A quality evaluation method, device, equipment, medium and product of a display device

Abstract

The application provides a quality evaluation method, device and equipment of a display device, a medium and a product, and relates to the field of computer vision. The method comprises the following steps: acquiring a first image, wherein the first image comprises at least a color separation light spot image of a display device; acquiring a plurality of color separation features based on the first image, wherein the color separation features are used to represent the color dispersion degree of the color separation light spot image in different directions; inputting the plurality of color separation features into a first prediction model to obtain first grading data; inputting the first image into a second prediction model to obtain second grading data; and obtaining a target grading result based on the first grading data and the second grading data, wherein the target grading result is used to represent the color separation grade of the display device.

Description

Technical Field

[0001] This application relates to the field of computer vision, and more specifically, to a method, apparatus, device, medium, and product for quality assessment of display devices. Background Technology

[0002] Color separation grading is a crucial aspect of display quality assessment. Its goal is to score the degree of color separation in a given image or set of images reflected from the display screen, based on how ambient light is reflected. Color separation can reduce the display's ambient contrast, especially in the presence of point light sources, producing noticeable color separation spots that negatively impact the viewer's experience.

[0003] However, existing methods for classifying color separation phenomena in displays rely on manually constructed features or deep convolutional neural networks to obtain classification results. These methods suffer from poor accuracy and can easily lead to inconsistent predictions for different display models with the same color separation level. Therefore, improving the classification accuracy of display color separation classification tasks has become a pressing issue in this field. Summary of the Invention

[0004] This application provides a method, apparatus, device, medium, and product for quality assessment of display devices, aiming to solve the problem of how to improve the grading accuracy of color separation and grading tasks for display screens.

[0005] A first aspect of this application provides a method for quality assessment of a display device, the method comprising:

[0006] Acquire a first image, the first image including at least a color separation spot image of the display device;

[0007] Based on the first image, multiple color separation features are obtained, which are used to characterize the degree of color dispersion of the color-separated spot image in different directions;

[0008] The multiple color separation features are input into the first prediction model to obtain the first graded data;

[0009] The first image is input into the second prediction model to obtain the second classification data;

[0010] Based on the first grading data and the second grading data, a target grading result is obtained, which is used to characterize the color separation level of the display device.

[0011] In one optional implementation, obtaining multiple color separation features based on the first image includes:

[0012] The first image is subjected to image normalization processing to convert white pixels in the first image into black pixels, thereby obtaining the second image;

[0013] Based on multiple pixels in different directions in the second image and the center pixel of the color separation spot, the multiple color separation features are obtained respectively.

[0014] In one optional implementation, the step of performing image normalization processing on the first image to convert white pixels in the first image into black pixels to obtain the second image includes:

[0015] Within the target color space, obtain the pixel value of each pixel in the first image corresponding to each color channel;

[0016] Based on the pixel values ​​corresponding to each pixel in each color channel, a bias pixel value is determined for each pixel, wherein the bias pixel value is the minimum value among the pixel values ​​corresponding to each pixel in each color channel.

[0017] Based on the bias pixel value corresponding to each pixel, the pixel values ​​corresponding to each pixel in each color channel are processed to obtain the second image.

[0018] In one optional implementation, the step of obtaining the multiple color separation features based on multiple pixels in different directions in the second image and the center pixel of the color separation spot includes:

[0019] Obtain the pixel values ​​of multiple pixels in the second image in the target direction. The target direction is any one of the first direction, the second direction, the third direction, and the fourth direction. The target direction passes through the center of the color separation spot. The first direction and the second direction extend in opposite directions in the vertical direction, and the third direction and the fourth direction extend in opposite directions in the horizontal direction.

[0020] The first and second moments of the multiple pixels in the target direction and the center pixel of the color separation spot are obtained as the color separation features of the second image in the target direction.

[0021] In one optional implementation, the first prediction model is a random forest model, and the step of inputting the plurality of color separation features into the first prediction model to obtain the first hierarchical data includes:

[0022] The multiple color separation features are preprocessed;

[0023] The multiple color separation features are input into the first prediction model, and the multiple color separation features are processed based on the multiple decision trees in the first prediction model to obtain multiple classification results;

[0024] The multiple classification results are processed based on a voting mechanism to obtain the first classification data, which is used to characterize the probability distribution of the display device belonging to different color separation levels.

[0025] In one optional implementation, the second prediction model is a recurrent convolutional residual network, and the step of inputting the first image into the second prediction model to obtain the second hierarchical data includes:

[0026] Perform polar coordinate transformation on the first image to obtain the third image;

[0027] The third image is input into the second prediction model, and the third image is subjected to cyclic convolution processing based on the second prediction model to obtain the first feature, which is the image feature corresponding to the third image with translation invariance;

[0028] The first feature is processed based on the second prediction model to obtain the second classification data, which is used to characterize the probability distribution of the display device belonging to different color separation levels.

[0029] In one optional implementation, obtaining the target classification result based on the first classification data and the second classification data includes:

[0030] Based on the first graded data, determine the first confidence level corresponding to the prediction results of different color separation levels in the first graded data;

[0031] Based on the second graded data, determine the second confidence level corresponding to the prediction results of different color separation levels in the second graded data;

[0032] The first and second graded data are processed based on the first and second confidence levels to obtain the target graded result.

[0033] In one optional implementation, the step of processing the first classification data and the second classification data based on the first confidence level and the second confidence level to obtain the target classification result includes:

[0034] Using the first confidence level as the weight of the prediction result corresponding to the first confidence level in the first graded data and the second confidence level as the weight of the prediction result corresponding to the second confidence level in the second graded data, the prediction results corresponding to the first confidence level in the first graded data and the prediction results corresponding to the second confidence level in the second graded data for the same color separation level are fused to obtain the graded result for each color separation level.

[0035] The maximum value among all the color separation level classification results is taken as the target classification result.

[0036] In one optional implementation, acquiring the first image includes:

[0037] The base and image acquisition device are calibrated based on the calibration board;

[0038] The display device is placed on the base, and images of the display device are periodically acquired based on a moving point light source to obtain multiple initial images. Each initial image includes at least a color separation spot image of the display device, and the color separation spot images in different initial images have different orientations in the display device.

[0039] The initial image is preprocessed to obtain the first image.

[0040] In one optional implementation, preprocessing the initial image to obtain the first image includes:

[0041] Based on a standard plane, an affine transformation is performed on the initial image to obtain a first preprocessed image;

[0042] Edge detection is performed on the first preprocessed image to obtain an image of the region in the first preprocessed image that contains the edge of the display device, which is then used as the second preprocessed image.

[0043] Brightness detection is performed on the second preprocessed image to obtain the region image in the second preprocessed image whose brightness is higher than a preset brightness threshold, which is then used as the first image.

[0044] In one optional implementation, the step of performing an affine transformation on the initial image based on a standard plane to obtain a first preprocessed image includes:

[0045] Based on the calibration plate, calibration results are generated for the base and the image acquisition device;

[0046] Based on the calibration results, a homography matrix is ​​constructed between the plane where the display device is located and the standard plane;

[0047] Based on the homography matrix, an affine transformation is performed on the initial image to obtain the first preprocessed image.

[0048] In an optional implementation, the step of performing edge detection on the first preprocessed image to obtain a region image containing the edge of the display device in the first preprocessed image, as a second preprocessed image, includes:

[0049] Based on the edge detection algorithm, the edge information of the display device in the first preprocessed image is obtained as the first edge information;

[0050] Based on the first edge information, the first preprocessed image is processed to obtain a separated image;

[0051] Based on the defect detection algorithm, the separated image is iteratively detected to obtain edge information located outside the region image corresponding to the display device in the separated image, which is used as the second edge information;

[0052] Based on the second edge information, the separated image is processed to obtain the second preprocessed image.

[0053] A second aspect of this application provides a quality assessment apparatus for a display device, the apparatus comprising:

[0054] An image acquisition module is used to acquire a first image, wherein the first image includes at least a color separation spot image of the display device;

[0055] The feature extraction module is used to obtain multiple color separation features based on the first image, wherein the color separation features are used to characterize the degree of color dispersion of the color separation spot image in different directions;

[0056] The first prediction module is used to input the multiple color separation features into the first prediction model to obtain the first graded data;

[0057] The second prediction module is used to input the first image into the second prediction model to obtain the second hierarchical data;

[0058] A grading module is used to obtain a target grading result based on the first grading data and the second grading data, wherein the target grading result is used to characterize the color separation level of the display device.

[0059] In one optional implementation, the feature extraction module includes:

[0060] The image normalization submodule is used to perform image normalization processing on the first image to convert white pixels in the first image into black pixels, thereby obtaining the second image.

[0061] The feature extraction submodule is used to obtain the multiple color separation features based on multiple pixels in different directions in the second image and the center pixel of the color separation spot.

[0062] In one optional implementation, the image normalization submodule includes:

[0063] The pixel value determination unit is used to obtain the pixel value of each pixel in the first image in each color channel within the target color space;

[0064] The offset pixel value unit is used to determine the offset pixel value corresponding to each pixel based on the pixel value corresponding to each pixel in each color channel, wherein the offset pixel value is the minimum value among the pixel values ​​corresponding to each pixel in each color channel.

[0065] An image normalization unit is used to process the pixel values ​​corresponding to each pixel in each color channel based on the bias pixel value corresponding to each pixel to obtain the second image.

[0066] In one optional implementation, the feature extraction submodule includes:

[0067] A target direction pixel value acquisition unit is used to acquire the pixel values ​​of multiple pixels of the second image in a target direction. The target direction is any one of a first direction, a second direction, a third direction, and a fourth direction. The target direction passes through the center of the color separation spot. The first direction and the second direction extend in opposite directions in the vertical direction, and the third direction and the fourth direction extend in opposite directions in the horizontal direction.

[0068] The feature extraction unit is used to obtain the first and second moments of the multiple pixels in the target direction and the center pixel of the color separation spot, as the color separation feature of the second image in the target direction.

[0069] In one optional implementation, the first prediction module includes:

[0070] The preprocessing submodule is used to preprocess the multiple color separation features;

[0071] The first classification submodule is used to input the multiple color separation features into the first prediction model, and process the multiple color separation features based on multiple decision trees in the first prediction model to obtain multiple classification results;

[0072] The first graded data submodule is used to process the multiple classification results based on a voting mechanism to obtain the first graded data, which is used to characterize the probability distribution of the display device belonging to different color separation levels.

[0073] In one optional implementation, the second prediction module includes:

[0074] The coordinate transformation submodule is used to perform polar coordinate transformation on the first image to obtain the third image;

[0075] The convolution submodule is used to input the third image into the second prediction model, and perform cyclic convolution processing on the third image based on the second prediction model to obtain a first feature, wherein the first feature is an image feature corresponding to the third image with translation invariance;

[0076] The second graded data submodule is used to process the first feature based on the second prediction model to obtain the second graded data, which is used to characterize the probability distribution of the display device belonging to different color separation levels.

[0077] In one optional implementation, the hierarchical module includes:

[0078] The first confidence level submodule is used to determine the first confidence level corresponding to the prediction results of different color separation levels in the first graded data based on the first graded data.

[0079] The second confidence level submodule is used to determine the second confidence level corresponding to the prediction results of different color separation levels in the second classification data based on the second classification data.

[0080] The grading submodule is used to process the first grading data and the second grading data based on the first confidence level and the second confidence level to obtain the target grading result.

[0081] In one optional implementation, the hierarchical submodule includes:

[0082] The fusion unit is used to fuse the prediction results corresponding to the first confidence level in the first graded data and the prediction results corresponding to the second confidence level in the second graded data for categories with the same color separation level, respectively, to obtain the grading result for each color separation level.

[0083] A grading unit is used to take the maximum value among the grading results of all color separation levels as the target grading result.

[0084] In one optional implementation, the image acquisition module includes:

[0085] The calibration submodule is used to calibrate the base and image acquisition device based on the calibration board;

[0086] The image acquisition submodule is used to place the display device on the base and periodically acquire images of the display device based on a moving point light source to obtain multiple initial images. The initial images include at least the color separation spot image of the display device, and the color separation spot images in different initial images have different orientations in the display device.

[0087] The image acquisition submodule is used to preprocess the initial image to obtain the first image.

[0088] In one optional implementation, the image acquisition submodule includes:

[0089] An affine transformation unit is used to perform an affine transformation on the initial image based on a standard plane to obtain a first preprocessed image;

[0090] An edge detection unit is used to perform edge detection on the first preprocessed image and obtain an image of the region containing the edge of the display device in the first preprocessed image, as a second preprocessed image.

[0091] The image acquisition unit is used to perform brightness detection on the second preprocessed image and acquire the region image in the second preprocessed image whose brightness is higher than a preset brightness threshold, as the first image.

[0092] In one optional embodiment, the affine transformation unit includes:

[0093] The calibration subunit is used to generate calibration results for the base and the image acquisition device based on the calibration plate;

[0094] A homography matrix construction subunit is used to construct a homography matrix between the plane where the display device is located and the standard plane based on the calibration results;

[0095] The first preprocessing image subunit is used to perform an affine transformation on the initial image based on the homography matrix to obtain the first preprocessed image.

[0096] In one optional implementation, the edge detection unit includes:

[0097] The first edge information subunit is used to obtain the edge information of the display device in the first preprocessed image based on the edge detection algorithm, as the first edge information;

[0098] A separate image subunit is used to process the first preprocessed image based on the first edge information to obtain a separate image;

[0099] The second edge information subunit is used to iteratively detect the separated image based on the defect detection algorithm to obtain edge information in the separated image that is outside the region image corresponding to the display device, as the second edge information;

[0100] The second preprocessing image subunit is used to process the separated image based on the second edge information to obtain the second preprocessed image.

[0101] A third aspect of this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps in the quality evaluation method for a display device according to any one of the first aspects of this application.

[0102] A fourth aspect of this application provides a computer-readable storage medium having a computer program / instructions stored thereon, which, when executed by a processor, implements the steps in the quality assessment method for a display device as described in any one of the first aspects of this application.

[0103] A fifth aspect of this application provides a computer program product, including a computer program / instructions, which, when executed by a processor, implement the steps in the quality evaluation method for a display device as described in any one of the first aspects of this application.

[0104] Beneficial effects:

[0105] This application provides a method, apparatus, device, medium, and product for quality assessment of a display device. The method includes: acquiring a first image, the first image including at least a color separation spot image of the display device; acquiring multiple color separation features based on the first image, the color separation features being used to characterize the degree of color dispersion of the color separation spot image in different directions; inputting the multiple color separation features into a first prediction model to obtain first grading data; inputting the first image into a second prediction model to obtain second grading data; and obtaining a target grading result based on the first grading data and the second grading data, the target grading result being used to characterize the color separation level of the display device. This application obtains the first grading data by processing the color separation features characterizing the degree of color dispersion, ensuring the interpretability of the grading result; and obtains the second grading data based on the second prediction model, effectively improving the prediction quality of the grading result, thereby giving the grading result obtained based on the first grading data and the second grading data higher grading accuracy and effectively improving the performance of the model in performing the color separation grading task. Attached Figure Description

[0106] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0107] Figure 1 This is a flowchart illustrating the steps of a quality assessment method for a display device according to an embodiment of this application;

[0108] Figure 2 This is a flowchart illustrating the acquisition of a first image in a quality assessment method for a display device according to an embodiment of this application;

[0109] Figure 3 This is a schematic diagram illustrating the acquisition of multiple color separation features in a quality assessment method for a display device according to an embodiment of this application;

[0110] Figure 4 This is a schematic diagram illustrating the acquisition of second-level data in a quality assessment method for a display device according to an embodiment of this application;

[0111] Figure 5 This is a schematic diagram illustrating the fusion of first-level data and second-level data in a quality assessment method for a display device according to an embodiment of this application.

[0112] Figure 6 This is a schematic diagram of obtaining an initial image in a quality assessment method for a display device according to an embodiment of this application;

[0113] Figure 7 This is a schematic diagram of a quality assessment device for a display device according to an embodiment of this application;

[0114] Figure 8 This is a schematic diagram of an electronic device according to an embodiment of this application. Detailed Implementation

[0115] The technical solutions of 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, 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.

[0116] In the accompanying drawings, the size of constituent elements, the thickness of layers, or areas may sometimes be exaggerated for clarity. Therefore, any implementation of this disclosure is not necessarily limited to the dimensions shown in the drawings, and the shapes and sizes of the components in the drawings do not reflect true proportions. Furthermore, the drawings schematically illustrate ideal examples, and any implementation of this disclosure is not limited to the shapes or values ​​shown in the drawings.

[0117] Color separation grading is a crucial aspect of display quality assessment. Its goal is to score the degree of color separation in a given image or set of images reflected from the display screen, based on how ambient light is reflected. Color separation can reduce the display's ambient contrast, especially in the presence of point light sources, producing noticeable color separation spots that negatively impact the viewer's experience.

[0118] Building upon conventional display screen quality assessment tasks, the task of classifying color separation phenomena requires not only insensitivity to the rotation angle of the display screen and the color separation spot (i.e., rotational invariance), but also the ability to correctly classify various types of display screens and color separation spot shapes (i.e., universality). This task has broad application prospects in display screen quality assessment, especially in the development and quality assessment of novel low-energy, polarizer-free display screens, and possesses significant research value.

[0119] However, existing methods for classifying color separation phenomena in displays rely on manually constructing features or using deep convolutional neural networks to obtain classification results. These methods have poor classification accuracy and can easily lead to inconsistent prediction results for different display models with the same color separation level.

[0120] In view of this, embodiments of this application propose a method for quality evaluation of a display device. Figure 1 This application illustrates a flowchart of a quality assessment method for a display device according to an embodiment of the present application. Figure 1 As shown, the method includes the following steps:

[0121] S101. Obtain a first image, wherein the first image includes at least a color separation spot image of the display device.

[0122] S102. Based on the first image, obtain multiple color separation features, which are used to characterize the degree of color dispersion of the color separation spot image in different directions.

[0123] S103. Input the multiple color separation features into the first prediction model to obtain the first graded data.

[0124] S104. Input the first image into the second prediction model to obtain the second classification data.

[0125] S105. Based on the first classification data and the second classification data, a target classification result is obtained, wherein the target classification result is used to characterize the color separation level of the display device.

[0126] In this embodiment, the first image includes at least a color-separated light spot image of the display device. The display device is used to display the image and includes, but is not limited to, any product or component with display functionality, such as a television, laptop, tablet, mobile phone, personal digital assistant (PDA), navigator, wearable device, or virtual reality (VR) device. Those skilled in the art can select the appropriate device based on its actual application; further details are omitted here. The display device includes a display panel, which can be a liquid crystal display (LCD), an organic light-emitting diode (OLED), a quantum dot light-emitting diode (QLED), etc. This embodiment does not specifically limit the type of display panel.

[0127] In the quality evaluation of display devices, the impact of color separation on the display device is an important quality evaluation indicator. Color separation reduces the environmental contrast of the display device, thereby affecting the user experience. The classification of color separation can be reflected by color separation spots on the display device. Color separation spots refer to local colored spots on the display device caused by color separation. The color of the color separation spots may be red, green, blue or a mixture of them.

[0128] In this embodiment, since the color-separating spots are radially distributed, their physical characteristics can be obtained as color-separating features to ensure interpretability. These color-separating features characterize the degree of color dispersion of the color-separating spots in different directions. Furthermore, the first prediction model in this embodiment is an ensemble learning model. By processing the color-separating features using this model, as many color-separating features related to the color-separation level as possible can be captured comprehensively, resulting in high accuracy and robustness of the output first-level data.

[0129] On the other hand, to ensure that the quality assessment method in this embodiment still performs well in the color separation grading task for specific color separation spots, this embodiment processes the first image using a second prediction model. This second prediction model is an end-to-end trained neural network model that can autonomously learn different color separation spot patterns, ensuring good performance even for specific color separation spots. Finally, the first grading data, which has strong interpretability, and the second grading data, which has good grading performance for specific color separation spots, are combined to obtain the final grading result. This allows the quality assessment method in this embodiment to have higher grading accuracy for color separation spots on different display devices, effectively improving the model's performance in performing the color separation grading task.

[0130] To enable those skilled in the art to better understand the technical solution of this application, the following is a detailed description of a quality assessment method for a display device provided in this application:

[0131] In specific implementation step S101, an initial image is first acquired. The initial image is an image of the display device that has been acquired, and the color separation spot is formed on the display device. The initial image includes at least the color separation spot image of the display device, so as to acquire the color separation spot image on the display device based on the initial image.

[0132] In some optional implementations, the rotation angle and orientation of the initial image can affect the first image obtained based on the initial image. This can cause the rotation angle and orientation of the color separation spot image of the first image to have an adverse effect on the subsequent classification of the color separation spot, thus affecting the classification accuracy of the model. For example, if the initial image is an image acquired by the display device under offset, tilt, or other orientations, the color separation spot image in the first image obtained based on the initial image will have a certain rotation angle (lacking rotation invariance). As a result, when the first image is used as input data for subsequent processing, the processing result will be affected by the rotation angle, leading to misjudgment results. This will cause the accuracy of the output color separation spot classification result to decrease, thus affecting the classification accuracy of the model.

[0133] Therefore, in this embodiment of the application, the initial image is obtained by performing calibration processing on the image during the acquisition process. Specifically, Figure 6 This illustration shows a schematic diagram of acquiring an initial image in a quality assessment method for a display device according to an embodiment of this application, as shown below. Figure 6As shown, firstly, the base and image acquisition device are calibrated based on a calibration board. The base is used to hold the calibration board and the display device, and the image acquisition device is used to acquire images of the area on the base within its field of view. For example, the image acquisition device is a camera or other device with image acquisition capabilities. Then, the display device is placed on the base, and images of the display device are periodically acquired based on a moving point light source, resulting in multiple initial images. The color separation spot images in different initial images have different postures on the display device. It should be noted that the different postures of the color separation spot images in different initial images are caused by the moving point light source illuminating the display device at different positions during the periodic image acquisition by the image acquisition device. Based on these different initial images, first images with different postures are obtained. Therefore, the color separation level of the display device is judged by comprehensively considering multiple grading results obtained from these multiple first images with different postures, further improving the accuracy of the results.

[0134] In this embodiment of the application, after calibrating the base and the image acquisition device based on the calibration plate, the display device placed on the base is positioned horizontally and with a suitable image size, thus ensuring that the rotation angle and orientation of the initial image do not affect the first image obtained based on the initial image.

[0135] In this embodiment of the application, although the calibration process during the acquisition of the initial image eliminates to some extent the adverse effects of the display device's posture on the rotation invariance of the first image, in order to improve the rotation invariance of the first image, the initial image is preprocessed after acquisition to obtain the first image. Specifically, Figure 2 This application illustrates a flowchart of a method for quality assessment of a display device according to an embodiment of the present application, in which a first image is acquired. Figure 2 As shown, firstly, after obtaining the initial image, an affine transformation is performed on the initial image based on a standard plane to eliminate the image distortion effect of the initial image caused by the tilt angle that may exist between the plane where the display device is located and the image acquisition device, thereby obtaining a first preprocessed image.

[0136] In some optional implementations, the first preprocessed image is acquired as follows: a calibration result is generated for the base and the image acquisition device based on the calibration plate; a homography matrix is ​​constructed between the plane where the display device is located and the standard plane based on the calibration result, the homography matrix being used to realize the transformation between the plane where the display device is located and the standard plane; and an affine transformation is performed on the initial image based on the homography matrix to eliminate the image distortion effect of the initial image, thereby obtaining the first preprocessed image.

[0137] In this embodiment, because the image acquisition device is limited by its field of view during the acquisition of the initial image, the initial image may contain background images outside the display device, such as images of other devices connected to the display device. Information from these background images is irrelevant to obtaining the grading results of the color separation spot, and therefore needs to be removed to improve the grading accuracy of the model. After obtaining the first preprocessed image, edge detection is performed on the first preprocessed image to obtain the region image containing the edge of the display device, thereby removing images located outside the edge of the display device, which becomes the second preprocessed image.

[0138] In some optional embodiments, the second preprocessed image is acquired as follows: based on an edge detection algorithm, the edge information of the display device in the first preprocessed image is acquired as first edge information; based on the first edge information, the first preprocessed image is processed to obtain a separated image. For example, the first edge information may be a closed detection box enclosed by multiple key points located on the edge of the display device; based on the detection box, the first preprocessed image is cropped to retain the image area within the detection box, thus obtaining the separated image. It should be noted that the above example is merely an optional method provided to enable those skilled in the art to better understand the solution of this application. The specific method of acquiring the separated image based on the first edge information is not limited to cropping and can be determined according to the actual situation; this application does not impose any limitations here.

[0139] Optionally, the edge detection algorithm can be the Canny edge detection algorithm, or other detection methods that can achieve the same edge detection effect; this application does not limit this. For example, the process of edge detection on the first preprocessed image based on the Canny edge detection algorithm is as follows: First, Gaussian filtering is applied to the input first preprocessed image to smooth the image and remove noise; then, the magnitude and direction of the gradient are calculated on the smoothed first preprocessed image; subsequently, non-maximum suppression is performed in the gradient direction to eliminate pixels that may not constitute edges. This step makes the edges finer while preserving the maximum value of the local gradient; next, a dual-threshold strategy is used to determine which edge pixels are true edges, obtaining the first edge information. Based on the two thresholds (high threshold and low threshold), the gradient magnitude is divided into three categories: strong edges, weak edges, and non-edges; finally, the first edge information is refined by connecting strong edge pixels, and inconspicuous edges are eliminated. Obtaining the first edge information based on the Canny edge detection algorithm can find the true edges of the display device in the first preprocessed image to the greatest extent possible, has a certain resistance to noise, and generates refined edges.

[0140] Since other display devices are connected to the edges of the display device, the edges of the separated image may contain portions of those other devices. Therefore, after obtaining the separated image, iterative detection is performed on the separated image based on a defect detection algorithm to continuously detect edge information located outside the region image corresponding to the display device, which serves as the second edge information. Optionally, the defect detection algorithm can be a convex defect detection algorithm. Subsequently, based on the second edge information, the separated image is processed by cropping it and solving for the minimum bounding rectangle of the cropped image to obtain the second preprocessed image, which corresponds to the image of the display device.

[0141] After obtaining the second preprocessed image, the first image is obtained based on the second preprocessed image. Specifically, since the first image is an image in the second preprocessed image that includes the color separation spot region, when the display device forms the color separation spot, the brightness of the area where the color separation spot is located is significantly greater than that of other areas under the action of the point light source. Therefore, brightness detection is performed on the second preprocessed image, and the brightness of each pixel in the second preprocessed image is judged based on a preset brightness threshold. If the brightness of a pixel is greater than or equal to the preset brightness threshold, it indicates that the pixel is a pixel in the area where the color separation spot is located; if the brightness of a pixel is less than the preset brightness threshold, it indicates that the pixel is not a pixel in the area where the color separation spot is located. The image of the area in the second preprocessed image whose brightness is higher than the preset brightness threshold is obtained as the first image.

[0142] In specific implementation step S102, after acquiring the first image, multiple color separation features are extracted from the first image. Specifically, Figure 3 This illustration shows a schematic diagram of obtaining multiple color separation features in a quality assessment method for a display device according to an embodiment of this application, such as... Figure 3 As shown, the color separation spot on the display device includes a white spot at the center caused by overexposure, and a color separation texture pattern radially distributed around the white spot. The color separation texture pattern is highly correlated with the degree of color separation of the color separation spot. Therefore, to improve the model's grading accuracy, when extracting the color separation features of the first image, image standardization is performed on the first image to convert white pixels into black pixels, resulting in the second image. This embodiment of the application eliminates white pixels with low correlation to the degree of color separation by standardizing the first image, allowing the model to focus more on the color separation texture pattern when extracting color separation features, thereby improving the accuracy of the grading data obtained based on the color separation features.

[0143] In some optional implementations, the second image is obtained by standardizing the pixels in the first image within the target color space. Specifically, firstly, the pixel values ​​corresponding to each pixel in the first image in each color channel are obtained within the target color space; based on the pixel values ​​corresponding to each pixel in each color channel, a bias pixel value is determined for each pixel, where the bias pixel value is the minimum value among the pixel values ​​corresponding to each pixel in each color channel; based on the bias pixel value, the pixel values ​​corresponding to each pixel in each color channel are processed to obtain the second image. Optionally, the bias pixel value is subtracted from the pixel value of each pixel in each color channel. In this case, the pixel value of a white pixel is converted to a black pixel after subtracting the bias pixel value, thereby achieving the standardization process.

[0144] For example, if the target color space is RGB, then the pixel value (r, g, b) of each pixel is obtained. The bias pixel value is the minimum value a = Min(r, g, b) among the pixel values ​​(r, g, b) of each color channel of each pixel. Based on the bias pixel value a, the pixel values ​​of the pixels are processed to obtain the standardized pixel values ​​(ra, ga, ba). The pixel value of a white pixel is (255, 255, 255). Based on the pixel value of the white pixel, the bias pixel value corresponding to the white pixel is determined to be 255. Then, based on the bias pixel value, the pixel value of the white pixel is processed to obtain the standardized pixel value (0, 0, 0), corresponding to the black pixel, thereby removing the white pixels in the first image.

[0145] It should be noted that the above example is only an optional implementation method given to enable those skilled in the art to better understand the solution of this application. The specific method for determining the target color space and bias value when converting the first image into the second image can be determined according to the actual situation, as long as the white pixels are converted into black pixels.

[0146] After obtaining the second image, since the color separation texture pattern in the second image is radially distributed relative to the central light spot, the multiple color separation features can be obtained based on multiple pixels in different directions in the second image and the central pixel of the color separation light spot. Specifically, as follows... Figure 3As shown, the pixel values ​​of multiple pixels in the second image along the target direction are obtained, wherein the different target directions include multiple different directions passing through the center of the color separation spot; subsequently, the first and second moments of the multiple pixels along the target direction and the central pixel of the color separation spot are obtained as the color separation feature of the second image in the target direction, wherein the first and second moments include the first and second moments of the target pixel corresponding to different color channels, and the first and second moments of the target pixel are determined based on the distance between the target pixel and the center of the color separation spot along the target direction, and the pixel value of the target pixel in each color channel.

[0147] In some optional implementations, the target direction is any one of a first direction, a second direction, a third direction, and a fourth direction. The target direction passes through the center of the color separation spot. The first and second directions extend in opposite directions vertically, and the third and fourth directions extend in opposite directions horizontally. Color separation features along the first, second, third, and fourth directions in the second image are acquired respectively, serving as multiple color separation features corresponding to the second image. For example, the second image along the first direction (red)...

[0148] The color separation features (first and second moments) on the color channels are obtained using the following formula:

[0149]

[0150] Where M1 is the first moment of the pixel, M2 is the second moment of the pixel, L1 is the first direction, l is the distance between the pixel along the first direction and the center pixel of the color separation spot, and r l This is the pixel value of the red color channel of the pixel.

[0151] It is easy to understand that the color separation features of other directions and / or other color channels can also be obtained by referring to the above formula, simply by replacing the pixel values ​​of the target direction and / or the color channel accordingly.

[0152] In this embodiment of the application, by acquiring the color separation features of the second image in different target directions, the divergence of the color separation spot in each direction is accurately characterized, which better reflects the color separation degree of the color separation spot. The grading result obtained based on the multiple color separation features has strong interpretability and improves the grading accuracy.

[0153] In specific implementation step S103, after obtaining the multiple color separation features, the multiple color separation features are input into a first prediction model. The first prediction model is an ensemble learning model, such as a random forest model or other generalized linear models. The ensemble learning model has good interpretability. The first prediction model makes predictions based on the multiple color separation features to obtain the first graded data. Taking the random forest model as an example, after obtaining the color separation features, the multiple color separation features are first preprocessed; the multiple color separation features are input into the first prediction model, which is a random forest model. The random forest model consists of multiple decision trees constructed based on different feature subsets. The multiple decision trees in the first prediction model process the multiple color separation features respectively to obtain multiple classification results; the multiple classification results are processed based on a voting mechanism to obtain the first graded data. The first graded data is used to characterize the probability distribution of the display device belonging to different color separation levels.

[0154] In this embodiment of the application, the random forest model is used to classify the multiple color separation features, which can capture as many high-end effective features related to the color separation level as possible in a more comprehensive way. Moreover, the random forest model generates the first graded data based on a voting mechanism, which has high accuracy and robustness.

[0155] In some optional implementations, the random forest model is trained according to the following steps: obtaining a training dataset, the training data including color separation features corresponding to multiple color separation spot images and color separation levels of the multiple color separation spot images; generating multiple different feature subsets based on the training dataset; constructing multiple decision trees based on the multiple different feature subsets, and integrating the multiple decision trees into an original random forest model; training and optimizing the original random forest model based on the multiple feature subsets to obtain an optimized random forest model.

[0156] In specific implementation step S104, the first image is processed based on the second prediction model to obtain the second classification data. The second prediction model is a neural network model, obtained through end-to-end training and optimization using a data-driven approach. It can autonomously learn different color separation spot patterns, thus exhibiting good performance for specific color separation spots. Processing the first image based on the second prediction model enhances feature extraction and representation capabilities for image classification tasks, thereby efficiently modeling the relationship between the input first image and the color separation level, improving color separation prediction performance, and enhancing classification accuracy. Optionally, the second prediction model is a recurrent convolutional residual network.

[0157] In some optional embodiments, the color-separated light spots on the display device mostly exhibit radial symmetry but are not uniform, and the color-separated light spots have radial rays in a specific direction. In actual judgment, the rating should not be affected regardless of the angle at which the light spots are presented; therefore, this requires the model to have rotational invariance in recognizing the color-separated light spot pattern. Therefore, before inputting the first image into the second prediction model, the first image needs to be processed to ensure that the image processed by the second prediction model has rotational invariance. Figure 4 This illustration shows a schematic diagram of obtaining second-level data in a quality assessment method for a display device according to an embodiment of this application, as shown below. Figure 4 As shown, the first image is first transformed using polar coordinates to obtain the third image. Specifically, the center point of the color separation spot in the first image is identified, and the first image is radially unfolded with the center point as the origin, transforming the Cartesian coordinate system to the polar coordinate system to obtain the third image. Subsequently, the third image is input into the second prediction model, and a circular convolution process is performed on the third image based on the second prediction model to obtain a first feature. The first feature is an image feature corresponding to the third image that has translation invariance. Because the third image corresponding to the first feature has translation invariance, the image in the corresponding Cartesian coordinate system has rotation invariance, allowing the image input into the second prediction model to better adapt to the model characteristics of the second prediction model, effectively improving model performance.

[0158] After obtaining the first feature corresponding to the translation invariance, the first feature is processed based on the second prediction model to obtain the second classification data. The second classification data is used to characterize the probability distribution of the display device belonging to different color separation levels. The second prediction model introduces a cyclic filling mechanism to address the shortcomings of convolutional neural networks in edge information processing, and to better adapt to the head-to-tail connection problem after the first image is expanded based on polar coordinates. Furthermore, the second prediction model includes multiple basic blocks and bottleneck blocks, which are used to form a residual network. Residual connections effectively solve the gradient vanishing problem of the second prediction model. In this embodiment, processing the first image through the cyclic convolutional residual network can better achieve the rotation invariance of the color separation spot in the first image, more efficiently model the relationship between the input first image and the color separation level, and improve classification accuracy.

[0159] In the specific implementation step S105, although the first prediction model has strong interpretability, it may make certain misjudgments in the classification results for specific light spots, and the model's versatility (the model's ability to correctly classify various types of screens and color separation light spot shapes) is not strong enough; although the second prediction model has good performance for specific light spots, the second prediction model is a black box model with poor interpretability. Therefore, after obtaining the first classification data based on the first prediction model and the second classification data based on the second prediction model, it is necessary to fuse the first classification model and the second classification model to obtain the final target classification result, so that the quality assessment method provided in this application embodiment can simultaneously take into account versatility and interpretability, and further improve the model performance.

[0160] In some alternative implementations, Figure 5 This illustration shows a schematic diagram of the fusion of first-level data and second-level data in a quality assessment method for a display device according to an embodiment of this application. Figure 5 As shown, the first and second graded data are fused based on confidence levels to obtain the target graded result. Specifically, based on the first graded data, a first confidence level is determined for each prediction result of different color separation levels in the first graded data. Optionally, when the first prediction model is a random forest model, the first confidence level is determined based on the proportion of samples in the leaf nodes of each decision tree in the random forest model. The higher the proportion of samples representing the majority class in the leaf nodes, the higher the first confidence level of the prediction result corresponding to the color separation level in the first graded data. Based on the second graded data, a second confidence level is determined for each prediction result of different color separation levels in the second graded data. The second confidence level is determined by the softmax probability of the prediction result of the color separation level.

[0161] After determining the first confidence level of the category of each color separation level in the first grading data and the second confidence level of the category of each color separation level in the second grading data, the first grading data and the second grading data are fused according to the category of color separation level to obtain the grading results corresponding to the categories of different color separation levels. The final target grading result is obtained based on the grading results.

[0162] In some optional implementations, after obtaining the first confidence level and the second confidence level, the first confidence level is used as the weight of the prediction result corresponding to the first confidence level in the first grading data, and the second confidence level is used as the weight of the prediction result corresponding to the second confidence level in the second grading data. The prediction results corresponding to the first confidence level in the first grading data and the prediction results corresponding to the second confidence level in the second grading data are then fused for categories with the same color separation level to obtain a grading result for each color separation level. The maximum value among all the grading results for all color separation levels is taken as the target grading result. Optionally, the grading result is determined according to the following formula:

[0163] in,

[0164] Where c represents the level of color separation (e.g., 10 levels from 1 to 10), k is the index of different models (e.g., the first prediction model or the second prediction model), and x i Indicates a specific sample, w represents the confidence score of class c when the sample is modeled with index value k. k,i This represents the weight held by the model with index value k during fusion after the sample has undergone normalization. Furthermore, through w... k,i By fusing the confidence scores of different models, P is obtained. (c) (x i ).

[0165] To demonstrate the superior performance of this application's embodiments in the color separation phenomenon grading task, a specialized evaluation dataset was used, and relatively accurate color separation levels of the images were obtained as sample labels through manual coarse annotation and multi-round iterative rescoring. Specifically, firstly, color separation spot images from various locations on the display device were acquired, and each image underwent standardization processing after acquisition, including automatic alignment and cropping, to ensure data consistency and quality. The resulting dataset consisted of 204 display devices and approximately 10,000 images. In the initial annotation stage, each image was initially labeled with a color separation level, and the images were divided into 5 color separation levels based on the degree of color separation. Due to inconsistencies in the initial labels and a large number of noisy labels, multi-round iterative rescoring was employed. Samples with low model confidence were manually blindly rated for color separation, and this process was repeated multiple times to obtain relatively accurate data labels. The labeled proprietary data was used for model training and testing.

[0166] Table 1. Performance data of the quality assessment method provided in this application embodiment on a proprietary dataset.

[0167]

[0168] Table 1 shows the performance data of the quality assessment method provided in this application embodiment on a proprietary dataset. As shown in Table 1, the target classification results obtained by the quality assessment method based on this application embodiment have higher accuracy compared to single convolutional network or random forest models. This application effectively combines the interpretability of the random forest model and the flexibility and adaptability of the deep recurrent convolutional neural network model by constructing a multi-model fusion strategy, significantly improving the accuracy and stability of color separation phenomenon classification. The accuracy of the fusion model on the test set is higher than that of using the random forest model or the deep recurrent convolutional neural network model alone. Especially when dealing with specific light spots and complex backgrounds, the fusion model shows stronger robustness and generalization ability.

[0169] This application provides a method, apparatus, device, medium, and product for quality assessment of a display device. The method includes: acquiring a first image, the first image including at least a color separation spot image of the display device; acquiring multiple color separation features based on the first image, the color separation features being used to characterize the degree of color dispersion of the color separation spot image in different directions; inputting the multiple color separation features into a first prediction model to obtain first grading data; inputting the first image into a second prediction model to obtain second grading data; and obtaining a target grading result based on the first grading data and the second grading data, the target grading result being used to characterize the color separation level of the display device. This application obtains the first grading data by processing the color separation features characterizing the degree of color dispersion, ensuring the interpretability of the grading result; and obtains the second grading data based on the second prediction model, effectively improving the prediction quality of the grading result, thereby giving the grading result obtained based on the first grading data and the second grading data higher grading accuracy and effectively improving the performance of the model in performing the color separation grading task.

[0170] Based on the same inventive concept, embodiments of this application disclose a quality evaluation device for a display device. Figure 7 A schematic diagram of a quality assessment device for a display device according to an embodiment of this application is shown, as follows: Figure 7 As shown, the device includes:

[0171] An image acquisition module is used to acquire a first image, wherein the first image includes at least a color separation spot image of the display device;

[0172] The feature extraction module is used to obtain multiple color separation features based on the first image, wherein the color separation features are used to characterize the degree of color dispersion of the color separation spot image in different directions;

[0173] The first prediction module is used to input the multiple color separation features into the first prediction model to obtain the first graded data;

[0174] The second prediction module is used to input the first image into the second prediction model to obtain the second hierarchical data;

[0175] A grading module is used to obtain a target grading result based on the first grading data and the second grading data, wherein the target grading result is used to characterize the color separation level of the display device.

[0176] In one optional implementation, the feature extraction module includes:

[0177] The image normalization submodule is used to perform image normalization processing on the first image to convert white pixels in the first image into black pixels, thereby obtaining the second image.

[0178] The feature extraction submodule is used to obtain the multiple color separation features based on multiple pixels in different directions in the second image and the center pixel of the color separation spot.

[0179] In one optional implementation, the image normalization submodule includes:

[0180] The pixel value determination unit is used to obtain the pixel value of each pixel in the first image in each color channel within the target color space;

[0181] The offset pixel value unit is used to determine the offset pixel value corresponding to each pixel based on the pixel value corresponding to each pixel in each color channel, wherein the offset pixel value is the minimum value among the pixel values ​​corresponding to each pixel in each color channel.

[0182] An image normalization unit is used to process the pixel values ​​corresponding to each pixel in each color channel based on the bias pixel value corresponding to each pixel to obtain the second image.

[0183] In one optional implementation, the feature extraction submodule includes:

[0184] A target direction pixel value acquisition unit is used to acquire the pixel values ​​of multiple pixels of the second image in a target direction. The target direction is any one of a first direction, a second direction, a third direction, and a fourth direction. The target direction passes through the center of the color separation spot. The first direction and the second direction extend in opposite directions in the vertical direction, and the third direction and the fourth direction extend in opposite directions in the horizontal direction.

[0185] The feature extraction unit is used to obtain the first and second moments of the multiple pixels in the target direction and the center pixel of the color separation spot, as the color separation feature of the second image in the target direction.

[0186] In one optional implementation, the first prediction module includes:

[0187] The preprocessing submodule is used to preprocess the multiple color separation features;

[0188] The first classification submodule is used to input the multiple color separation features into the first prediction model, and process the multiple color separation features based on multiple decision trees in the first prediction model to obtain multiple classification results;

[0189] The first graded data submodule is used to process the multiple classification results based on a voting mechanism to obtain the first graded data, which is used to characterize the probability distribution of the display device belonging to different color separation levels.

[0190] In one optional implementation, the second prediction module includes:

[0191] The coordinate transformation submodule is used to perform polar coordinate transformation on the first image to obtain the third image;

[0192] The convolution submodule is used to input the third image into the second prediction model, and perform cyclic convolution processing on the third image based on the second prediction model to obtain a first feature, wherein the first feature is an image feature corresponding to the third image with translation invariance;

[0193] The second graded data submodule is used to process the first feature based on the second prediction model to obtain the second graded data, which is used to characterize the probability distribution of the display device belonging to different color separation levels.

[0194] In one optional implementation, the hierarchical module includes:

[0195] The first confidence level submodule is used to determine the first confidence level corresponding to the prediction results of different color separation levels in the first graded data based on the first graded data.

[0196] The second confidence level submodule is used to determine the second confidence level corresponding to the prediction results of different color separation levels in the second classification data based on the second classification data.

[0197] The grading submodule is used to process the first grading data and the second grading data based on the first confidence level and the second confidence level to obtain the target grading result.

[0198] In one optional implementation, the hierarchical submodule includes:

[0199] The fusion unit is used to fuse the prediction results corresponding to the first confidence level in the first graded data and the prediction results corresponding to the second confidence level in the second graded data for categories with the same color separation level, respectively, to obtain the grading result for each color separation level.

[0200] A grading unit is used to take the maximum value among the grading results of all color separation levels as the target grading result.

[0201] In one optional implementation, the image acquisition module includes:

[0202] The calibration submodule is used to calibrate the base and image acquisition device based on the calibration board;

[0203] The image acquisition submodule is used to place the display device on the base and periodically acquire images of the display device based on a moving point light source to obtain multiple initial images. The initial images include at least the color separation spot image of the display device, and the color separation spot images in different initial images have different orientations in the display device.

[0204] The image acquisition submodule is used to preprocess the initial image to obtain the first image.

[0205] In one optional implementation, the image acquisition submodule includes:

[0206] An affine transformation unit is used to perform an affine transformation on the initial image based on a standard plane to obtain a first preprocessed image;

[0207] An edge detection unit is used to perform edge detection on the first preprocessed image and obtain an image of the region containing the edge of the display device in the first preprocessed image, as a second preprocessed image.

[0208] The image acquisition unit is used to perform brightness detection on the second preprocessed image and acquire the region image in the second preprocessed image whose brightness is higher than a preset brightness threshold, as the first image.

[0209] In one optional embodiment, the affine transformation unit includes:

[0210] The calibration subunit is used to generate calibration results for the base and the image acquisition device based on the calibration plate;

[0211] A homography matrix construction subunit is used to construct a homography matrix between the plane where the display device is located and the standard plane based on the calibration results;

[0212] The first preprocessing image subunit is used to perform an affine transformation on the initial image based on the homography matrix to obtain the first preprocessed image.

[0213] In one optional implementation, the edge detection unit includes:

[0214] The first edge information subunit is used to obtain the edge information of the display device in the first preprocessed image based on the edge detection algorithm, as the first edge information;

[0215] A separate image subunit is used to process the first preprocessed image based on the first edge information to obtain a separate image;

[0216] The second edge information subunit is used to iteratively detect the separated image based on the defect detection algorithm to obtain edge information in the separated image that is outside the region image corresponding to the display device, as the second edge information;

[0217] The second preprocessing image subunit is used to process the separated image based on the second edge information to obtain the second preprocessed image.

[0218] Based on the same inventive concept, this application discloses an electronic device. Figure 8 A schematic diagram of an electronic device according to an embodiment of this application is shown, such as... Figure 8 As shown, the electronic device 100 includes a memory 110 and a processor 120. The memory 110 and the processor 120 are connected via a bus. The memory 110 stores at least one instruction, at least one program, code set, or instruction set. The at least one instruction, the at least one program, the code set, or the instruction set can be loaded and run on the processor 120 to implement the steps in the quality evaluation method for the display device disclosed in the embodiments of this application.

[0219] Based on the same inventive concept, embodiments of this application disclose a computer-readable storage medium storing at least one instruction, at least one program, code set, or instruction set thereon. The at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by a processor to implement the steps in the quality evaluation method for the display device disclosed in embodiments of this application.

[0220] Based on the same inventive concept, embodiments of this application disclose a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps in the quality evaluation method for a display device disclosed in embodiments of this application.

[0221] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0222] This application describes embodiments with reference to flowchart illustrations and / or block diagrams of methods, apparatuses, electronic devices, and computer program products according to embodiments of this application. It should 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 terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0223] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal 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.

[0224] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment 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.

[0225] Although preferred embodiments of the present application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present application.

[0226] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal 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 terminal 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 terminal device that includes said element.

[0227] The above provides a detailed description of the quality evaluation method, apparatus, equipment, medium, and product of a display device provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for quality assessment of a display device, characterized in that, The method includes: Acquire a first image, the first image including at least a color separation spot image of the display device; Based on the first image, multiple color separation features are obtained, which are used to characterize the degree of color dispersion of the color-separated spot image in different directions; The multiple color separation features are input into the first prediction model to obtain the first graded data; The first image is input into the second prediction model to obtain the second classification data; Based on the first grading data and the second grading data, a target grading result is obtained, which is used to characterize the color separation level of the display device.

2. The quality assessment method for a display device according to claim 1, characterized in that, The step of obtaining multiple color separation features based on the first image includes: The first image is subjected to image normalization processing to convert white pixels in the first image into black pixels, thereby obtaining the second image; Based on multiple pixels in different directions in the second image and the center pixel of the color separation spot, the multiple color separation features are obtained respectively.

3. The quality assessment method for a display device according to claim 2, characterized in that, The step of performing image normalization processing on the first image to convert white pixels in the first image into black pixels to obtain the second image includes: Within the target color space, obtain the pixel value of each pixel in the first image corresponding to each color channel; Based on the pixel values ​​corresponding to each pixel in each color channel, a bias pixel value is determined for each pixel, wherein the bias pixel value is the minimum value among the pixel values ​​corresponding to each pixel in each color channel. Based on the bias pixel value corresponding to each pixel, the pixel values ​​corresponding to each pixel in each color channel are processed to obtain the second image.

4. The quality assessment method for a display device according to claim 2, characterized in that, The step of obtaining multiple color separation features based on multiple pixels in different directions in the second image and the center pixel of the color separation spot includes: Obtain the pixel values ​​of multiple pixels in the second image in the target direction. The target direction is any one of the first direction, the second direction, the third direction, and the fourth direction. The target direction passes through the center of the color separation spot. The first direction and the second direction extend in opposite directions in the vertical direction, and the third direction and the fourth direction extend in opposite directions in the horizontal direction. The first and second moments of the multiple pixels in the target direction and the center pixel of the color separation spot are obtained as the color separation features of the second image in the target direction.

5. The quality assessment method for a display device according to claim 1, characterized in that, The first prediction model is a random forest model. The step of inputting the multiple color separation features into the first prediction model to obtain the first hierarchical data includes: The multiple color separation features are preprocessed; The multiple color separation features are input into the first prediction model, and the multiple color separation features are processed based on the multiple decision trees in the first prediction model to obtain multiple classification results; The multiple classification results are processed based on a voting mechanism to obtain the first classification data, which is used to characterize the probability distribution of the display device belonging to different color separation levels.

6. The quality assessment method for a display device according to claim 1, characterized in that, The second prediction model is a recurrent convolutional residual network. The step of inputting the first image into the second prediction model to obtain the second graded data includes: Perform polar coordinate transformation on the first image to obtain the third image; The third image is input into the second prediction model, and the third image is subjected to cyclic convolution processing based on the second prediction model to obtain the first feature, which is the image feature corresponding to the third image with translation invariance; The first feature is processed based on the second prediction model to obtain the second classification data, which is used to characterize the probability distribution of the display device belonging to different color separation levels.

7. The quality assessment method for a display device according to claim 1, characterized in that, The step of obtaining the target classification result based on the first classification data and the second classification data includes: Based on the first graded data, determine the first confidence level corresponding to the prediction results of different color separation levels in the first graded data; Based on the second graded data, determine the second confidence level corresponding to the prediction results of different color separation levels in the second graded data; The first and second graded data are processed based on the first and second confidence levels to obtain the target graded result.

8. The quality assessment method for a display device according to claim 7, characterized in that, The process of processing the first and second graded data based on the first and second confidence levels to obtain the target graded result includes: Using the first confidence level as the weight of the prediction result corresponding to the first confidence level in the first graded data and the second confidence level as the weight of the prediction result corresponding to the second confidence level in the second graded data, the prediction results corresponding to the first confidence level in the first graded data and the prediction results corresponding to the second confidence level in the second graded data for the same color separation level are fused to obtain the graded result for each color separation level. The maximum value among all the color separation level classification results is taken as the target classification result.

9. The quality assessment method for a display device according to claim 1, characterized in that, The acquisition of the first image includes: The base and image acquisition device are calibrated based on the calibration board; The display device is placed on the base, and images of the display device are periodically acquired based on a moving point light source to obtain multiple initial images. Each initial image includes at least a color separation spot image of the display device, and the color separation spot images in different initial images have different orientations in the display device. The initial image is preprocessed to obtain the first image.

10. The quality assessment method for a display device according to claim 9, characterized in that, The step of preprocessing the initial image to obtain the first image includes: Based on a standard plane, an affine transformation is performed on the initial image to obtain a first preprocessed image; Edge detection is performed on the first preprocessed image to obtain an image of the region in the first preprocessed image that contains the edge of the display device, which is then used as the second preprocessed image. Brightness detection is performed on the second preprocessed image to obtain the region image in the second preprocessed image whose brightness is higher than a preset brightness threshold, which is then used as the first image.

11. The method for quality assessment of a display device according to claim 9, characterized in that, The step of performing an affine transformation on the initial image based on a standard plane to obtain a first preprocessed image includes: Based on the calibration plate, calibration results are generated for the base and the image acquisition device; Based on the calibration results, a homography matrix is ​​constructed between the plane where the display device is located and the standard plane; Based on the homography matrix, an affine transformation is performed on the initial image to obtain the first preprocessed image.

12. The quality assessment method for a display device according to claim 9, characterized in that, The step of performing edge detection on the first preprocessed image to obtain an image of the region containing the edge of the display device in the first preprocessed image, as the second preprocessed image, includes: Based on the edge detection algorithm, the edge information of the display device in the first preprocessed image is obtained as the first edge information; Based on the first edge information, the first preprocessed image is processed to obtain a separated image; Based on the defect detection algorithm, the separated image is iteratively detected to obtain edge information located outside the region image corresponding to the display device in the separated image, which is used as the second edge information; Based on the second edge information, the separated image is processed to obtain the second preprocessed image.

13. A quality assessment device for a display device, characterized in that, The device includes: An image acquisition module is used to acquire a first image, wherein the first image includes at least a color separation spot image of the display device; The feature extraction module is used to obtain multiple color separation features based on the first image, wherein the color separation features are used to characterize the degree of color dispersion of the color separation spot image in different directions; The first prediction module is used to input the multiple color separation features into the first prediction model to obtain the first graded data; The second prediction module is used to input the first image into the second prediction model to obtain the second hierarchical data; A grading module is used to obtain a target grading result based on the first grading data and the second grading data, wherein the target grading result is used to characterize the color separation level of the display device.

14. An electronic device, characterized in that, The device includes a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps in the quality evaluation method for the display device according to any one of claims 1-12.

15. A computer-readable storage medium having a computer program / instructions stored thereon, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps in the quality evaluation method for the display device according to any one of claims 1-12.

16. A computer program product, characterized in that, Includes a computer program / instruction that, when executed by a processor, implements the steps in the quality evaluation method for the display device according to any one of claims 1-12.

Images