schematic diagram of the physical structure of an electronic device. The electronic device may be a smart terminal, and its internal structure diagram may be as Figure 9 shown. The electronic device includes a processor, a memory, and a network interface 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 terminal through a network connection. When the computer program is executed by the processor, it realizes an adaptive color management method based on scene recognition, and the method includes:

[0163] Obtain image data, and call a deep learning model to process the image data to extract key visual features in the image data. The key visual features include color distribution features, texture features, and edge features;

[0164] The weighted sum and fusion processing of various key visual features are performed through an attention mechanism model to generate a fused feature vector, which contains the scene information of the image data and is used for scene type mapping;

[0165] Based on the fused feature vector, the scene type mapping is classified to obtain a scene classification result, and the Diffusion model is trained based on the scene classification result and the color mapping tables corresponding to multiple types of scenes to fit the scene-color distribution characteristics;

[0166] The scene type recognition result of the real-time scene is used as the input of the trained diffusion model to generate a dynamic color mapping table, and the color mapping table is adjusted based on the dynamic color mapping table to obtain an adjusted color mapping table;

[0167] The adjusted color mapping table is applied to the image data to adjust the visual attributes of the image data.

[0168] Those skilled in the art can understand that Figure 9 the structure shown in

[0169] is only a block diagram of some structures related to the solution of the present invention, and does not constitute a limitation on the electronic device to which the solution of the present invention is applied. The specific electronic device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0170] On the other hand, the present invention also provides a computer storage medium storing a computer program, which when executed by a processor implements an adaptive color management method based on scene recognition. The method includes:

[0171] The image data is obtained, and a deep learning model is called to process the image data to extract the key visual features in the image data. The key visual features include color distribution features, texture features, and edge features;

[0172] The weighted sum and fusion processing of various key visual features are performed through an attention mechanism model to generate a fused feature vector, which contains the scene information of the image data and is used for scene type mapping;

[0173] Based on the fused feature vector, the scene type mapping is classified to obtain a scene classification result, and the Diffusion model is trained based on the scene classification result and the color mapping tables corresponding to multiple types of scenes to fit the scene-color distribution characteristics;

[0174] Apply the adjusted color mapping table to the image data to adjust the visual attributes of the image data.

[0175] In another aspect, a computer program product or a computer program is provided. The computer program product or the computer program includes computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium. When the processor executes the computer instructions, an adaptive color management method based on scene recognition is implemented. The method includes:

[0176] Obtain image data and call a deep learning model to process the image data to extract key visual features in the image data. The key visual features include color distribution features, texture features, and edge features;

[0177] Perform weighted sum and fusion processing on various key visual features through an attention mechanism model to generate a fusion feature vector. The fusion feature vector contains scene information of the image data and is used for scene type mapping;

[0178] Classify the scene type mapping based on the fusion feature vector to obtain a scene classification result, and train a Diffusion model based on the scene classification result and color mapping tables corresponding to multiple types of scenes to fit the scene-color distribution features;

[0179] Use the scene type recognition result of the real-time scene as the input of the trained diffusion model to generate a dynamic color mapping table, and adjust the color mapping table based on the dynamic color mapping table to obtain an adjusted color mapping table;

[0180] Apply the adjusted color mapping table to the image data to adjust the visual attributes of the image data.

[0181] 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 by the present invention 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.

[0182] By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), Synchlink DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct Rambus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), among others.

[0183] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0184] The above-described embodiments merely represent several implementation manners of the present invention. Their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the patent for the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent for the present invention shall be subject to the appended claims.

Claims

1. An adaptive color management method based on scene recognition, characterized in that The method includes: Obtain image data, and call a deep learning model to process the image data to extract key visual features in the image data, where the key visual features include color distribution features, texture features, and edge features; Perform weighted sum and fusion processing on various key visual features through an attention mechanism model to generate a fused feature vector, where the fused feature vector contains the scene information of the image data and is used for scene type mapping; Classify the scene type mapping based on the fused feature vector to obtain a scene classification result, and train a Diffusion model based on the scene classification result and a color mapping table corresponding to multiple type scenes to fit the scene-color distribution features; Use the scene type recognition result of the real-time scene as the input of the trained diffusion model to generate a dynamic color mapping table, and adjust the color mapping table based on the dynamic color mapping table to obtain an adjusted color mapping table; Apply the adjusted color mapping table to the image data to adjust the visual attributes of the image data.

2. The adaptive color management method based on scene recognition according to claim 1, characterized in that The obtaining of the image data and calling a deep learning model to process the image data to extract key visual features in the image data includes: Obtain the number of pixel points in the image data and the color value of each pixel point, and call the Kronecker function to count the occurrence frequency of pixel points with the same color value in the image data to obtain a color histogram for characterizing the color distribution features; Obtain the number of pixel pairs in the image data at any distance and direction and the gray value of each pixel pair, and generate the texture features of the image data through a gray-level co-occurrence matrix according to the number of pixel pairs and the gray value of each pixel pair.

3. The adaptive color management method based on scene recognition according to claim 2, characterized in that The obtaining of the image data and calling a deep learning model to process the image data to extract key visual features in the image data further includes: Obtain the pixel intensity at any position in the image data, and calculate the gradient of the pixel intensity to determine the position of the object boundary in the image data to obtain the edge features; Respectively obtain the weight coefficients of the color distribution features, texture features, and edge features, and calculate the integrated feature vector of the color distribution features, texture features, and edge features based on the weight coefficients; Wherein, the weight coefficients are obtained by the YOLOv8 network through extraction and setting of the color distribution features, texture features, and edge features, and are used to adjust the contribution ratio of different features in the integrated feature vector.

4. The adaptive color management method based on scene recognition according to claim 3, wherein, The performing of weighted sum and fusion processing on various key visual features through an attention mechanism model to generate a fused feature vector includes: Extract feature vectors from similar pixel regions in the image data, and call the attention mechanism model to calculate the weight assignment of the feature vectors, where the weight assignment is used to adjust the contribution ratio of each feature vector in the fused feature vector; Each of the feature vectors and the weight assignment corresponding to the feature vector are obtained through training with label data, and weighted sum fusion processing is performed on all the feature vectors in the image data to solidify the fused feature vector matrix; Among them, the similar pixel region is a region composed of pixel points with a similarity exceeding a first threshold after the image data is processed by an image segmentation model, and the feature vector includes the color distribution feature, texture feature, and edge feature of the similar pixel region.

5. The adaptive color management method based on scene recognition according to claim 4, wherein Classifying the scene type mapping based on the fused feature vector to obtain a scene classification result, and training the Diffusion model based on the scene classification result and the color mapping tables corresponding to multiple types of scenes to fit the scene-color distribution feature, including: Taking the fused feature vector matrix as the input of a meta-learning model and a multi-class classifier to output a classification output probability for characterizing the first type of scene in the image data, and obtaining the preset color distribution information and adjustment factor of the first type of scene according to the existing color mapping table under the first type of scene; Training the Diffusion model to generate the dynamic color mapping table according to the classification output probability, preset color distribution information, adjustment factor, scene illumination intensity variable, and color mapping table label corresponding to each type of scene in the image data to obtain a trained diffusion model; Among them, the first type of scene is any determined scene type in the image data, the classification output probability is the probability that the first type of scene is the determined scene type, and the adjustment factor is used to adjust the contribution degree of the scene classification result to the color mapping table.

6. The adaptive color management method based on scene recognition according to claim 5, characterized in that Using the scene type recognition result of the real-time scene as the input of the trained diffusion model to generate a dynamic color mapping table, and adjusting the color mapping table based on the dynamic color mapping table to obtain an adjusted color mapping table, including: Invoking the trained diffusion model to process the real-time data of the current scene to generate the dynamic color mapping table, and adjusting the color mapping table in combination with the learning rate parameter to obtain the adjusted color mapping table; Among them, the learning rate parameter is used to control the speed and amplitude of the color mapping table adjustment.

7. The adaptive color management method based on scene recognition according to claim 6, wherein Applying the adjusted color mapping table to the image data to adjust the visual attributes of the image data, including: Invoking a transformation function to calculate the adjusted color mapping table and the color distribution of the image data to map the color distribution in the image data to the color space in the adjusted color mapping table to obtain color-adjusted image data.

8. An adaptive color management device based on scene recognition, characterized in that The device includes: A feature extraction module, configured to obtain image data and call a deep learning model to process the image data to extract key visual features in the image data, where the key visual features include color distribution features, texture features, and edge features; A feature fusion module, which is used to perform weighted sum and fusion processing on various types of the key visual features through an attention mechanism model to generate a fused feature vector. The fused feature vector contains the scene information of the image data and is used for scene type mapping; A color mapping table generation module, which is used to classify the scene type mapping based on the fused feature vector to obtain a scene classification result, and train a Diffusion model based on the scene classification result and color mapping tables corresponding to multiple types of scenes to fit the scene-color distribution characteristics; A color mapping table adjustment module, which is used to use the scene type recognition result of the real-time scene as the input of the trained diffusion model to generate a dynamic color mapping table, and adjust the color mapping table based on the dynamic color mapping table to obtain an adjusted color mapping table; An image color adjustment module, which is used to apply the adjusted color mapping table to the image data to adjust the visual attributes of the image data.

9. An electronic device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer storage medium stores a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the method according to any one of claims 1 to 7.

Images