Display panel color compensation model, training method, compensation method and device

Abstract

The invention discloses a display panel color compensation model, a training method, a compensation method and a device, and the model comprises a three-channel separation network which is used for extracting color image data from display image data of a display panel collected by a color camera, and the color image data comprises channel image data of at least one of R, G and B channels; the image processing network comprises at least one processing branch corresponding to each channel and is used for performing color compensation on the channel image data and outputting a corresponding channel compensation image; and the image fusion network is used for obtaining a color compensation image through image fusion based on the channel compensation image obtained by the image processing network so as to perform color compensation on the display panel based on the color compensation image. The color channel image data of the display image data acquired by the color camera are subjected to AI compensation and then fused to obtain the color compensation image, and the problem that the defect of uneven chromaticity still exists after single gray scale brightness compensation is avoided.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, in particular to a display panel color compensation model, a training method, a compensation method and a device. BACKGROUND

[0002] This section is intended to provide background information to facilitate a better understanding of embodiments of this application as set forth in the claim section. The description herein does not constitute admission of prior art.

[0003] A new generation of display devices is affected by process differences, evaporation shadows and quantum dot coating unevenness, and the RGB three-color luminance and color coordinates of different sub-pixels are inconsistent at the same gray level. Existing defect compensation methods include manual sampling compensation methods and single gray level defect compensation methods. The manual sampling compensation method is selected by engineers, and the compensation is manually adjusted, which is time-consuming and ineffective. The single color (gray level) compensation method of the display panel can only compensate for the luminance difference, and there are still defects of color difference uniformity in the display panel after luminance compensation. SUMMARY

[0004] One purpose of the present application is to provide a display panel color compensation model, which fuses color compensation images obtained by AI compensation of color channel image data of display image data collected by a color camera, thereby avoiding the problem of color difference uniformity after single gray level luminance compensation. Another purpose of the present application is to provide a display panel color compensation model training method. Another purpose of the present application is to provide a display panel color compensation method. Another purpose of the present application is to provide a display panel color compensation model construction device. Another purpose of the present application is to provide a display panel color compensation device.

[0005] To achieve the above purpose, the present application discloses a display panel color compensation model, comprising: a three-channel separation network for extracting color image data from display image data of a display panel collected by a color camera, wherein the color image data comprises channel image data of at least one of R, G and B channels; an image processing network comprising at least one processing branch corresponding to each channel, for color compensation of the channel image data, and outputting a corresponding channel compensation image; an image fusion network for obtaining a color compensation image by image fusion based on the channel compensation image obtained by the image processing network, so as to compensate the color of the display panel based on the color compensation image.

[0006] Optionally, the processing branch of the image processing network comprises a convolution layer, a bottleneck layer and a jump layer. The convolutional layer is used to adjust the number of channels and resolution of the channel image data; The bottleneck layer performs feature amplification on the channel image data adjusted by the convolutional layer; The skip layer connects the image data output by the bottleneck layer with the image data output by the convolutional layer to obtain a channel-compensated image.

[0007] Optionally, the convolutional layer includes: a first convolutional layer, a pooling layer, and a second convolutional layer; The first convolutional layer expands the number of channels in the channel image data; The pooling layer performs image compression on the channel image data after the channel number is expanded; The second convolutional layer adjusts the number of image channels in the compressed image data.

[0008] Optionally, the skip layer includes a first skip layer, a second skip layer, and a third skip layer; The first skip layer fuses the image data output by the bottleneck layer with the image data output by the second convolutional layer to obtain a first connected image; The second skip layer fuses the first connected image output by the first skip layer with the image data output by the first convolutional layer to obtain the second connected image; The third skip layer performs feature fusion between the second connected image output by the second skip layer and the channel image data to obtain a channel-compensated image.

[0009] Optionally, the image fusion network includes a color balance sub-network, used to calculate the adjustment coefficients of brightness and chromaticity of each pixel in each channel compensation image obtained by the image processing network, perform pixel value correction on each pixel in each channel compensation image based on the adjustment coefficients, and perform image fusion on the channel compensation images after pixel value correction to obtain a color compensation image.

[0010] This application also discloses a training method for the display panel color compensation model as described above, the method comprising: Acquire standard image data obtained by a color camera from standard color parts, and form standard data; Acquire the actual image data captured by the color camera and displayed on the corresponding grayscale display panel; A training dataset is formed based on the standard data and the actual image data. The color compensation model is obtained by training the pre-built compensation model framework using the training dataset.

[0011] Optionally, the compensation model framework further includes smoothing constraints on the color compensation image, and the step of training the pre-built compensation model framework using the training dataset to obtain the color compensation model includes: The training dataset is input into a pre-built compensation model framework for multiple iterative training until a preset training termination condition is reached to obtain the color compensation model. In each iterative training process, the model parameters of the compensation model framework are adjusted based on the smoothing constraint condition.

[0012] This application also discloses a display panel color compensation method using the display panel color compensation model described above, comprising: Obtain the actual color image data from the image displayed on the display panel to be compensated, captured by a color camera; The actual color image data is input into the trained compensation model to obtain the color-compensated image; Based on the color compensation image, determine the color compensation value of each channel corresponding to each pixel in the display panel, and perform color compensation on the display panel.

[0013] This application also discloses a display panel color compensation model construction device, including: The data processing module is used to acquire standard image data obtained by the color camera from standard color parts and form standard data; The image acquisition module is used to acquire the actual image data captured by the color camera and the corresponding grayscale display panel. The model training module is used to form a training dataset based on the standard data and the actual image data, and to train the pre-built compensation model framework using the training dataset to obtain the color compensation model.

[0014] This application also discloses a display panel color compensation device, comprising: The data acquisition module is used to acquire the actual color image data obtained by capturing the image displayed on the display panel to be compensated through a color camera; The data prediction module is used to input the actual color image data into the trained compensation model to obtain a color-compensated image; The color compensation module is used to determine the color compensation value of each channel corresponding to each pixel in the display panel based on the color compensation image, and to perform color compensation on the display panel.

[0015] This application discloses a display panel color compensation model. The model uses a three-channel separation network to extract image data from at least one of the R, G, and B channels individually. The image processing network includes at least one independent processing branch corresponding to each channel. This allows each processing branch to perform color compensation on the corresponding color channel's image data while maintaining low computational load, improving the accuracy of single-channel compensation and reducing compensation deviations caused by inter-channel feature interference. Finally, an image fusion network fuses the channel compensation images of each channel to obtain a color compensation image, achieving accurate color compensation for the display panel. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings: Figure 1 This is a flowchart illustrating the color compensation model for the display panel in an embodiment of this application. Figure 2 The results of R16 before and after compensation are shown in a specific embodiment of the color compensation model for the display panel in this application. Figure 3 The results of B16 before and after compensation are shown in a specific embodiment of the color compensation model for the display panel in this application. Figure 4 The results of G16 before and after compensation are shown in a specific embodiment of the color compensation model for the display panel in this application. Figure 5 This application provides a specific embodiment of the display panel color compensation model, including an image processing network. Figure 6 This is a flowchart of the display panel color compensation model training method according to an embodiment of this application; Figure 7 This is a flowchart of a display panel color compensation method according to an embodiment of this application; Figure 8 This is a block diagram of the display panel color compensation model construction device according to an embodiment of this application; Figure 9 This is a module structure diagram of the display panel color compensation device according to an embodiment of this application; Figure 10 This is a schematic diagram of the structure of a computer device used to implement embodiments of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. In the description of the embodiments of this application, unless otherwise stated, " / " means "or," for example, A / B can mean A or B; "and / or" in the text is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Furthermore, in the description of the embodiments of this application, "multiple" refers to two or more than two.

[0018] It should be understood that the terms "first," "second," etc., in the specification, claims, and drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.

[0019] In this application, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described in this application can be combined with other embodiments.

[0020] To address at least one of the problems existing in the prior art, according to one aspect of this application, this embodiment discloses a display panel color compensation model. For example... Figure 1 As shown in this embodiment, the model includes a three-channel separation network, an image processing network, and an image fusion network.

[0021] The three-channel separation network extracts color image data from the display image data of the display panel acquired by the color camera. The color image data includes channel image data of at least one of the R, G and B channels.

[0022] Specifically, the core of the three-channel separation network is to extract the image data of three monochrome channels—red (R), green (G), and blue (B)—from the raw display image data acquired by a color camera. The raw display image data acquired by a color camera is typically in Bayer (Bayer Color Filter Array) format. The raw data of a Bayer format image consists of monochrome pixels arranged according to a specific pattern, usually in 2×2 repeating units, containing 2 G pixels, 1 R pixel, and 1 B pixel. There are four arrangement patterns for the 2×2 repeating units in the Bayer format: RGGB (first row R, G; second row G, B), GRBG (first row G, R; second row B, G), GBRG (first row G, B; second row R, G), and BGGR (first row B, G; second row G, R). The three-channel separation network needs to identify the arrangement pattern and then use the Bayer interpolation de-mosaic algorithm to complete the separation and extraction of the R, G, and B channels. The Bayer interpolation demosaic algorithm first identifies the Bayer array type corresponding to each pixel in the image, performs interpolation calculations on missing color components, and supplements each pixel with complete R, G, and B color information. Finally, it separates the data into three single-channel image data: the R channel retains only the red component information of each pixel, the G channel retains only the green component information, and the B channel retains only the blue component information. The color camera capturing the display panel image data can capture the entire screen at once, analyzing the uniformity of brightness and color across the entire display panel, and locating abnormal areas such as bright spots, dark spots, and color spots in local pixels. The color camera only needs to capture the display panel once to extract image data containing the three monochrome channels (R, G, and B), eliminating the need for separate captures of each channel, thus shortening the color compensation time.

[0023] The image processing network includes at least one processing branch corresponding to each channel, used to perform color compensation on the channel image data and output the corresponding channel-compensated image.

[0024] In a specific example, the processing branch of the image processing network uses a single-channel image compensation network based on the U-Net framework. Independent processing branches are set up for the image data of the R, G, and B channels respectively. The processing branches are trained specifically based on the data features of the corresponding channels to obtain processing branches for different color compensations, thereby improving the compensation effect of image data of each color channel.

[0025] It should be noted that the network structure of each processing branch can be set to be the same or different and consistent. Those skilled in the art can select different network architectures for color channel image data compensation according to the actual situation to generate image processing networks. This application does not limit this.

[0026] In practical applications, for each display panel, white screen images at 32, 64, and 128 gray levels can be acquired in a darkroom environment. The acquired white screen images at each gray level are then subjected to R, G, and B primary color component separation to obtain the original image data of the corresponding gray level's R channel, G channel, and B channel. During training, the original image data of the R, G, and B channels are used as input, and the image data of channels without brightness or color unevenness defects are used as the supervised target. Through end-to-end direct learning, each processing branch learns the compensation rules for the defects of the corresponding channel. The brightness, contrast, and Gamma curve of the image are dynamically adjusted through the compensation algorithm. At the same time, training is conducted on low gray level samples in the training data, and the pixel parameters at low gray levels are adjusted through the compensation algorithm to improve the details in dark fields and the smoothness of gray level transitions. Finally, the R channel compensated image, G channel compensated image, and B channel compensated image after eliminating the defects of the corresponding channel are output.

[0027] It should be noted that the acquisition process of the defect-free channel image data is as follows: A high-quality reference camera is first calibrated in a darkroom environment using a uniform white board and gray card as references. Multiple images with varying positions and angles are taken using a defect-free display panel as the subject under different exposure, light source, and temperature conditions. Simultaneously, data enhancement algorithms such as brightness, contrast, and white balance perturbation are used to eliminate the effects of imaging non-uniformity of the camera sensor itself, expand the training dataset, and simulate different scenarios in actual applications, thereby enhancing the generalization of the model. Finally, defect-free channel image data with uniform brightness and standard color is formed, which serves as the supervised target for training the image processing network.

[0028] An image fusion network is used to obtain a color compensation image from the channel compensation image obtained by the image processing network, and to perform color compensation on the display panel based on the color compensation image.

[0029] In a specific example, such as Figures 2-4 As shown, Figure 2 The images show the results before and after R16 compensation. Before compensation, the dark details in the R16 channel are blurred and there are local brightness banding. After compensation, the dark textures are clear and the grayscale transitions are continuous. Figure 3 The images show the results before and after B16 compensation. Before compensation, the B16 channel exhibited localized color shift; after compensation, the color coordinates returned to the standard range. Figure 4 The images show the results before and after G16 compensation. Before compensation, the local brightness of the G16 channel was uneven, while after compensation, the brightness distribution was uniform.

[0030] The display panel color compensation model of this application extracts image data of at least one of the R, G, and B channels individually through a three-channel separation network. The image processing network includes at least one independent processing branch for each channel, which enables each processing branch to perform color compensation on the corresponding color channel image data while ensuring low computational load. This improves the accuracy of single-channel compensation and reduces compensation deviations caused by inter-channel feature interference. Finally, the image fusion network fuses the channel compensation images of each channel to obtain the color compensation image, achieving accurate color compensation for the display panel.

[0031] In an optional implementation, the image fusion network may include a built-in color balance sub-network. The color balance sub-network is used to calculate the adjustment coefficients for the brightness and chromaticity of each pixel in each channel compensated image obtained by the image processing network, correct the pixel value of each pixel in each channel compensated image based on the adjustment coefficients, and perform image fusion on the corrected channel compensated images to obtain a color compensated image.

[0032] Specifically, the image fusion network receives the compensated R-channel, G-channel, and B-channel images. First, through a built-in color balance sub-network, it analyzes the brightness distribution and color coordinate parameters of each channel, calculates the brightness and chromaticity adjustment coefficients between channels, and corrects the pixel values ​​of each channel's compensated image point by point. This eliminates color deviations such as brightness imbalance and color coordinate shift caused by independent channel compensation. The R, G, and B channel compensated images are then input into the back-end weighted fusion module of the fusion network to generate a color compensated image. Using this color compensated image as a reference, it is compared with the display panel captured by the color camera to obtain compensation values. Based on these compensation values, the pixels of the compensated display panel are made consistent with the pixels in the color compensated image. The compensation values ​​are then loaded into the display driver IC, which adjusts the voltage and current of the panel's pixel circuits pixel by pixel or region by region, ultimately eliminating the brightness and chromaticity defects of the display panel.

[0033] It should be noted that after the R, G, and B channel image data are compensated independently, differences in color defect characteristics of each channel, such as local overbrightness in the R channel and local underbrightness in the G channel, will result in new brightness unevenness in the superimposed grayscale image. In a specific example, the adjustment coefficients for brightness and chromaticity between channels can be determined as follows: The color balance subnetwork first extracts the brightness values ​​of the compensated images of each channel, calculates the brightness weight coefficient of each channel based on the target brightness, and adjusts the brightness values ​​of the compensated images of each channel so that the adjusted brightness values ​​of each channel are close to the target brightness, thus making the brightness of the final fused color compensated image uniform. At the same time, in order to eliminate local deviations in color coordinates, the color balance subnetwork pre-sets a maximum allowable deviation threshold for color coordinates as a color gamut constraint. It extracts the compensated color coordinates of each channel and calculates their deviation value from the target color coordinates. If the deviation of a certain channel exceeds the threshold, the pixel value of that channel is corrected, and then the color coordinates of the compensation results of each channel are fine-tuned so that the pixel color coordinates of the fused image fall within the target color gamut, ensuring consistent chromaticity at the same gray level. Those skilled in the art can adjust the brightness adjustment coefficient and the chromaticity adjustment coefficient according to actual needs, and this application does not limit this.

[0034] In an optional implementation, the image processing network includes convolutional layers, bottleneck layers, and skip layers. The convolutional layers adjust the number of channels and resolution of the channel image data; the bottleneck layers perform feature amplification on the channel image data adjusted by the convolutional layers; and the skip layers connect the image data output by the bottleneck layers with the image data output by the convolutional layers to obtain a channel-compensated image.

[0035] Specifically, the convolutional layer adjusts the number of channels and resolution of the input R, G, and B channel original image data. While adjusting the number of channels to extract image features, it controls the downsampling degree to avoid over-compression and loss of details, ultimately outputting image data containing shallow features such as the original brightness distribution and texture contours. The bottleneck layer performs feature amplification processing on the R, G, and B channel image data adjusted by the convolutional layer, preserving image details such as single-pixel color defects and low-grayscale dark areas, enabling the model to identify and learn the compensation rules for single-pixel and other detail defects, ultimately outputting deep feature image data containing defect locations. The skip layer fuses the deep feature image data output by the bottleneck layer with the shallow feature image data output by the convolutional layer, eliminating color defects while avoiding loss of image details during compensation, ultimately outputting channel-compensated images for each of the R, G, and B channels with eliminated color defects and complete image details.

[0036] In an optional implementation, the convolutional layer includes: a first convolutional layer, a pooling layer, and a second convolutional layer. The first convolutional layer expands the number of channels in the channel image data; the pooling layer compresses the expanded channel image data; and the second convolutional layer adjusts the number of image channels in the compressed channel image data.

[0037] In a specific example, such as Figure 5 As shown, the input image data of the channel is 64×64×3, the resolution is 64×64, and the number of channels is 3. After convolution in the first convolutional layer L1, the parameters change to 128×128×500, the resolution is magnified to 128×128, and the number of channels is expanded to 500. Image data containing shallow features such as the original brightness distribution and texture contours of the image is extracted. Only one pooling layer L2 is retained in the middle of the convolutional layer. After max pooling calculation, the number of channels remains unchanged, and the parameters change to 64×64×500. Entering the second convolutional layer L3, the parameters change to 256×256×128, the resolution is magnified by 4 times to 256×256, and the number of channels is reduced to 128. Deep features such as the location of defects are extracted. The bottleneck layer L4 performs feature amplification on the channel image data adjusted after the convolutional layer, preserving image details. It receives 256×256×128 features from the output of the convolutional layer, performs feature amplification processing, and transposes the convolution to increase the resolution, outputting 1024×1024×32 features. The resolution is increased by 4 times to 1024×1024, the number of channels is reduced to 32, and more image details are preserved.

[0038] In an optional implementation, the skip layer includes a first skip layer L5, a second skip layer L6, and a third skip layer L7. The first skip layer L5 fuses the image data output from the bottleneck layer L4 with the image data output from the second convolutional layer L3 to obtain a first connected image; the second skip layer L6 fuses the first connected image output from the first skip layer L5 with the image data output from the first convolutional layer L1 to obtain a second connected image; and the third skip layer L7 fuses the second connected image output from the second skip layer with the channel image data to obtain a channel-compensated image.

[0039] In a specific example, such as Figure 5As shown, the skip layer connects the image data output from the bottleneck layer L4 with the image data output from the convolutional layer to obtain a channel-compensated image. The first skip layer L5 performs feature fusion on the 1024×1024×32 image data output from the bottleneck layer L4 and the 256×256×128 image data output from the second convolutional layer L3. Before fusion, the image is transposed by upsampling and modulation operations on the decoder to adjust the size, reducing the resolution from 1024×1024 to 256×256 by a factor of 4, and expanding the number of channels to 256, outputting a 256×256×128 first connected image. The second skip layer... Layer L6 performs feature fusion between the 256×256×128 image data output from the first skip layer L5 and the 128×128×500 image data output from the convolutional layer. After upsampling to adjust the size and number of channels, it outputs a second connected image of 128×128×256. The third skip layer L7 performs feature fusion between the 128×128×256 image data output from the second skip layer L6 and the input 64×64×3 channel image data. After upsampling to restore the input resolution and adjusting the number of channels to 512, it outputs a channel-compensated image of 64×64×512.

[0040] According to another aspect of this application, such as Figure 6 As shown, this embodiment also discloses a method for training a display panel color compensation model, including: S100: Acquire standard image data obtained by a color camera from a standard color component, and form standard data.

[0041] Specifically, the standard color component is a standard color chart containing consistent and accurate color parameters. When shooting the standard color component, the color camera needs to be in a darkroom environment. The captured standard image data contains color information from the R, G, and B channels, forming standard data.

[0042] S200: Acquires the actual image data captured by the color camera and displayed on the corresponding grayscale display panel.

[0043] Specifically, the grayscale display panel displays images according to preset grayscale parameters, which include multiple grayscale levels. When the color camera captures the image displayed on the grayscale display panel, the shooting environment is consistent with that of a standard color image. The captured actual image data undergoes channel separation processing to extract the actual image data of the R channel, G channel, and B channel. The R channel image data reflects the actual display of the R channel on the display panel, the G channel image data reflects the actual display of the G channel, and the B channel image data reflects the actual display of the B channel.

[0044] S300: A training dataset is formed based on the standard data and the actual image data, and the color compensation model is obtained by training the pre-built compensation model framework using the training dataset.

[0045] Specifically, standard data is used as the supervised target, and actual image data for each channel is used as the training input. For the R channel, the actual image data of the R channel is used as input, and the corresponding R channel color information in the standard data is used as the target; for the G channel, the actual image data of the G channel is used as input, and the corresponding G channel color information in the standard data is used as the target; for the B channel, the actual image data of the B channel is used as input, and the corresponding B channel color information in the standard data is used as the target, thus constructing the training dataset. The compensation model framework is the image processing network described above. This network includes convolutional layers, bottleneck layers, and skip layers, which can process the input channel image data and output channel-compensated images. The input data from the training dataset is input into the compensation model framework to obtain the model's predicted output. The loss value between the predicted output and the corresponding target output in the training dataset is calculated. Based on the loss value, the parameters in the compensation model framework are adjusted through the backpropagation algorithm until the loss value reaches a preset threshold or the training reaches a preset number of iterations, thus obtaining the trained compensation model.

[0046] In an optional implementation, the compensation model framework includes convolutional layers, bottleneck layers, and skip layers; The convolutional layer comprises a first convolutional layer, a pooling layer, and a second convolutional layer. The skip layer comprises a first skip layer, a second skip layer, and a third skip layer.

[0047] Since the principle of solving the problem in this embodiment is similar to that of the above model, this embodiment can refer to the implementation of the model, and will not be described again here.

[0048] In an optional implementation, the loss function of the compensation model includes mean absolute error and mean squared error.

[0049] Specifically, the loss function quantifies the deviation between the channel-compensated image output by the model and the standard data. Mean absolute error (MAE) reflects the difference in the global grayscale distribution of the image, while mean squared error (MSE) amplifies the weight of local defects. Combining MAE and MSE allows for the evaluation of the model's effectiveness in compensating for color deviation, brightness unevenness, and other issues. MAE measures the overall grayscale deviation between the channel-compensated image output by the model and the standard data; the calculation formula is as follows: Where N is the total number of pixels in a single-channel image. To compensate for the grayscale value of the k-th pixel in the model output, This represents the grayscale value of the k-th pixel in the standard data. Mean Squared Error (MSE) is used to amplify the weight of areas with significant local differences, improving the repair of local defects.

[0050] The formula for calculating the mean square error is: Where N is the total number of pixels in a single-channel image. To compensate for the grayscale value of the k-th pixel in the model output, Let be the grayscale value of the k-th pixel in the standard data. The smoothing constraint term S is the sum of the grayscale differences between adjacent pixels, S = ∑∣I(i,j) I(i±1,j)∣+∑∣I(i,j) I(i,j±1)∣。 The total loss function is L=α×MAE+β×MSE+γ×S, where α, β, and γ are the weight coefficients of MAE, MSE, and the smoothing constraint term S, respectively, and satisfy α+β+γ=1. It can be dynamically adjusted according to the defects of the display panel. If there are many local defects, β can be increased. If the overall image needs to be smoothed, γ can be increased.

[0051] In a specific example, in step S300, the actual image data of each of the R, G, and B channels are input into the compensation model to obtain the initial compensation image of each channel. The total loss is calculated according to L=α×MAE+β×MSE+γ×S. Based on the backpropagation algorithm, the parameters of the model such as the convolution kernel weights and bias terms are adjusted according to the total loss to minimize the total loss. Finally, the above steps are repeated until the loss value converges, training is stopped, and the final compensation model is obtained.

[0052] The compensation model framework also includes smoothing constraints for the color compensation image, and the step of training the pre-constructed compensation model framework using the training dataset to obtain the color compensation model includes: The training dataset is input into a pre-built compensation model framework for multiple iterative training until a preset training termination condition is reached to obtain the color compensation model. In each iterative training process, the model parameters of the compensation model framework are adjusted based on the smoothing constraint condition.

[0053] In this method, a smoothing constraint is added to limit the grayscale value difference between adjacent pixels in the channel compensation image. Let the coordinates of any pixel in the compensation image be (i, j), and its grayscale value be denoted as I(i, j). Centered on pixel (i, j), select its four adjacent pixels (i+1, j), ... 1, j), (i, j+1), (i, j) 1) As a computational object. Calculate the absolute difference in grayscale values ​​between the center pixel and each of its neighboring pixels, i.e., |I(i,j) I(i+1, j)∣、∣I(i, j) I(i 1,j)∣、∣I(i,j) I(i, j+1)∣、∣I(i, j) I(i,j) 1) |. The sum of the absolute differences mentioned above is limited to a preset threshold T, i.e., ∑ |I(i,j) I(i±1,j)∣+∑∣I(i,j) I(i,j±1)∣≤T. The threshold T needs to be set according to parameters such as the noise level of the display panel.

[0054] According to another aspect of this application, such as Figure 7 As shown, this embodiment also discloses a display panel color compensation method, including: S111: Obtain the actual color image data obtained from the image displayed on the display panel to be compensated, which was captured by the color camera.

[0055] S112: Input the actual color image data into the trained compensation model to obtain a color compensation image.

[0056] S113: Determine the color compensation value of each channel corresponding to each pixel in the display panel based on the color compensation image, and perform color compensation on the display panel.

[0057] Since the principle behind this method is similar to that of the methods described above, the implementation of this method can be found in the section on method implementation, and will not be repeated here.

[0058] Based on the same principle, such as Figure 8 As shown, this embodiment also discloses a display panel color compensation model construction device, including: Data processing module 10: Used to acquire standard image data obtained by the color camera from the standard color part, and form standard data.

[0059] Image acquisition module 20: Used to acquire the actual image data obtained by capturing the image displayed on the grayscale display panel corresponding to the color camera.

[0060] Model training module 30: used to form a training dataset based on the standard data and the actual image data, and to train the pre-built compensation model framework using the training dataset to obtain the color compensation model.

[0061] Since the principle by which this device solves the problem is similar to the methods described above, the implementation of this device can be found in the implementation of the methods, and will not be repeated here.

[0062] Based on the same principle, such as Figure 9As shown, this embodiment also discloses a display panel color compensation device, including a data acquisition module 11, a data prediction module 12, and a color compensation module 13.

[0063] The data acquisition module 11 is used to acquire the actual color image data obtained by acquiring the image displayed on the display panel to be compensated through the color camera.

[0064] The data prediction module 12 is used to input the actual color image data into the trained compensation model to obtain a color compensation image.

[0065] The color compensation module 13 is used to determine the color compensation value of each channel corresponding to each pixel in the display panel based on the color compensation image, and to perform color compensation on the display panel.

[0066] Since the principle by which this device solves the problem is similar to the methods described above, the implementation of this device can be found in the implementation of the methods, and will not be repeated here.

[0067] Based on the same principle, this application also discloses a display panel color compensation system. The system includes a color camera, a display panel color compensation model construction device as described in this embodiment, and a display panel color compensation device as described in this embodiment.

[0068] Since the principle by which this system solves the problem is similar to the methods and apparatus described above, the implementation of this system can be found in the implementation of the methods and apparatus, and will not be repeated here.

[0069] This application also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described method.

[0070] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0071] Those skilled in the art will understand that the embodiments of this application can be provided as methods, systems, or computer programs, producing the systems, apparatuses, modules, or units described in the above embodiments. Specifically, they can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer device; specifically, a computer device can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.

[0072] In a typical example, the computer device specifically includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the method executed by the client as described above, or the method executed by the server as described above.

[0073] The following is for reference. Figure 10 It shows a schematic diagram of the structure of a computer programmable logic device 600 suitable for implementing embodiments of the present application.

[0074] like Figure 10 As shown, the computer programmable logic device 600 includes a central processing unit (CPU) 601, which can perform various appropriate tasks and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage section 608 into a random access memory (RAM) 603. The RAM 603 also stores various programs and data required for the operation of the computer programmable logic device 600. The CPU 601, ROM 602, and RAM 603 are interconnected via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0075] The following components are connected to I / O interface 605: an input section 606 including a keyboard, mouse, etc.; an output section 607 including a cathode ray tube (CRT), liquid crystal feedback (LCD), etc., and speakers, etc.; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a LAN card, modem, etc. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to I / O interface 605 as needed. A removable medium 611, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 610 as needed so that computer programs read from it can be installed in storage section 608 as needed.

[0076] In particular, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program tangibly embodied on a machine-readable medium, the computer program including program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 609, and / or installed from removable medium 611.

[0077] Computer-readable media include both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage, programmable logic devices, or any other non-transferable media that can be used to store information accessible by a computer programmable logic device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0078] For ease of description, the above devices are described separately by function as various units. Of course, in implementing this application, the functions of each unit can be implemented in one or more software and / or hardware.

[0079] This application is described with reference to flowchart illustrations and / or block diagrams of methods, programmable logic devices (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing programmable logic device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing programmable logic device, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0080] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing programmable logic device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0081] These computer program instructions can also be loaded onto a computer or other programmable data processing logic device, causing a series of operational steps to be executed on the computer or other programmable logic device to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable logic device for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0082] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or programmable logic device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or programmable logic device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or programmable logic device that includes said element.

[0083] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0084] This application can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. This application can also be practiced in distributed computing environments where tasks are performed by remotely processed programmable logic devices connected via a communication network. In a distributed computing environment, program modules can reside in local and remote computer storage media, including storage for programmable logic devices.

[0085] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0086] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A color compensation model for a display panel, characterized in that, include: A three-channel separation network extracts color image data from the display image data of the display panel acquired by the color camera. The color image data includes channel image data of at least one of the R, G, and B channels. An image processing network includes at least one processing branch corresponding to each channel, used to perform color compensation on the channel image data and output the corresponding channel-compensated image; An image fusion network is used to obtain a color compensation image from the channel compensation image obtained by the image processing network, and to perform color compensation on the display panel based on the color compensation image.

2. The display panel color compensation model according to claim 1, characterized in that, The processing branches of the image processing network include convolutional layers, bottleneck layers, and skip layers; The convolutional layer is used to adjust the number of channels and resolution of the channel image data; The bottleneck layer performs feature amplification on the channel image data adjusted by the convolutional layer; The skip layer connects the image data output by the bottleneck layer with the image data output by the convolutional layer to obtain a channel-compensated image.

3. The display panel color compensation model according to claim 2, characterized in that, The convolutional layer includes: a first convolutional layer, a pooling layer, and a second convolutional layer; The first convolutional layer expands the number of channels in the channel image data; The pooling layer performs image compression on the channel image data after the channel number is expanded; The second convolutional layer adjusts the number of image channels in the compressed image data.

4. The display panel color compensation model according to claim 3, characterized in that, The skip layers include a first skip layer, a second skip layer, and a third skip layer; The first skip layer fuses the image data output by the bottleneck layer with the image data output by the second convolutional layer to obtain a first connected image; The second skip layer fuses the first connected image output by the first skip layer with the image data output by the first convolutional layer to obtain the second connected image; The third skip layer performs feature fusion between the second connected image output by the second skip layer and the channel image data to obtain a channel-compensated image.

5. The display panel color compensation model according to claim 1, characterized in that, The image fusion network includes a color balance sub-network, which is used to calculate the adjustment coefficients of brightness and chromaticity of each pixel in each channel compensation image obtained by the image processing network, correct the pixel value of each pixel in each channel compensation image based on the adjustment coefficients, and perform image fusion on the channel compensation images after pixel value correction to obtain a color compensation image.

6. A training method for a display panel color compensation model as described in any one of claims 1-5, characterized in that, The method includes: Acquire standard image data obtained by a color camera from standard color parts, and form standard data; Acquire the actual image data captured by the color camera and displayed on the corresponding grayscale display panel; A training dataset is formed based on the standard data and the actual image data. The color compensation model is obtained by training the pre-built compensation model framework using the training dataset.

7. The training method for the display panel color compensation model according to claim 6, characterized in that, The compensation model framework also includes smoothing constraints for the color compensation image, and the step of training the pre-constructed compensation model framework using the training dataset to obtain the color compensation model includes: The training dataset is input into a pre-built compensation model framework for multiple iterative training until a preset training termination condition is reached to obtain the color compensation model. In each iterative training process, the model parameters of the compensation model framework are adjusted based on the smoothing constraint condition.

8. A display panel color compensation method employing the display panel color compensation model as described in any one of claims 1-5, characterized in that, include: Obtain the actual color image data from the image displayed on the display panel to be compensated, captured by a color camera; The actual color image data is input into the trained compensation model to obtain the color compensation image; Based on the color compensation image, determine the color compensation value of each channel corresponding to each pixel in the display panel, and perform color compensation on the display panel.

9. A display panel color compensation model construction device, characterized in that, include: The data processing module is used to acquire standard image data obtained by the color camera from standard color parts and form standard data; The image acquisition module is used to acquire the actual image data captured by the color camera and the corresponding grayscale display panel. The model training module is used to form a training dataset based on the standard data and the actual image data, and to train the pre-built compensation model framework using the training dataset to obtain the color compensation model.

10. A display panel color compensation device, characterized in that, include: The data acquisition module is used to acquire the actual color image data obtained by capturing the image displayed on the display panel to be compensated through a color camera; The data prediction module is used to input the actual color image data into the trained compensation model to obtain a color-compensated image; The color compensation module is used to determine the color compensation value of each channel corresponding to each pixel in the display panel based on the color compensation image, and to perform color compensation on the display panel.

Images