Screen color correction method and electronic device

Abstract

The application provides a screen color correction method and an electronic device. The method comprises the following steps: obtaining a panel type, running state information and color information of a screen to be corrected; constructing screen feature information of the screen to be corrected based on the panel type, the running state information and the color information; inputting the screen feature information into a color correction model to obtain a color correction parameter output by the color correction model; wherein the color correction model processes the screen feature information based on a decision tree of a corresponding branch selected based on the panel type to obtain the color correction parameter; and performing color correction on the screen to be corrected based on the color correction parameter. The application automatically generates a color correction parameter by using a color correction model based on a decision tree of a corresponding branch selected based on a panel type, realizes adaptive color correction for different panel types and running states, avoids the problem of low color correction accuracy caused by the dependence of a traditional scheme on manual experience, and improves color correction accuracy.

Description

Technical Field

[0001] This application relates to the field of screen calibration technology, and in particular to a screen color calibration method and electronic device. Background Technology

[0002] Color accuracy is one of the core indicators for measuring the color performance of a display device, directly affecting the accuracy of visual experience (such as watching movies and reading) and professional work (such as post-production photography and print design). Color accuracy is usually measured by color difference values; the smaller the color difference value, the closer the displayed colors are to their intended state, and the better the color accuracy. However, due to differences in screen hardware and natural aging during use, the actual displayed colors often deviate from their intended state, leading to increased color difference values. Therefore, screen calibration is necessary to reduce color difference values. Currently, screen calibration usually relies on the experience of engineers, but the human eye has limited ability to distinguish color differences, making it difficult to accurately judge the degree of color difference, resulting in inaccurate calibration and the screen's color accuracy failing to reach the ideal level after calibration. Summary of the Invention

[0003] This application provides a screen color calibration method and an electronic device to at least solve the above-mentioned technical problems existing in the prior art.

[0004] A first aspect of this application provides a screen color calibration method, the method comprising: Obtain the panel type, operating status information, and color information of the screen to be calibrated; Based on the panel type, the operating status information, and the color information, the screen feature information of the screen to be calibrated is constructed; The screen feature information is input into the color calibration model to obtain the color calibration parameters output by the color calibration model; wherein, the color calibration model processes the screen feature information based on the decision tree that selects the corresponding branch of the panel type to obtain the color calibration parameters; The screen to be calibrated is calibrated based on the calibration parameters.

[0005] In one possible implementation, the color calibration model is trained through the following operations: Obtain a training sample set, which includes multiple training samples, each of which includes a sample screen, screen feature information of the sample screen, and real color correction parameters; Using panel type as the splitting feature of the root node, branches corresponding to each panel type are split out, and the training sample set is divided into branches corresponding to each panel type according to the panel type. Within each branch corresponding to each panel type, based on the set feature importance weights, the sample weights of each training sample corresponding to each branch, and the set decision tree hyperparameters corresponding to each branch, the decision tree construction is performed iteratively with the goal of minimizing the multi-objective hierarchical loss function, resulting in multiple decision trees corresponding to each branch. All decision trees constructed from each branch are combined into the color correction model.

[0006] In one possible implementation, the screen feature information includes the following features: panel type, screen model shown in the operating status information, screen temperature shown in the operating status information, and ambient light intensity shown in the operating status information; The set feature importance weight is the feature weight assigned to each feature in the screen feature information; wherein, the feature weight of each feature is used to characterize the importance of each feature to color calibration.

[0007] In one possible implementation, during the iterative process of constructing a decision tree, the method further includes, for each previously constructed decision tree: In response to the completion of the construction of the previous decision trees, all the previously constructed decision trees will be used as an ensemble model to infer on the training sample set to obtain the predicted color correction parameters for each training sample. Based on the error between the predicted color correction parameters and the actual color correction parameters of each training sample, the sample weights of the training samples are updated, and the updated sample weights are used to construct a decision tree later.

[0008] In one possible implementation, the method further includes: Obtain the verification sample set; During the process of constructing the decision tree in a loop, the function value of the multi-objective hierarchical loss function is obtained based on the validation sample set; If the change in the function value for a consecutive first set number of rounds is less than a set change threshold, the learning rate of the decision tree will be halved. In response to the function value changing less than the set change threshold for a second consecutive set number of rounds, the construction of the decision tree is stopped, and all decision trees are combined into the color correction model, wherein the second set number of rounds is greater than the first set number of rounds.

[0009] In one possible implementation, combining all decision trees constructed from each branch into the color correction model includes: Obtain the contribution of each decision tree on the validation sample set, remove decision trees whose contribution is lower than the set score threshold, and combine the remaining decision trees into the color correction model.

[0010] In one possible implementation, the actual color calibration parameters include red channel gain parameters, green channel gain parameters, blue channel gain parameters, gamma offset, and color temperature compensation parameters; The multi-objective hierarchical loss function is related to the prediction error of each parameter in the true color correction parameters and the smoothness loss term.

[0011] In one possible implementation, the screen feature information is further constructed based on a time decay factor; the time decay factor is obtained through the following operation: Obtain the usage time of the screen to be calibrated and the attenuation coefficient corresponding to the panel type; The time decay factor is determined based on the usage duration and the decay coefficient.

[0012] In one possible implementation, the panel type includes organic light-emitting diodes, sub-millimeter light-emitting diodes, and liquid crystal displays; Obtain the color information of the screen to be calibrated, including: In response to the panel type being organic light-emitting diode (OLED), a first predetermined number of images are selected from the basic image library and the OLED-specific image library; In response to the panel type being sub-millimeter light-emitting diode, a second set number of images are selected from the basic image library and the sub-millimeter light-emitting diode dedicated image library; In response to the panel type being a liquid crystal display, a third predetermined number of images are selected from the basic image library; In response to the ambient light intensity shown in the operating status information being greater than the set strong light threshold, a fourth set number of images are selected from the strong light adaptation image library; After merging and deduplicating all selected images, a color calibration image set is obtained. Each frame in the color calibration frame set is displayed sequentially, and the color parameter information of each frame is collected as the color information; The number of images in the color calibration image set does not exceed a set threshold.

[0013] In one possible implementation, the method further includes: Collect the color information of the screen to be calibrated after color calibration, compare it with the standard color information, and determine the color difference value; In response to the color difference value being greater than the set color difference threshold, the screen feature information is updated based on the newly acquired color information, and the color calibration parameters are re-determined.

[0014] A second aspect of this application provides a screen color calibration apparatus, the apparatus comprising: The acquisition module is used to acquire the panel type, operating status information, and color information of the screen to be calibrated. The construction module is used to construct the screen feature information of the screen to be calibrated based on the panel type, the running status information and the color information; The parameter determination module is used to input the screen feature information into the color calibration model to obtain the color calibration parameters output by the color calibration model; wherein, the color calibration model processes the screen feature information based on the decision tree that selects the corresponding branch of the panel type to obtain the color calibration parameters; The color calibration module is used to calibrate the screen to be calibrated based on the color calibration parameters.

[0015] A third aspect of this application provides an electronic device comprising: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method described in this application.

[0016] A fourth aspect of this application provides a non-transitory computer-readable storage medium storing computer instructions for causing the computer to perform the methods described in this application.

[0017] The screen calibration method and electronic device of this application acquire the panel type, operating status information, and color information of the screen to be calibrated, and construct unified screen feature information based on the above information. Then, it uses a pre-trained calibration model with a panel type-based specialized architecture to automatically infer the calibration parameters adapted to the current screen, and finally writes the calibration parameters into the screen driver chip to complete the calibration. This achieves adaptive and intelligent calibration processing for screens of different panel types and in different states, avoiding the problems of low calibration accuracy, poor generalization ability, and inability to adapt to new panels and dynamic environments caused by reliance on manual experience in traditional solutions. It significantly improves the accuracy, stability, and automation of screen calibration.

[0018] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent from the following description. Attached Figure Description

[0019] The above and other objects, features, and advantages of exemplary embodiments of this application will become readily apparent from the following detailed description taken in conjunction with the accompanying drawings. Several embodiments of this application are illustrated in the drawings by way of example and not limitation, in which: In the accompanying drawings, the same or corresponding reference numerals indicate the same or corresponding parts.

[0020] Figure 1 A schematic diagram illustrating the implementation flow of the screen color calibration method provided in an embodiment of this application is shown; Figure 2 This illustration shows a schematic diagram of the implementation process of training the color calibration model of the screen color calibration method provided in the embodiments of this application; Figure 3 A schematic diagram illustrating the color calibration effect of the screen color calibration method provided in this application embodiment is shown; Figure 4 A schematic diagram of the composition structure of the screen color calibration device provided in the embodiments of this application is shown; Figure 5 A schematic diagram of the composition structure of an electronic device according to an embodiment of this application is shown. Detailed Implementation

[0021] To make the objectives, features, and advantages of this application more apparent and understandable, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0022] First, the application scenarios of this application embodiment are explained. Screen color accuracy is a core indicator for measuring the color performance of a device, directly affecting everyday visual experiences such as watching movies and reading, as well as the accuracy of color output in professional scenarios such as post-production photography, print design, and medical imaging. Color accuracy is quantified by the ΔE color difference value; the smaller the ΔE value, the higher the degree of conformity between the screen display color and the standard color. As users' requirements for color accuracy continue to increase, existing color calibration methods can no longer meet the needs of professional scenarios, generally exhibiting large ΔE values ​​and discrete ΔE values ​​distributed around 0.7 after multiple calibrations. The issue between 2.1.

[0023] For example, related technologies propose a fixed-screen color calibration scheme based on engineers' experience. This scheme is designed by engineers for mainstream screen models such as Liquid Crystal Display (LCD) and Organic Light-Emitting Diode (OLED), with preset values ​​of 30. The system uses 50 fixed color calibration frames, collects actual color data using a colorimeter, calculates the ΔE value, and then writes the corresponding color compensation parameters into the screen driver chip to complete a single calibration. However, this solution has significant drawbacks: the color calibration frames rely on manual experience and cannot adapt to the characteristics of new panels such as Mini LED (sub-millimeter light-emitting diode) local backlighting and Micro OLED (micro organic light-emitting diode) pixel self-emissive nonlinearity, resulting in ΔE values ​​generally greater than 1.5 after color calibration of OLED screens; the stability of multiple color calibrations of the same model of screen is poor, with significant ΔE value dispersion; it lacks dynamic compensation and self-optimization capabilities, and does not incorporate real-time status parameters such as ambient light and screen temperature, resulting in insufficient color calibration accuracy, stability, and scene adaptability.

[0024] Therefore, to solve the above-mentioned technical problems, this application provides a screen calibration method and an electronic device. This screen calibration method can be executed by an electronic device with data processing and peripheral control capabilities, such as a laptop computer, a monitor, or a dedicated calibration device.

[0025] Figure 1 A schematic diagram illustrating the implementation flow of the screen color calibration method provided in this application embodiment is shown.

[0026] refer to Figure 1 This application first provides a screen color calibration method, which includes: Operation 101: Obtain the panel type, operating status information, and color information of the screen to be calibrated.

[0027] Among them, "screen to be calibrated" refers to any display screen that needs to be corrected for color accuracy, including but not limited to laptop screens, desktop monitors, TV screens, tablet screens, smartphone screens, medical imaging monitors, professional photography monitors, and vehicle displays.

[0028] Panel type represents the category of display panel used in the screen to be calibrated. Different panel types differ fundamentally in their light-emitting principles, driving methods, brightness response curves (Gamma characteristics), color gamut coverage, and aging patterns. Therefore, considering that a single color calibration strategy cannot be universally applied to all panel types, this application designs adaptive color calibration processing logic for different panel types. Panel types include, but are not limited to, OLED, Mini-LED, Micro-LED, LCD, and other emerging display panels.

[0029] In addition to panel type, to adapt to complex and ever-changing usage environments and the screen's real-time status, it is also necessary to acquire operational status information. Operational status information characterizes the working environment and physical parameters of the screen to be calibrated at the current moment, including but not limited to ambient light intensity, screen temperature, continuous screen usage time, and current screen brightness setting. Ambient light intensity can be collected by an ambient light sensor; screen temperature can be collected by a temperature sensor attached to the screen panel; continuous screen usage time can be obtained through a timing module or by reading the working time recorded in the screen driver chip; and the current screen brightness setting can be obtained by reading the brightness control register of the screen driver chip or through the operating system's display interface.

[0030] Color information refers to the actual color output data of the screen to be calibrated when displaying a specific image, and is the core input of the calibration operation. The essence of calibration is to measure the actual values ​​of the colors displayed on the screen and compare them with the reference values ​​of the standard color space to calculate compensation parameters used to correct deviations. Therefore, the accuracy of the color information directly determines the calibration effect. In this application, the color information can be obtained through a high-precision color analyzer, such as a CA410, CA310, or similar device. This high-precision color analyzer needs to be communicatively connected to the electronic device executing the method of this application. The color information includes at least the actual values ​​of the red, green, and blue primary color channels displayed on the screen.

[0031] Operation 102: Based on panel type, operating status information, and color information, construct the screen feature information of the screen to be calibrated.

[0032] The acquired heterogeneous feature data, such as panel type, operating status information, and color information, needs to be processed into a standardized format recognizable by the color calibration model, forming screen feature information that the color calibration model can recognize. For example, categorical features such as panel type and screen model are encoded and converted (e.g., one-hot encoding), while continuous features such as temperature, light intensity, and usage time are normalized. Finally, all processed features are integrated into a single, continuous feature vector, which serves as the screen feature information recognizable by the color calibration model.

[0033] Operation 103 inputs screen feature information into the color calibration model to obtain color calibration parameters output by the color calibration model; wherein, the color calibration model processes the screen feature information based on the decision tree that selects the corresponding branch according to the panel type to obtain the color calibration parameters.

[0034] To avoid the problems of low color calibration accuracy and poor adaptability caused by relying on human experience, this application pre-trains a dedicated color calibration model. This model is an improved random forest model, which internally adopts the subtree specialization architecture based on panel type designed in this application. After receiving screen feature information, the color calibration model first parses the panel type features, then selects the corresponding training branch within the model according to the panel type, and only activates the decision tree under that branch for inference calculation. The decision trees of other non-corresponding panel branches do not participate in the calculation, and finally outputs the optimal color calibration parameters adapted to the current screen to be calibrated.

[0035] Operation 104: Perform color calibration on the screen to be calibrated based on the calibration parameters.

[0036] Among them, color calibration parameters represent the correction amounts required to make the screen display colors approximate the standard color space reference values, including but not limited to the red channel gain parameter (Red channel gain, Green channel gain parameter Blue channel gain parameter ), Gamma offset ) and color temperature compensation parameters (Color temperature compensation, The red channel gain parameter adjusts the magnification of the screen's red channel to correct deviations in the red component; the green channel gain parameter adjusts the magnification of the green channel; and the blue channel gain parameter adjusts the magnification of the blue channel. These three gain parameters work together to correct the screen's color bias in the three primary colors of red, green, and blue. The gamma offset adjusts the screen's gamma curve, i.e., the transition characteristics of screen brightness from dark to bright, to improve the accuracy of grayscale representation. The color temperature compensation parameter corrects color temperature shift caused by screen temperature changes or other factors, ensuring that the screen's white point matches the target color temperature (e.g., 6500K).

[0037] Standard color space reference values ​​represent the standard color values ​​defined in a recognized and ideal color reference system. For example, in the sRGB color space, standard red, green, blue, and each grayscale point all have clearly defined theoretical tristimulus values ​​or RGB values; corresponding reference values ​​also exist in different color spaces such as DCI-P3, Adobe RGB, or Rec.709. These reference values ​​represent the accurate colors that a screen should display in that color space and serve as a reference for measuring the degree of deviation of the actual displayed colors.

[0038] After obtaining the aforementioned color calibration parameters, the electronic device executing the method of this application writes the color calibration parameters into the corresponding register of the screen driver chip of the screen to be calibrated via a communication interface (such as an I2C bus, a display data channel DDC, or a custom communication protocol). Upon receiving these parameters, the screen driver chip dynamically adjusts its internal processing logic based on the parameter values. This includes adjusting the digital gain or analog amplification factor of the RGB three channels, modifying the Gamma lookup table (LUT) curve, and adjusting the white balance or color temperature settings, so that the actual displayed color of the screen matches the standard color space reference value, thus completing the screen color calibration.

[0039] Thus, this embodiment of the application obtains the panel type, operating status information, and color information of the screen to be calibrated, and constructs unified screen feature information based on the above information. Then, it uses a pre-trained calibration model with a panel type-based specialized architecture to automatically infer calibration parameters suitable for the current screen. Finally, the calibration parameters are written into the screen driver chip to complete the calibration. This achieves adaptive and intelligent calibration processing for screens of different panel types and states, avoiding the problems of low calibration accuracy, poor generalization ability, and inability to adapt to new panels and dynamic environments caused by relying on human experience. It significantly improves the accuracy, stability, and automation of screen calibration.

[0040] Figure 2 This diagram illustrates the implementation flow of the training operation of the color calibration model in the screen color calibration method provided in this embodiment.

[0041] refer to Figure 2 In one embodiment of this application, the color calibration model can be trained automatically by the electronic device executing the method of this application, or it can be pre-trained on other devices (such as servers, cloud computing platforms, or dedicated training equipment) to reduce the computational burden on the device, facilitate model version management, or utilize more powerful computing resources. The trained model is then deployed to the electronic device executing the screen color calibration method of this application. This application does not limit this approach. Specifically, taking the scenario where the color calibration model is trained automatically by the electronic device executing the method of this application as an example, the color calibration model is trained through the following operations: Operation 201: Obtain the training sample set. The training sample set includes multiple training samples. Each training sample includes a sample screen, the screen feature information of the sample screen, and the true color correction parameters.

[0042] Specifically, a large number of training samples covering various screen models, panel types, and environmental conditions are collected using a high-precision color analyzer (such as CA410, CA310, or similar devices). Screen feature information and corresponding real color calibration parameters are constructed for each training sample to form a training sample set. In this embodiment, the screen feature information of each training sample may include multiple features, such as screen model, panel type, actual RGB (red, green, blue) values ​​of the screen collected by the high-precision color analyzer, ambient light intensity, and screen temperature.

[0043] The true color calibration parameters for each training sample are calculated based on color physics. For example, these true color calibration parameters may include red channel gain parameters, green channel gain parameters, blue channel gain parameters, gamma offset, and color temperature compensation parameters. These parameters are obtained based on reference values ​​in standard color spaces (e.g., sRGB, DCI-P3, Adobe RGB) according to the following physical calculations: , , .

[0044] in, This represents the gain parameter for the red channel. This represents the gain parameter of the green channel. This represents the gain parameter of the blue channel. The actual RGB values ​​of the screen. The base RGB values ​​for the standard color space. The gain coefficient, adapted to the panel type, was obtained through experimental calibration.

[0045]

[0046] in, This is the gamma offset. The actual gamma value of the screen. The target gamma value (e.g., 2.2) is divided by 2 to control the correction step size and avoid overcorrection.

[0047]

[0048] in, These are the color temperature compensation parameters. Set the target color temperature (e.g., 6500K). This represents the screen temperature (in °C), and 25 is the reference temperature. The temperature-color temperature compensation coefficient was obtained through experimental calibration.

[0049] In the above manner, a correspondence between screen feature information and real color calibration parameters is established for each training sample, forming a training sample set that can be used for supervised learning. It is understood that the above calculation method is merely an example of this application, and those skilled in the art can use other color physics formulas to calculate the real color calibration parameters according to actual needs, as long as the optimal calibration parameters can be obtained based on the collected actual values ​​and the target standard values.

[0050] Operation 202 uses panel type as the splitting feature of the root node to split out branches corresponding to each panel type, and divides the training sample set into branches corresponding to each panel type according to the panel type.

[0051] The color calibration model in this application uses a random forest as its basic model structure. Unlike conventional random forests, which randomly select the optimal splitting feature from all features at the root node of each decision tree, this application employs a specific design for the model training method tailored to the characteristics of screen color calibration scenarios. Specifically, when constructing each decision tree in the random forest, the root node is forced to split according to the panel type feature, rather than being randomly selected from features. The number of possible panel type values ​​determines the number of branches, such as OLED, LCD, Mini-LED, etc. Each branch contains only training samples corresponding to the panel type, and subsequent decision trees grown within that branch will specifically learn the color calibration rules for that panel type. Operation 203: Within each branch corresponding to each panel type, based on the set feature importance weights, the sample weights of each training sample corresponding to each branch, and the set decision tree hyperparameters corresponding to each branch, the decision tree construction is performed iteratively with the goal of minimizing the multi-objective hierarchical loss function, resulting in multiple decision trees corresponding to each branch.

[0052] The screen feature information includes various types of features, each with varying degrees of impact on color correction. Therefore, during training, this application pre-configures feature importance weights to prioritize features with greater impact on color correction during node splitting. These weights assign different priorities to different features during node splitting. Sample weights are used to dynamically adjust the focus on difficult samples during training, making the model more focused on samples that are difficult to fit. The decision tree hyperparameters for each branch include, but are not limited to, the maximum tree depth and the minimum number of samples per leaf node, and can be configured according to the characteristics of different panel types. The multi-objective hierarchical loss function simultaneously constrains different types of parameters in the color correction parameters, guiding the model to generate more accurate color correction parameters.

[0053] Specifically, for each panel type branch, the training process is as follows: First, based on the sample weights of each training sample (which can be set to equal weights initially), weighted sampling with replacement is performed on the training samples of that branch to generate the training subset for the current decision tree to be built.

[0054] Then, a decision tree is recursively constructed starting from the root node. At each node, each feature in the screen feature information, such as panel type, color information, ambient light intensity, and screen temperature, is traversed, and the multi-objective hierarchical loss function value that the child node can achieve after using this feature as the splitting criterion is evaluated. The feature that maximizes the decrease in the multi-objective hierarchical loss function is selected as the splitting feature for the current node. In this evaluation process, a feature importance weight is introduced, and the decrease in the multi-objective hierarchical loss function for different features is multiplied by the corresponding weight factor to prioritize splitting features that have a greater impact on the color correction effect.

[0055] After determining the splitting features, the samples of the current node are divided into different child nodes according to the different values ​​of the splitting features. This splitting process is executed recursively until the stopping condition specified by the hyperparameters of the decision tree corresponding to that branch is reached, such as the maximum depth of the tree or the minimum number of samples contained in a leaf node. When the splitting stops, the current node becomes a leaf node, and the mean of the color correction parameters of all training samples within that leaf node is used as the output value of that leaf node. At this point, a decision tree is constructed.

[0056] Repeat the weighted sampling and decision tree construction cycle described above until the preset number of decision trees is reached. Through this process, a set of decision trees specifically adapted to the color correction rules of that panel type is generated within each panel type branch.

[0057] Operation 204 combines all the decision trees built from each branch into a color correction model.

[0058] After constructing all decision trees, the decision trees generated under each panel type branch are merged to form a complete color calibration model. When used, the color calibration model will activate only the decision tree under the corresponding branch for inference based on the panel type field in the input, thereby achieving specialized color calibration based on the subtree of the panel type.

[0059] Thus, the color calibration model obtained through the above training operations can use decision trees with different branches for different panel types, effectively improving color calibration accuracy and generalization ability.

[0060] In one embodiment of this application, the screen feature information includes the following features: panel type, screen model shown in the operation status information, screen temperature shown in the operation status information, and ambient light intensity shown in the operation status information; the feature importance weight is set as the feature weight assigned to each feature in the screen feature information; wherein, the feature weight of each feature is used to characterize the importance of each feature to color calibration.

[0061] Model training in this application embodiment ( Figure 2 Corresponding description) and model reasoning ( Figure 1 The screen feature information structure used in the corresponding description is consistent, including panel type and features such as screen model, screen temperature, and ambient light intensity as shown in the operating status information. Accordingly, the feature importance weight is set as the feature weight assigned to each feature in the above screen feature information, which is used to characterize the importance of each feature to color calibration.

[0062] For example, considering that different features have significantly different degrees of influence on screen color deviation—for instance, panel type almost determines the basic curve of color response, while screen model only brings minor calibration differences—this application adopts a hierarchical configuration strategy for setting feature importance weights. Each feature in the screen feature information is exemplarily divided into three levels, and feature weights are configured hierarchically: Tier 1 (High Importance): This includes panel type, ambient light intensity, and screen temperature. These are the fundamental factors affecting color performance and are given a high feature weight, such as 1.5.

[0063] Second level (medium importance): Includes the actual red channel value, actual green channel value, and actual blue channel value in color information (i.e., the aforementioned...). They are the direct carriers of color, but their influence is modulated by first-level factors, giving them a moderate feature weight, such as 1.2.

[0064] The third tier (low importance): includes screen model. It only represents minor tuning differences within the same panel technology, assigning a weight to basic features, such as 1.0.

[0065] Thus, with the above configuration, the varying degrees of influence of different features on the color correction effect can be considered during node splitting. Specifically, the weighted splitting gain can be calculated using the following formula:

[0066] in, The original splitting gain is calculated with the objective of minimizing the multi-objective hierarchical loss function. This represents the feature weights corresponding to the current feature. After obtaining the weighted split gain, the node can be selected during splitting. The largest feature is used as the splitting feature of the current node. In this way, when building a decision tree, splitting can be prioritized along the feature direction that has the greatest impact on color output, thereby capturing the main contradictions in the data more quickly, improving learning efficiency and model accuracy, and enhancing the generalization ability to different display panels.

[0067] In one embodiment of this application, during the process of cyclically executing decision tree construction, in response to the completion of the construction of the previously constructed decision tree, all previously constructed decision trees are used as an integrated model to perform inference on the training sample set to obtain the predicted color correction parameters of each training sample; based on the error between the predicted color correction parameters and the true color correction parameters of each training sample, the sample weights of the training samples are updated, and the updated sample weights are used to construct decision trees in the subsequent stages.

[0068] Specifically, this application introduces a dynamic sample weight update mechanism to focus on difficult samples that are hard to fit in the training sample set. For example, extreme color blocks at the edge of the color space, or images with extremely strong nonlinear color response displayed by new panels (such as MicroOLED), are all considered difficult samples. Conventional random forests treat all samples equally, which can easily lead to underfitting of the model on these samples. To address this, this application designs sample weights and corresponding sample weight update methods, as follows: First, initialize the sample weights of all training samples to equal values, for example, set the initial weights to 1. After each round of decision tree construction, use the ensemble model composed of all the constructed decision trees to perform inference on the training sample set to obtain the predicted color correction parameters for each training sample. And combined with the actual color calibration parameters of the sample Calculate the absolute error This absolute error reflects the difficulty of fitting the training samples to the current ensemble model; the larger the error, the more difficult it is for the training samples to be correctly fitted by the current ensemble model.

[0069] For difficult samples with large errors, their sample weights are updated according to the following formula before the next round of decision tree construction:

[0070] in, For the first Training samples The weight, These are preset hyperparameters used to control the intensity of focusing on difficult samples. Through the above sample weight update method, samples with larger errors are assigned higher weights in the next round.

[0071] Thus, through the above configuration of dynamic sample weight update and weighted sampling, in the subsequent decision tree construction, the probability of each training sample being selected will be proportional to its current sample weight, enabling the subsequent decision tree to pay more attention to the difficult samples with poor fitting results, thereby effectively improving the boundary performance of the model in difficult scenarios such as edge color blocks and new panels, and enhancing the adaptability and color calibration accuracy of different display technologies.

[0072] In one embodiment of this application, the following progressive early stopping operation is also performed to avoid overfitting due to overtraining and the inability to fully utilize the model's potential due to blindly setting training rounds: obtaining a validation sample set; obtaining the function value of the multi-objective hierarchical loss function based on the validation sample set during the cyclic execution of decision tree construction; halving the learning rate of decision tree construction in response to the function value changing less than a set change threshold for a first set number of consecutive rounds; stopping the construction of decision trees in response to the function value changing less than a set change threshold for a second set number of consecutive rounds, combining all decision trees into a color correction model, with the second set number of rounds being greater than the first set number of rounds.

[0073] Specifically, during the iterative construction of decision trees, if there are too many training rounds, the model is prone to overfitting to noise in the training data, leading to a decrease in generalization ability; if there are insufficient training rounds, the model may not have fully learned the patterns in the data and cannot achieve optimal performance. Therefore, this application's embodiments introduce a progressive early stopping mechanism. After each round of decision tree construction, the multi-objective hierarchical loss function value of the current ensemble model is calculated using the validation sample set. When the change value of the multi-objective hierarchical loss function value for a consecutive first set of rounds (e.g., three consecutive rounds) is less than a preset change threshold (i.e., the loss no longer decreases significantly), it indicates that the model's convergence speed has slowed down. At this point, the learning rate (e.g., the shrinkage rate of the leaf node output values) for decision tree construction is halved, and subsequent tree construction continues with a finer step size. If the change value of the multi-objective hierarchical loss function value for a consecutive second set of rounds (e.g., five consecutive rounds) is still less than the threshold, and the second set of rounds is greater than the first set of rounds, it is determined that the model is close to its optimal state, and continuing training may lead to overfitting risks. Therefore, the construction of the decision trees is immediately terminated, and all constructed decision trees are combined into the final color correction model.

[0074] By employing the aforementioned progressive early stopping strategy, this application can effectively control model complexity, avoid unnecessary training rounds, and prevent overfitting while ensuring model accuracy.

[0075] In one embodiment of this application, the above operation 204, which combines all decision trees constructed by each branch into a color correction model, includes: obtaining the contribution of each decision tree on the validation sample set, removing decision trees whose contribution is lower than a set score threshold, and combining the remaining decision trees into a color correction model.

[0076] Specifically, after constructing all decision trees, some decision trees may contribute negligibly to the final prediction result. These redundant trees not only increase storage volume but also reduce computational efficiency during online inference. Therefore, this application further introduces an importance pruning mechanism. For each constructed decision tree, its contribution to the validation sample set is calculated. This contribution can be represented as the total feature importance score of the decision tree on the validation set, or its marginal contribution to the prediction result of the ensemble model. The contribution of each tree is compared with a preset score threshold (e.g., 0.01). All decision trees with contributions below the threshold are removed, and only decision trees with higher contributions are retained. The remaining decision trees are then recombined into the final color correction model.

[0077] Thus, by implementing the aforementioned importance pruning configuration, the storage size of the model can be effectively reduced while ensuring color calibration accuracy. Experiments show that the model size can be reduced by about 25% after pruning, while improving online inference speed by about 35%. This makes the color calibration model more suitable for resource-constrained embedded environments, such as laptops, monitors, or dedicated color calibration equipment.

[0078] In one embodiment of this application, the true color calibration parameters include red channel gain parameters, green channel gain parameters, blue channel gain parameters, gamma offset, and color temperature compensation parameters; the multi-objective hierarchical loss function is related to the prediction error of each parameter in the true color calibration parameters and the smoothness loss term.

[0079] Specifically, considering that different color correction parameters have different physical meanings, numerical ranges, and degrees of influence on the final color correction effect, using a single loss function (such as mean absolute error) to equally penalize the prediction errors of all parameters would fail to guide the model to focus on key parameters. Therefore, this application proposes a multi-objective hierarchical loss function with parameter importance weighting, the expression of which is as follows:

[0080] in, Indicates the gain parameters of the red, green, and blue channels ( The mean absolute error between the predicted value and the corresponding true color calibration parameters in the training samples. Since the RGB gain parameter has the most direct and critical impact on color accuracy, this application assigns it the highest weighting coefficient. ,For example . Indicates gamma offset ( The gamma parameter is the mean absolute error between the predicted value and the corresponding true color correction parameters in the training samples. The gamma parameter is crucial for the accuracy of grayscale transitions, therefore it is assigned the second highest weight coefficient. ,For example . Indicates color temperature compensation parameters ( The mean absolute error between the predicted value and the corresponding true color calibration parameters in the training samples. Although the color temperature parameter is affected by ambient temperature and needs compensation, its importance is relatively low, and a weighting coefficient is assigned. ,For example . This is a smoothness loss term used to constrain the predicted values ​​of color correction parameters from undergoing drastic changes in adjacent regions of the color space (e.g., adjacent gray levels or similar color patches), thereby ensuring a natural and smooth color transition in the corrected image and avoiding color blocks or banding. This application assigns weighting coefficients to the smoothness loss. ,For example The above weighting coefficients These are all preset hyperparameters, which can be adjusted according to the actual application scenario to balance the importance of each loss term.

[0081] Thus, through the aforementioned multi-objective hierarchical loss function, the model can concentrate its limited expressive power on key color correction parameters such as RGB gain and gamma shift during training, while improving the visual quality of the color correction results through smoothness constraints, and avoiding image quality degradation caused by parameter mutations.

[0082] In one embodiment of this application, the screen feature information is further constructed based on a time decay factor; the time decay factor is obtained through the following operations: obtaining the usage time of the screen to be calibrated and the decay coefficient corresponding to the panel type; and determining the time decay factor based on the usage time and the decay coefficient.

[0083] Specifically, during use, screens experience slow color shifts due to the natural aging of their luminescent materials, such as brightness decay, color temperature shift, or color gamut shrinkage. Existing color calibration methods only provide a one-time correction for the current screen state and cannot detect or predict these long-term changes. Therefore, this application quantifies the screen's aging effect as a time decay factor and incorporates it into the construction of screen feature information. This time decay factor is calculated using the screen's cumulative usage time (e.g., in hours) and the decay coefficient corresponding to the panel type. The decay coefficient can be calibrated by conducting accelerated aging experiments on different panel types (such as OLED, LCD, and Mini-LED).

[0084] For example, in the data preprocessing stage, the time decay factor is treated as a new feature dimension and combined with the aforementioned panel type, operating status information (ambient light intensity, screen temperature, screen model, etc.) and color information (actual RGB values) to form complete screen feature information. The calculation method for this time decay factor can be expressed as follows:

[0085] in, To determine the degradation coefficient obtained through accelerated aging experiments, different panel types The value can be independently calibrated to reflect its aging characteristics.

[0086] Thus, by configuring the time decay factor as described above, the model can dynamically compensate for color deviations caused by aging, achieving a leap from calibrating only the current state to predicting and compensating for future states, supporting preventative calibration, and extending the lifespan of the device's color accuracy.

[0087] In one embodiment of this application, after completing the above-mentioned decision tree construction and combination, the trained color calibration model is further evaluated and saved. Specifically, a pre-reserved test sample set is obtained (the samples in this test sample set do not participate in the training process), the screen feature information of each sample in the test sample set is input into the color calibration model to obtain the color calibration parameters predicted by the model, and the mean absolute error (MAE) between the predicted values ​​and the actual color calibration parameters labeled in the samples is calculated. For example, the accuracy requirement can be set to an MAE of no more than 0.03 to ensure that the model has sufficient color calibration accuracy. When the model meets the accuracy requirement, it is serialized and saved as a model file for loading and calling in the online color calibration stage.

[0088] In one embodiment of this application, the operation 101 described above, obtaining the color information of the screen to be calibrated, includes: in response to the panel type being an organic light-emitting diode (OLED), selecting a first predetermined number of images from a basic image library and an OLED-specific image library; in response to the panel type being a sub-millimeter light-emitting diode (EMD), selecting a second predetermined number of images from a basic image library and an EED-specific image library; in response to the panel type being a liquid crystal display (LCD), selecting a third predetermined number of images from a basic image library; in response to the ambient light intensity indicated by the operating status information being greater than a predetermined intensity threshold, additionally selecting a fourth predetermined number of images from an intensity adaptation image library; merging all selected images and removing duplicates to obtain a color calibration image set; sequentially displaying each image in the color calibration image set and collecting the color parameter information of each image as color information; wherein the number of images in the color calibration image set does not exceed a predetermined threshold.

[0089] To efficiently and specifically acquire screen color information, this application dynamically selects calibration images based on panel type and ambient light intensity, rather than using a fixed set of images. Specifically, this application pre-sets multiple image libraries for different panel types and ambient light conditions, including a basic image library, dedicated image libraries for different panel types, and a strong light adaptation image library. The basic image library covers basic color and brightness responses, the dedicated image library captures the non-linear or zoned backlight characteristics of specific panel types, and the strong light adaptation image library enhances calibration performance in strong light environments. The selection rules are as follows:

[0090] in, This represents the final set of color-corrected images selected. Represents the basic image library; This represents a dedicated image library for organic light-emitting diodes (OLEDs), including dark scenes, highly saturated colors, and fast-switching images. This represents a dedicated image library for sub-millimeter LEDs, including images featuring local highlights, light and dark transitions, and dynamic backlighting. This indicates the panel type, which can be OLED, Mini LED, or LCD. The formula means: based on the panel type... Depending on the panel type, the union of the corresponding basic image library and the dedicated image library is selected as the color calibration image set; when the panel type is LCD, only the basic image library is used.

[0091] In addition, when the ambient light is stronger than the set strong light threshold (e.g., 500 lux), based on the above rule filtering results, an additional fourth set number (e.g., 2 sets) of images are selected from the strong light adaptation image library to ensure the calibration effect under strong light conditions.

[0092] In order to ensure the real-time nature of the color calibration process and the efficiency of data acquisition, the number of images in the selected color calibration image set shall not exceed a preset threshold (e.g., 20 sets). This threshold can be flexibly configured according to the actual application scenario, hardware performance and color calibration accuracy requirements.

[0093] After determining the color calibration screen set, each screen is displayed sequentially, and the actual red, green, and blue channel values ​​displayed on the screen for each screen are collected using a high-precision color analyzer (e.g., CA410). This data is used as color information. Through the dynamic filtering mechanism described above, the most representative color data can be obtained with a small number of images (e.g., 15-20 sets), which ensures both acquisition efficiency and provides high-quality input for subsequent color calibration models.

[0094] In one embodiment of this application, color calibration is a closed-loop iterative process, which is only completed when the color difference value after calibration meets the preset accuracy requirements. Therefore, after each color calibration, the color information of the screen to be calibrated is also collected and compared with the standard color information to determine the color difference value; in response to the color difference value being greater than the set color difference threshold, the screen feature information is updated based on the newly collected color information, and the color calibration parameters are re-determined.

[0095] The standard color information refers to the aforementioned standard color space reference values, such as the standard color values ​​defined in standard color spaces like sRGB, DCI-P3, or AdobeRGB (e.g., standard RGB values, standard color temperature, standard gamma values). After the color calibration parameters are written to the screen driver chip and take effect, the actual color information of the current screen is collected again by a high-precision color analyzer and compared with the standard color information to calculate the color difference value. For example, a color difference threshold is set. .like If the color calibration is successful, the process ends. If the current iteration count does not exceed the preset limit (e.g., 3 times), the screen feature information is reconstructed based on the newly acquired color information (e.g., the updated actual RGB values), and then input into the color calibration model for inference to obtain updated color calibration parameters, which are then reissued. This closed-loop verification process is repeated until the color difference value meets the requirements or reaches the iteration limit. Through this closed-loop iterative optimization mechanism, this application can ensure the final reliability of each color calibration result, overcoming the shortcomings of unstable results in traditional single-loop calibration.

[0096] Figure 3 A schematic diagram illustrating the color calibration effect of the screen color calibration method provided in this application embodiment is shown.

[0097] refer to Figure 3 To verify the actual effect of the screen color calibration method of this application, the embodiments of this application conducted multiple batches of color difference value ΔE tracking experiments on a certain model of screen produced using this method. Figure 3 The horizontal axis represents different production or validation batches, including the PCR validation stage, the first mass production stage, and multiple subsequent mass production batches (December 2024 to March 2025); the left vertical axis represents the color difference value ΔE, ranging from 0.3 to 1.1; the right vertical axis represents the output of each batch (unit: units), ranging from 0 to 5000.

[0098] The graph shows various curves or data points, each corresponding to a different color space or baseline value: "Input" represents the baseline color difference value of the uncalibrated original screen; "Target" represents the target color difference value (usually set as the ideal calibration target, e.g., ΔE=0); "sRGB" represents the ΔE value measured after calibration in the sRGB color space; and "P3" represents the ΔE value measured after calibration in the DCI-P3 color space. Additionally, the bar chart represents the production volume of each batch.

[0099] Depend on Figure 3It can be seen that the Input ΔE value before calibration is significantly higher than that after calibration (the Input curve in the figure is located at a higher position, for example, above 0.9 in the PCR validation stage); the Target baseline is near 0. After calibration using the method of this application, in the PCR validation stage, the ΔE in sRGB space is approximately 0.74, and the ΔE in P3 space is approximately 0.77; in the first mass production stage, the ΔE in sRGB space drops to 0.68, and the ΔE in P3 space drops to 0.70; in subsequent batches (24-Dec, 25-Jan, 25-Feb, 25-Mar), the ΔE in sRGB space is stable between 0.58 and 0.75, and the ΔE in P3 space is stable between 0.50 and 0.75. The ΔE value of all batches is controlled below 0.8, and the fluctuation between batches is minimal, which is far better than the unstable performance of the ΔE discrete distribution of 0.7 to 2.1 in related calibration schemes. Meanwhile, production volume gradually increased from a small amount during the PCR validation phase to several thousand units, demonstrating that the proposed method maintains high precision and stability in color calibration even in mass production environments. In summary, the screen color calibration method proposed in this application demonstrates excellent color accuracy control and mass production consistency in experiments.

[0100] Figure 4 A schematic diagram of the composition structure of the screen color calibration device provided in the embodiments of this application is shown.

[0101] refer to Figure 4 This application also provides a screen calibration apparatus, which includes: The acquisition module 301 is used to acquire the panel type, operating status information and color information of the screen to be calibrated; The construction module 302 is used to construct the screen feature information of the screen to be calibrated based on the panel type, running status information and color information; The parameter determination module 303 is used to input screen feature information into the color calibration model and obtain the color calibration parameters output by the color calibration model; wherein, the color calibration model processes the screen feature information based on the decision tree that selects the corresponding branch of the panel type to obtain the color calibration parameters; The color calibration module 304 is used to calibrate the screen to be calibrated based on the color calibration parameters.

[0102] It should be noted that the description of the apparatus in this application embodiment is similar to the description of the method embodiment above, and has similar beneficial effects as the method embodiment, therefore it will not be repeated. For any technical details not covered in the screen calibration apparatus provided in this application embodiment, please refer to... Figures 1 to 3 The meaning is understood in accordance with the description of any of the accompanying drawings.

[0103] According to embodiments of this application, this application also provides an electronic device and a readable storage medium.

[0104] Figure 5 A schematic block diagram of an example electronic device 400 that can be used to implement embodiments of this application is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the application described and / or claimed herein.

[0105] like Figure 5 As shown, the electronic device 400 includes a computing unit 401, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 402 or a computer program loaded from a storage unit 408 into a random access memory (RAM) 403. The RAM 403 may also store various programs and data required for the operation of the electronic device 400. The computing unit 401, ROM 402, and RAM 403 are interconnected via a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0106] Multiple components in electronic device 400 are connected to I / O interface 405, including: input unit 406, such as keyboard, mouse, etc.; output unit 407, such as various types of displays, speakers, etc.; storage unit 408, such as disk, optical disk, etc.; and communication unit 409, such as network card, modem, wireless transceiver, etc. Communication unit 409 allows electronic device 400 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0107] The computing unit 401 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 401 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 401 performs the various methods and processes described above, such as screen calibration methods. For example, in some embodiments, the screen calibration method may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 408. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 400 via ROM 402 and / or communication unit 409. When the computer program is loaded into RAM 403 and executed by the computing unit 401, one or more steps of the screen calibration method described above may be performed. Alternatively, in other embodiments, the computing unit 401 may be configured to perform screen calibration methods by any other suitable means (e.g., by means of firmware).

[0108] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transferring data and instructions to the storage system, the at least one input device, and the at least one output device.

[0109] The program code used to implement the methods of this application may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0110] In the context of this application, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0111] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0112] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0113] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0114] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this application can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this application can be achieved, and this is not limited herein.

[0115] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "a plurality of" means two or more, unless otherwise explicitly specified.

[0116] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A screen color calibration method, characterized in that, The method includes: Obtain the panel type, operating status information, and color information of the screen to be calibrated; Based on the panel type, the operating status information, and the color information, the screen feature information of the screen to be calibrated is constructed; The screen feature information is input into the color calibration model to obtain the color calibration parameters output by the color calibration model; wherein, the color calibration model processes the screen feature information based on the decision tree that selects the corresponding branch of the panel type to obtain the color calibration parameters; The screen to be calibrated is calibrated based on the calibration parameters.

2. The method according to claim 1, characterized in that, The color calibration model was trained through the following operations: Obtain a training sample set, which includes multiple training samples, each of which includes a sample screen, screen feature information of the sample screen, and real color correction parameters; Using panel type as the splitting feature of the root node, branches corresponding to each panel type are split out, and the training sample set is divided into branches corresponding to each panel type according to the panel type. Within each branch corresponding to each panel type, based on the set feature importance weights, the sample weights of each training sample corresponding to each branch, and the set decision tree hyperparameters corresponding to each branch, the decision tree construction is performed iteratively with the goal of minimizing the multi-objective hierarchical loss function, resulting in multiple decision trees corresponding to each branch. All decision trees constructed from each branch are combined into the color correction model.

3. The method according to any one of claims 1-2, characterized in that, The screen feature information includes the following features: panel type, screen model shown in the operation status information, screen temperature shown in the operation status information, and ambient light intensity shown in the operation status information; The set feature importance weight is the feature weight assigned to each feature in the screen feature information; wherein, the feature weight of each feature is used to characterize the importance of each feature to color calibration.

4. The method according to claim 2, characterized in that, During the iterative process of building the decision tree, the method further includes, for the previously built decision tree: In response to the completion of the construction of the previous decision trees, all the previously constructed decision trees will be used as an ensemble model to infer on the training sample set to obtain the predicted color correction parameters for each training sample. Based on the error between the predicted color correction parameters and the actual color correction parameters of each training sample, the sample weights of the training samples are updated, and the updated sample weights are used to construct a decision tree later.

5. The method according to claim 2, characterized in that, The method further includes: Obtain the verification sample set; During the process of constructing the decision tree in a loop, the function value of the multi-objective hierarchical loss function is obtained based on the validation sample set; If the change in the function value for a consecutive first set number of rounds is less than a set change threshold, the learning rate of the decision tree will be halved. In response to the function value changing less than the set change threshold for a second consecutive set number of rounds, the construction of the decision tree is stopped, and all decision trees are combined into the color correction model, wherein the second set number of rounds is greater than the first set number of rounds.

6. The method according to claim 5, characterized in that, The process of combining all decision trees constructed from each branch into the color correction model includes: Obtain the contribution of each decision tree on the validation sample set, remove decision trees whose contribution is lower than the set score threshold, and combine the remaining decision trees into the color correction model.

7. The method according to claim 2, characterized in that, The actual color calibration parameters include red channel gain parameters, green channel gain parameters, blue channel gain parameters, gamma offset, and color temperature compensation parameters; The multi-objective hierarchical loss function is related to the prediction error of each parameter in the true color correction parameters and the smoothness loss term.

8. The method according to any one of claims 1-2, characterized in that, The screen feature information is also constructed based on a time decay factor; the time decay factor is obtained through the following operations: Obtain the usage time of the screen to be calibrated and the attenuation coefficient corresponding to the panel type; The time decay factor is determined based on the usage duration and the decay coefficient.

9. The method according to claim 1, characterized in that, The panel types include organic light-emitting diodes, sub-millimeter light-emitting diodes, and liquid crystal displays; Obtain the color information of the screen to be calibrated, including: In response to the panel type being organic light-emitting diode (OLED), a first predetermined number of images are selected from the basic image library and the OLED-specific image library; In response to the panel type being sub-millimeter light-emitting diode, a second set number of images are selected from the basic image library and the sub-millimeter light-emitting diode dedicated image library; In response to the panel type being a liquid crystal display, a third predetermined number of images are selected from the basic image library; In response to the ambient light intensity shown in the operating status information being greater than the set strong light threshold, a fourth set number of images are selected from the strong light adaptation image library; After merging and deduplicating all selected images, a color calibration image set is obtained. Each frame in the color calibration frame set is displayed sequentially, and the color parameter information of each frame is collected as the color information; The number of images in the color calibration image set does not exceed a set threshold.

10. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-9.

Images