Liquid crystal panel light intensity prediction model training method and picture display method

Abstract

This invention provides a method for training a light intensity prediction model and displaying an image for a liquid crystal panel, relating to the field of display technology. The method includes: acquiring a sample theoretical grayscale sequence for the liquid crystal panel; inputting the sample theoretical grayscale sequence into an initial light intensity prediction model to obtain a light intensity prediction sequence; and iteratively adjusting the model parameters of the initial light intensity prediction model based on the light intensity prediction sequence and the light intensity sequence labels corresponding to the sample theoretical grayscale sequence to obtain the light intensity prediction model. The light intensity sequence labels are obtained by inputting the sample theoretical grayscale sequence into a recursive formula, which is determined based on triplet data, including the actual light intensity value of the previous frame, the theoretical grayscale value of the current frame, and the actual light intensity value of the current frame. It is evident that the entire process of this invention does not rely on the internal material parameters of the liquid crystal panel, thereby improving the controllability of light intensity prediction for the liquid crystal panel.

Description

Technical Field

[0001] This invention relates to the field of display technology, and in particular to a method for training a light intensity prediction model for a liquid crystal panel and a method for displaying an image. Background Technology

[0002] To achieve depth perception without the need for visual aids, naked-eye 3D display technology typically relies on time-division multi-frame liquid crystal modulation. This approach requires the liquid crystal display panel to continuously output multiple frames (usually four or more frames) of image sequences with specific grayscale distributions within a short period (typically less than 25ms). The light field values ​​from these multiple frames are then integrated and averaged over time using the persistence of vision effect, ultimately forming a continuous depth hierarchy perceptible to the human eye in space. To meet these requirements, the refresh rate of the liquid crystal panel needs to be increased to 240Hz or higher (4.17ms per frame) or even 480Hz (2.08ms per frame). However, high refresh rates result in poor light intensity response characteristics for the liquid crystal panel.

[0003] In related technologies, a response time model based on the Ericksen-Leslie equation is usually established. By parameterizing the rotational viscosity coefficient and elastic constant of liquid crystal molecules, the dynamic phase delay characteristics during grayscale conversion are described, and a quantitative relationship between cell thickness, driving voltage and response time is established to improve the light intensity response characteristics of LCD panels at high refresh rates.

[0004] However, the aforementioned technologies rely on the internal material parameters of the LCD panel (such as rotational viscosity coefficient, dielectric anisotropy, elastic constant, etc.). These internal material parameters are core process secrets of panel manufacturers and are not easy to obtain, thereby reducing the controllability of the light intensity prediction of the LCD panel. Summary of the Invention

[0005] This invention provides a method for training a light intensity prediction model and a display method for liquid crystal panels, in order to solve the defects in the prior art that reduce the controllability of light intensity prediction for liquid crystal panels.

[0006] This invention provides a method for training a light intensity prediction model for liquid crystal panels, comprising the following steps.

[0007] Obtain the theoretical grayscale sequence of samples for the LCD panel; The sample theoretical grayscale sequence is input into the initial light intensity prediction model to obtain the light intensity prediction sequence output by the initial light intensity prediction model; Based on the light intensity prediction sequence and the light intensity sequence labels corresponding to the sample theoretical grayscale sequence, the model parameters of the initial light intensity prediction model are iteratively adjusted to obtain the light intensity prediction model. The light intensity sequence labels are obtained by inputting the sample theoretical grayscale sequence into a recursive formula. The recursive formula is determined based on triplet data. The triplet data includes the actual light intensity value of the previous frame, the theoretical grayscale value of the current frame, and the actual light intensity value of the current frame. The actual light intensity value of the previous frame and the actual light intensity value of the current frame are obtained by measuring the liquid crystal panel using a light measurement module.

[0008] According to the light intensity prediction model training method for liquid crystal panels provided by the present invention, the recursive relation is determined as follows: For each of the triplet data, the actual light intensity value of the previous frame and the theoretical grayscale value of the current frame in the triplet data are used as the independent variables of the polynomial, and the actual light intensity value of the current frame in the triplet data is used as the dependent variable of the polynomial. Based on the independent and dependent variables corresponding to each triplet data, fit the coefficient matrix of the polynomial; Based on the coefficient matrix, the recursive relation is determined.

[0009] According to the present invention, a method for training a light intensity prediction model for a liquid crystal panel is provided, the method further comprising: Generate a grayscale transition sequence that includes multiple grayscale values; Multiple gray-level transition pairs are generated based on the gray-level transition sequence, and each gray-level transition pair includes any two target gray-level values ​​in the gray-level transition sequence; For each grayscale transfer pair, the liquid crystal panel is controlled to alternately display a preset transfer time based on the two target grayscale values ​​in the grayscale transfer pair; Within the preset transfer time, the light intensity value of the displayed liquid crystal panel is collected by the light measurement module; Based on the light intensity values ​​corresponding to each grayscale transfer pair, the light intensity response curve is determined; Based on the light intensity response curve, the triplet data corresponding to each grayscale transition pair is determined.

[0010] According to the present invention, a method for training a light intensity prediction model for a liquid crystal panel is provided, wherein determining the triplet data corresponding to each grayscale transition pair based on the light intensity response curve includes: The light intensity response curve is segmented based on the preset transition time to obtain multiple segments; For each of the segments, the segments are denoised to obtain denoised segments; The light intensity values ​​collected within the same pixel time period in the denoised segment are integrated to obtain the single-frame light intensity value corresponding to each of the same pixel time periods. The same pixel time period is determined based on the sampling resolution of the light measurement module and the refresh rate of the liquid crystal panel. Based on the light intensity values ​​of each single frame corresponding to the denoised segment, the triplet data corresponding to the grayscale transition pair is determined.

[0011] According to the present invention, a method for training a light intensity prediction model for a liquid crystal panel is provided, wherein determining the triplet data corresponding to the grayscale transition pair based on the light intensity values ​​of each single frame corresponding to the denoised segment includes: Based on the light intensity values ​​of all single frames corresponding to the denoised segments, a reference value is determined. Based on the light intensity values ​​of all single frames corresponding to the denoised segments and the reference value, the actual light intensity value of the previous frame and the actual light intensity value of the current frame are determined. The actual light intensity value of the previous frame, the actual light intensity value of the current frame, and the theoretical gray level of the current frame are determined as the triplet data corresponding to the gray level transition pair.

[0012] According to the present invention, a method for training a light intensity prediction model for a liquid crystal panel, wherein determining the actual light intensity value of the previous frame and the actual light intensity value of the current frame based on all single-frame light intensity values ​​corresponding to the denoised segment and the reference value includes: If the theoretical grayscale value of the previous frame is less than or equal to the theoretical grayscale value of the current frame, determine the single-frame light intensity values ​​of the previous frame that are less than or equal to the reference value among all single-frame light intensity values ​​corresponding to the denoised segment, and determine the average value of the single-frame light intensity values ​​of the previous frame that are less than or equal to the reference value as the actual light intensity value of the previous frame. If the theoretical grayscale value of the previous frame is greater than the theoretical grayscale value of the current frame, determine the single-frame light intensity values ​​of all previous frames that are greater than the reference value among all single-frame light intensity values ​​corresponding to the denoised segment, and determine the average value of the single-frame light intensity values ​​of all previous frames that are greater than the reference value as the actual light intensity value of the previous frame. If the theoretical grayscale value of the previous frame is less than or equal to the theoretical grayscale value of the current frame, determine the single-frame light intensity values ​​of all current frames that are greater than the reference value from all single-frame light intensity values ​​corresponding to the denoised segment, and determine the average value of the single-frame light intensity values ​​of all current frames that are greater than the reference value as the actual light intensity value of the current frame. If the theoretical grayscale value of the previous frame is greater than the theoretical grayscale value of the current frame, determine the single-frame light intensity value of the current frame that is less than or equal to the reference value among all single-frame light intensity values ​​corresponding to the denoised segment, and determine the average value of the single-frame light intensity values ​​of the current frame that is less than or equal to the reference value as the actual light intensity value of the current frame.

[0013] According to the present invention, a method for training a light intensity prediction model for a liquid crystal panel is provided, the method further comprising: Starting from the first theoretical gray value in the theoretical gray value sequence, the system iterates through the sample, inputting the currently iterated theoretical gray value and the initial actual light intensity value into the recursive formula to obtain the first actual light intensity value of the sample output by the recursive formula. Input the actual light intensity value of the first sample and the theoretical grayscale value of the sample in the next traversal into the recursive formula to obtain the actual light intensity value of the second sample output by the recursive formula. Input the actual light intensity value of the second sample and the theoretical grayscale value of the sample to be traversed in the next iteration into the recursive formula to obtain the actual light intensity value of the third sample output by the recursive formula. Repeat the above steps until the last theoretical grayscale value of the last sample in the theoretical grayscale sequence is traversed to obtain the actual light intensity value of the last sample. The last sample's actual light intensity value is used as the new initial actual light intensity value to continue iterating until the preset number of iterations is reached, and finally the light intensity sequence label is obtained.

[0014] The present invention also provides a method for displaying images on a liquid crystal panel, comprising the following steps.

[0015] Obtain the target theoretical grayscale sequence for the 3D image; The target theoretical grayscale sequence is input into the light intensity prediction model to obtain the target light intensity sequence output by the light intensity prediction model; When the target theoretical grayscale sequence is adjusted based on the target light intensity sequence, the liquid crystal panel is controlled to display the three-dimensional image based on the adjusted theoretical grayscale sequence. The light intensity prediction model is obtained by training based on any of the methods described above.

[0016] According to a method for displaying an image on a liquid crystal panel provided by the present invention, the method further includes: If it is determined, based on the target light intensity sequence, that the target theoretical grayscale sequence should not be adjusted, the liquid crystal panel is controlled to display the three-dimensional image based on the target theoretical grayscale sequence.

[0017] The present invention also provides a training device for a light intensity prediction model of a liquid crystal panel, comprising: The first acquisition unit is used to acquire the theoretical grayscale sequence of samples for the liquid crystal panel; The first prediction unit is used to input the theoretical grayscale sequence of the sample into the initial light intensity prediction model to obtain the light intensity prediction sequence output by the initial light intensity prediction model. The training unit is used to iteratively adjust the model parameters of the initial light intensity prediction model based on the light intensity prediction sequence and the light intensity sequence labels corresponding to the theoretical grayscale sequence of the samples, so as to obtain the light intensity prediction model. The light intensity sequence labels are obtained by inputting the theoretical grayscale sequence of the samples into a recursive formula. The recursive formula is determined based on triplet data. The triplet data includes the actual light intensity value of the previous frame, the theoretical grayscale value of the current frame, and the actual light intensity value of the current frame. The actual light intensity value of the previous frame and the actual light intensity value of the current frame are obtained by measuring the liquid crystal panel based on the light measurement module.

[0018] The present invention also provides a screen display device for a liquid crystal panel, comprising: The second acquisition unit is used to acquire the target theoretical grayscale sequence for the three-dimensional image; The second prediction unit is used to input the target theoretical grayscale sequence into the light intensity prediction model to obtain the target light intensity sequence output by the light intensity prediction model. The first control unit is configured to control the liquid crystal panel to display the three-dimensional image based on the adjusted theoretical grayscale sequence when the target theoretical grayscale sequence is adjusted based on the target light intensity sequence. The light intensity prediction model is obtained by training based on any of the methods described above.

[0019] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements either the light intensity prediction model training method for a liquid crystal panel as described above, or the image display method for a liquid crystal panel as described above.

[0020] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements either the light intensity prediction model training method for a liquid crystal panel as described above, or the image display method for a liquid crystal panel as described above.

[0021] The present invention also provides a computer program product, including a computer program, which, when executed by a processor, implements either the light intensity prediction model training method for a liquid crystal panel as described above, or the image display method for a liquid crystal panel as described above.

[0022] The present invention provides a method for training a light intensity prediction model for a liquid crystal panel and a method for displaying an image. The method involves obtaining a sample theoretical grayscale sequence for the liquid crystal panel, inputting this sequence into an initial light intensity prediction model to obtain the output light intensity prediction sequence. Based on the light intensity prediction sequence and the light intensity sequence labels corresponding to the sample theoretical grayscale sequence, the model parameters of the initial light intensity prediction model are iteratively adjusted to obtain the light intensity prediction model. The light intensity sequence labels are obtained by inputting the sample theoretical grayscale sequence into a recursive formula. This recursive formula is determined based on triplet data, which includes the actual light intensity value of the previous frame, the theoretical grayscale value of the current frame, and the actual light intensity value of the current frame. As can be seen, the present invention measures the actual light intensity value of the previous frame and the actual light intensity value of the current frame by using a light measurement module to measure the liquid crystal panel. Then, based on the triplet data composed of the actual light intensity value of the previous frame, the theoretical grayscale value of the current frame, and the actual light intensity value of the current frame, a recursive relation is determined. Based on the recursive relation, the light intensity sequence label required to train the initial light intensity prediction model is obtained. Finally, a light intensity prediction model for the liquid crystal panel is trained. The whole process does not depend on the internal material parameters of the liquid crystal panel, thereby improving the controllability of the light intensity prediction of the liquid crystal panel. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in this 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 some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0024] Figure 1 This is one of the flowcharts illustrating the light intensity prediction model training method for liquid crystal panels provided in this embodiment of the invention.

[0025] Figure 2 This is the second flowchart illustrating the light intensity prediction model training method for liquid crystal panels provided in this embodiment of the invention.

[0026] Figure 3 This is a schematic diagram of the single-frame light intensity value obtained after integration provided in this embodiment.

[0027] Figure 4 This is a schematic diagram of the triplet data provided in this embodiment.

[0028] Figure 5 This is a flowchart illustrating a screen display method for a liquid crystal panel provided in an embodiment of the present invention.

[0029] Figure 6 This is a schematic diagram of the structure of the light intensity prediction model training device for liquid crystal panels provided in an embodiment of the present invention.

[0030] Figure 7 This is a schematic diagram of the structure of a screen display device for a liquid crystal panel provided in an embodiment of the present invention.

[0031] Figure 8 This is a schematic diagram of the physical structure of the electronic device provided in an embodiment of the present invention. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0033] The following is combined with Figures 1 to 4 This invention describes a method for training a light intensity prediction model for a liquid crystal panel. The subject executing this method can be an electronic device such as a terminal, tablet computer, computer, or server, or a light intensity prediction model training device for a liquid crystal panel installed in that electronic device. This device can be implemented through software, hardware, or a combination of both.

[0034] Figure 1 This is one of the flowcharts illustrating the light intensity prediction model training method for liquid crystal panels provided in this embodiment of the invention, such as... Figure 1 As shown, the training method for the light intensity prediction model of a liquid crystal panel includes the following steps: Step 101: Obtain the theoretical grayscale sequence of the sample for the LCD panel.

[0035] The LCD panel can be a high refresh rate LCD panel or a low refresh rate LCD panel. The high refresh rate is usually not lower than 120Hz (Hz), and the low refresh rate is usually not higher than 60Hz. This invention does not limit this.

[0036] For example, a sample theoretical grayscale sequence including multiple sample theoretical grayscale values ​​can be randomly generated. The length of the sample theoretical grayscale sequence can be M, that is, the sample theoretical grayscale sequence includes M frames of sample theoretical grayscale values, where M is an integer greater than or equal to 1. The number of sample theoretical grayscale sequences can be one or more, and the present invention does not limit this.

[0037] Step 102: Input the theoretical grayscale sequence of the sample into the initial light intensity prediction model to obtain the light intensity prediction sequence output by the initial light intensity prediction model.

[0038] For example, the obtained theoretical grayscale sequences of each sample are used as training datasets and input into the initial light intensity prediction model to obtain the light intensity prediction sequences corresponding to the theoretical grayscale sequences of each sample output by the initial light intensity prediction model.

[0039] It should be noted that the initial light intensity prediction model can be a lightweight multilayer perceptron (MLP) network. The activation function of the lightweight MLP network can be a rectified linear unit (ReLU), the output layer can be a linear layer, and the hidden layer dimension is preferably 128-256. Furthermore, the number of parameters in the lightweight MLP network can be adjusted based on the actual scenario requirements. For example, for embedded deployment scenarios (such as low-power graphics processing units (GPUs)), the hidden layers of the lightweight MLP network can be compressed to 64-32 nodes or INT8 quantization can be used, sacrificing some accuracy for lower inference overhead. Preferably, the hidden layer dimension of the lightweight MLP network is 128, which can ensure high fitting accuracy (test grayscale deviation less than 0.02) while maintaining low inference overhead (memory usage <500M). This invention does not impose any limitations on this.

[0040] Step 103: Based on the light intensity prediction sequence and the light intensity sequence labels corresponding to the sample theoretical grayscale sequence, iteratively adjust the model parameters of the initial light intensity prediction model to obtain the light intensity prediction model. The light intensity sequence labels are obtained by inputting the sample theoretical grayscale sequence into the recursive formula. The recursive formula is determined based on triplet data. The triplet data includes the actual light intensity value of the previous frame, the theoretical grayscale value of the current frame, and the actual light intensity value of the current frame. The actual light intensity value of the previous frame and the actual light intensity value of the current frame are obtained by measuring the liquid crystal panel using the light measurement module.

[0041] For example, after obtaining the light intensity prediction sequence output by the initial light intensity prediction model, a loss function is constructed based on the light intensity prediction sequence and the light intensity sequence labels corresponding to the theoretical grayscale sequences of the samples. Here, the light intensity sequence labels are obtained by inputting the theoretical grayscale sequences of the samples into a recursive relation determined based on triplet data. Each sample's theoretical grayscale sequence and its corresponding light intensity sequence label constitute a supervised learning dataset. The format of each sample is: input vector = [S1, S2, ..., S...]. M The output vector is [x1, x2, ..., x]. M The input vector refers to the theoretical grayscale sequence of the samples, including the theoretical grayscale values ​​of M frames of samples, S. M The output vector represents the theoretical grayscale value of the M-th frame sample, and refers to the light intensity sequence label corresponding to the theoretical grayscale sequence of the samples, including the light intensity label corresponding to each of the M-frame sample theoretical grayscale values, x.M This represents the light intensity label corresponding to the theoretical grayscale value of the sample in frame M.

[0042] The model parameters of the initial light intensity prediction model are iteratively adjusted based on the loss function until the convergence condition is met, and the final light intensity prediction model is trained. The loss function can be represented by the mean squared error. , Indicates mean square error. This represents the light intensity prediction sequence corresponding to the theoretical grayscale sequence of the t-th sample. This represents the light intensity sequence label corresponding to the theoretical grayscale sequence of the t-th sample. The number of theoretical grayscale sequences in the sample can be represented by the Adam optimizer, trained until the loss is stable, and an early stopping strategy can be used to prevent overfitting. After training, the prediction accuracy of the light intensity prediction model should not exceed 0.01 light intensity values ​​relative to the recursive relation.

[0043] It should be noted that this invention does not directly use the established recursive relation to optimize multi-frame display content. Instead, it uses the recursive relation to generate light intensity sequence labels corresponding to the theoretical grayscale sequences of the samples. The theoretical grayscale sequences of each sample and their corresponding light intensity sequence labels are used as the supervised learning dataset for training the initial light intensity prediction model. By constructing and training the initial light intensity prediction model, a paradigm shift from iterative recursion to single forward propagation is achieved, increasing the inference speed by at least 100%, while maintaining the prediction error below 0.02. This method has advantages such as high inference efficiency, low deployment difficulty, and hardware friendliness. Furthermore, its training objective directly serves multi-frame light intensity prediction, making it a key technical path connecting accurate recursion and efficient inference.

[0044] The present invention provides a method for training a light intensity prediction model for a liquid crystal panel. This method obtains a sample theoretical grayscale sequence for the liquid crystal panel, inputs this sample theoretical grayscale sequence into an initial light intensity prediction model, and obtains the light intensity prediction sequence output by the initial light intensity prediction model. Based on the light intensity sequence labels corresponding to the light intensity prediction sequence and the sample theoretical grayscale sequence, the model parameters of the initial light intensity prediction model are iteratively adjusted to obtain the light intensity prediction model. The light intensity sequence labels are obtained by inputting the sample theoretical grayscale sequence into a recursive formula. The recursive formula is determined based on triplet data, which includes the actual light intensity value of the previous frame, the theoretical grayscale value of the current frame, and the actual light intensity value of the current frame. As can be seen, the present invention measures the actual light intensity value of the previous frame and the actual light intensity value of the current frame by using a light measurement module to measure the liquid crystal panel. Then, based on the triplet data composed of the actual light intensity value of the previous frame, the theoretical grayscale value of the current frame, and the actual light intensity value of the current frame, a recursive relation is determined. Based on the recursive relation, the light intensity sequence label required to train the initial light intensity prediction model is obtained. Finally, a light intensity prediction model for the liquid crystal panel is trained. The whole process does not depend on the internal material parameters of the liquid crystal panel, thereby improving the controllability of the light intensity prediction of the liquid crystal panel.

[0045] In one embodiment, the recursive relation is determined as follows: For each of the triplet data, the actual light intensity value of the previous frame and the theoretical grayscale value of the current frame in the triplet data are used as the independent variables of the polynomial, and the actual light intensity value of the current frame in the triplet data is used as the dependent variable of the polynomial; based on the independent and dependent variables corresponding to each of the triplet data, the coefficient matrix of the polynomial is fitted; based on the coefficient matrix, the recursive relation is determined.

[0046] For example, construct a polynomial ,in, and To represent the order of a polynomial, preferably, The maximum value is 3. The maximum value is 5, that is Highest level 3 The highest level is 5. For each triplet data set, the actual light intensity value of the previous frame in the triplet data set is used. and the theoretical grayscale value of the current frame As the independent variable of the polynomial, and using the actual light intensity value of the current frame in the triplet data as the independent variable. The coefficient matrix of the polynomial is obtained by fitting the polynomial using the least squares method, with the coefficients being the dependent variable. The fitted coefficient matrix Substituting into the polynomial formula yields the explicit recurrence relation. This implementation method offers advantages such as high computational efficiency, analytical differentiability, and ease of hardware deployment. The input to this recurrence relation is the actual light intensity value of the previous frame. and the theoretical grayscale value of the current frame The actual light intensity value of the previous frame As the memory state of the current frame, and the theoretical grayscale value of the current frame. Together, they determine the output of the actual light intensity prediction value for the current frame, and the final output is the actual light intensity prediction value z. pred It can be based on the actual light intensity prediction value z pred Construct light intensity sequence labels.

[0047] It should be noted that, in addition to polynomial fitting, other fitting methods such as radial basis function (RBF) networks and Gaussian process regression (GPR) can also be used to establish the mapping of z = f(x, y) for the fitting process of the recursive relation. However, this invention prefers polynomial fitting, which can achieve a fitting effect close to the ideal state. The recursive relation is obtained based on polynomial fitting, and the coefficient of determination R2 ≈ 0.997 for the fitting process.

[0048] In this embodiment, a recursive relationship between frames can be established based on triplet data and polynomials, realizing the automatic determination of the recursive relationship. This facilitates the automatic generation of light intensity sequence labels corresponding to the theoretical grayscale sequence of the samples based on the recursive relationship. Furthermore, the theoretical grayscale sequence of the samples and the corresponding light intensity sequence labels are used as the supervised learning dataset for training the initial light intensity prediction model, thereby improving the efficiency and accuracy of obtaining the supervised learning dataset.

[0049] In one embodiment, Figure 2 This is a second schematic flowchart of the light intensity prediction model training method for liquid crystal panels provided in this embodiment of the invention, as shown below. Figure 2 As shown, prior to step 101 above, the method for training the light intensity prediction model for a liquid crystal panel further includes the following steps: Step 201: Generate a grayscale transition sequence that includes multiple grayscale values.

[0050] For example, a grayscale transfer sequence including multiple grayscale values ​​can be randomly generated for any image that the LCD panel needs to display. For example, the normalized grayscale space can be subdivided into grayscale transfer sequences G with a preset interval, such as a preset interval of 0.1, and the grayscale transfer sequence G = [0, 0.1, 0.2, 0.3, 0.4, 0.5...].

[0051] Step 202: Generate multiple grayscale transition pairs based on the grayscale transition sequence, wherein each grayscale transition pair includes any two target grayscale values ​​in the grayscale transition sequence.

[0052] For example, two target grayscale values ​​are randomly selected in the grayscale transition sequence G to form a grayscale transition pair (i→j), where i represents the theoretical grayscale value of the previous frame and j represents the theoretical grayscale value of the current frame.

[0053] It should be noted that any two target gray levels in the gray level transfer sequence G can form a gray level transfer pair. The more gray level transfer pairs there are, the better, so as to cover all target gray levels in the gray level transfer sequence.

[0054] Step 203: For each grayscale transfer pair, control the liquid crystal panel to alternately display a preset transfer time based on the two target grayscale values ​​in the grayscale transfer pair.

[0055] For example, for each grayscale transfer pair, the two target grayscale values ​​in the grayscale transfer pair are alternately sent to the LCD panel via the communication interface as driving signals. Taking the grayscale transfer pair (i→j) as an example, the theoretical grayscale value i of the previous frame is sent to the LCD panel first, then the theoretical grayscale value j of the current frame is sent to the LCD panel, then the theoretical grayscale value i of the previous frame is sent to the LCD panel again, and then the theoretical grayscale value j of the current frame is sent to the LCD panel again. The two target grayscale values ​​in the grayscale transfer pair are sent alternately in this way, so that the LCD panel displays based on the two alternately received target grayscale values, ensuring that each grayscale transfer pair alternately displays for a preset transfer time, for example, the preset transfer time is 1 second, to reduce the impact of noise.

[0056] Step 204: Within the preset transfer time, the light intensity value of the displayed liquid crystal panel is collected by the light measurement module.

[0057] For example, when the LCD panel displays two target grayscale values ​​alternately based on the grayscale transfer pair, the light measurement module is synchronously triggered to collect the light intensity value of the LCD panel after display within a preset transfer time. The light intensity value here can be understood as the brightness measurement value in the normalized 0-1 grayscale range. The light measurement module can be a photometer.

[0058] It should be noted that when collecting light intensity values ​​through the light measurement module, the LCD panel needs to be placed in a dark room environment, and then the detection port of the light measurement module should be placed close to the screen of the LCD panel. The sampling resolution Fs of the light measurement module needs to be set to ensure that it can capture the transient details of the light intensity. For example, the sampling resolution Fs of the light measurement module is 4000Hz. The light measurement module is connected to the electronic device, and when the light intensity value is collected, it is synchronized to the electronic device.

[0059] It should be noted that when the light measurement module collects light intensity values, it can use uniform sampling, that is, sampling within the normalized grayscale value range of 0 to 1 with a fixed step size, or it can use non-uniform sampling, that is, sampling within the normalized grayscale value range of 0 to 1 with different step sizes. However, theoretically, the more sufficient the grayscale sampling, the higher the accuracy of the final fitted recursive relationship. This invention does not limit this.

[0060] Step 205: Determine the light intensity response curve based on the light intensity value corresponding to each grayscale transfer pair.

[0061] For example, when the electronic device receives the light intensity values ​​corresponding to each grayscale transition pair sent by the optical measurement module, it arranges the light intensity values ​​based on the corresponding grayscale values ​​and plots the light intensity response curve.

[0062] Step 206: Based on the light intensity response curve, determine the triplet data corresponding to each grayscale transition pair.

[0063] For example, when obtaining the light intensity response curve, the light intensity response curve is segmented, and each segment corresponds to a gray level transition pair. Based on the segment, the actual light intensity value of the previous frame and the actual light intensity value of the current frame corresponding to the gray level transition pair are calculated. Then, the three data of the actual light intensity value of the previous frame, the actual light intensity value of the current frame, and the theoretical gray level of the current frame are combined to form the triplet data corresponding to the gray level transition pair. Using the same method, the triplet data corresponding to each gray level transition pair can be determined.

[0064] In this embodiment, the liquid crystal panel is controlled to alternately display two target grayscale values ​​in the grayscale transfer pair for a preset transfer time, so that the light measurement module can collect the light intensity value of the liquid crystal panel after display and finally determine the light intensity response curve. Based on the light intensity response curve, the triplet data corresponding to each grayscale transfer pair is determined, realizing the automatic acquisition of triplet data. Moreover, since the liquid crystal panel alternately displays two target grayscale values ​​in the grayscale transfer pair, the accuracy of triplet data acquisition can be improved.

[0065] In one embodiment, step 206 above determines the triplet data corresponding to each grayscale transition pair based on the light intensity response curve, which can be implemented in the following way: The light intensity response curve is segmented based on the preset transfer time to obtain multiple segments; for each segment, denoising processing is performed to obtain denoised segments; the light intensity values ​​collected within the same pixel time period in the denoised segments are integrated to obtain the single-frame light intensity value corresponding to each same pixel time period, wherein the same pixel time period is determined based on the sampling resolution of the light measurement module and the refresh rate of the liquid crystal panel; based on the single-frame light intensity values ​​corresponding to each denoised segment, the triplet data corresponding to the grayscale transfer pair is determined.

[0066] Optionally, the triplet data corresponding to the grayscale transition pair can be determined based on the light intensity values ​​of each single frame corresponding to the denoised segment, which can be achieved in the following way: Based on the light intensity values ​​of all single frames corresponding to the denoised segments, a reference value is determined; based on the light intensity values ​​of all single frames corresponding to the denoised segments and the reference value, the actual light intensity value of the previous frame and the actual light intensity value of the current frame are determined; the actual light intensity value of the previous frame, the actual light intensity value of the current frame, and the theoretical grayscale of the current frame are determined as the triplet data corresponding to the grayscale transition pair.

[0067] For example, the light intensity response curve is segmented according to a preset transition time to obtain multiple segments. For each segment, denoising is performed, preferably using percentage-based denoising to retain valid data, resulting in denoised segments. Then, considering the refresh rate of the LCD panel, the light intensity values ​​collected within the same pixel time period T (e.g., if the sampling resolution of the light measurement module is Fs and the refresh rate of the LCD panel is S, then T = Fs / S) are integrated, for example using mean integration, to obtain the single-frame light intensity value corresponding to each same pixel time period. Here, the single-frame light intensity value includes the single-frame light intensity value of the previous frame and the single-frame light intensity value of the current frame. Figure 3 This is a schematic diagram of the single-frame light intensity value obtained after integration provided in this embodiment. For each preset transition time, that is, for each grayscale transition pair, a reference value is determined based on all single-frame light intensity values ​​corresponding to the denoised segment. Then, based on the comparison between all single-frame light intensity values ​​corresponding to the denoised segment and the reference value, the actual light intensity value of the previous frame and the actual light intensity value of the current frame are determined. Subsequently, the actual light intensity value x of the previous frame, the actual light intensity value z of the current frame, and the corresponding theoretical grayscale y of the current frame are determined as the triplet data (x, y, z) corresponding to the grayscale transition pair. Figure 4 This is a schematic diagram of the triplet data provided in this embodiment, where each point corresponds to a triplet data.

[0068] In this embodiment, the light intensity response curve is segmented based on a preset transition time to obtain multiple segments. The segments are then denoised to obtain denoised segments. Triple data is then determined based on the light intensity values ​​of each single frame corresponding to the denoised segments. This avoids the influence of noise and improves the accuracy of triple data determination.

[0069] In one embodiment, the determination of the actual light intensity value of the previous frame and the actual light intensity value of the current frame based on all single-frame light intensity values ​​corresponding to the denoised segments and the reference value can be achieved in the following way: If the theoretical grayscale value of the previous frame is less than or equal to the theoretical grayscale value of the current frame, then among all the single-frame light intensity values ​​corresponding to the denoised segment, determine all the previous frame's single-frame light intensity values ​​that are less than or equal to the reference value, and determine the average of all the previous frame's single-frame light intensity values ​​that are less than or equal to the reference value as the previous frame's actual light intensity value; if the theoretical grayscale value of the previous frame is greater than the current frame's theoretical grayscale value, then among all the single-frame light intensity values ​​corresponding to the denoised segment, determine all the previous frame's single-frame light intensity values ​​that are greater than the reference value, and determine the average of all the previous frame's single-frame light intensity values ​​that are greater than the reference value as the previous frame's actual light intensity value.

[0070] For example, among all single-frame light intensity values ​​corresponding to the denoised segment, the maximum and minimum single-frame light intensity values ​​can be selected. Then, the average of the maximum and minimum single-frame light intensity values ​​is taken to obtain a reference value. If the theoretical grayscale value of the previous frame is less than or equal to the theoretical grayscale value of the current frame, all previous frame light intensity values ​​less than or equal to the reference value are selected from all single-frame light intensity values ​​corresponding to the denoised segment. The average of these previous frame light intensity values ​​is taken as the actual light intensity value of the previous frame. If the theoretical grayscale value of the previous frame is greater than the theoretical grayscale value of the current frame, all previous frame light intensity values ​​greater than the reference value are determined from all single-frame light intensity values ​​corresponding to the denoised segment. The average of these previous frame light intensity values ​​is taken as the actual light intensity value of the previous frame.

[0071] It should be noted that the median of all single-frame light intensity values ​​corresponding to the denoised segments can also be used as a reference value, but this invention does not limit this.

[0072] If the theoretical grayscale value of the previous frame is less than or equal to the theoretical grayscale value of the current frame, then among all the single-frame light intensity values ​​corresponding to the denoised segment, determine all the single-frame light intensity values ​​of the current frame that are greater than the reference value, and determine the average value of the single-frame light intensity values ​​of the current frame that are greater than the reference value as the actual light intensity value of the current frame; if the theoretical grayscale value of the previous frame is greater than the theoretical grayscale value of the current frame, then among all the single-frame light intensity values ​​corresponding to the denoised segment, determine all the single-frame light intensity values ​​of the current frame that are less than or equal to the reference value, and determine the average value of the single-frame light intensity values ​​of the current frame that are less than or equal to the reference value as the actual light intensity value of the current frame.

[0073] For example, if the theoretical grayscale value of the previous frame is less than or equal to the theoretical grayscale value of the current frame, select all single-frame light intensity values ​​of the current frame that are greater than the reference value from all single-frame light intensity values ​​corresponding to the denoised segment, and use the average of all single-frame light intensity values ​​of the current frame that are greater than the reference value as the actual light intensity value of the current frame. If the theoretical grayscale value of the previous frame is greater than the theoretical grayscale value of the current frame, determine all single-frame light intensity values ​​of the current frame that are less than or equal to the reference value from all single-frame light intensity values ​​corresponding to the denoised segment, and use the average of all single-frame light intensity values ​​of the current frame that are less than or equal to the reference value as the actual light intensity value of the current frame.

[0074] In this embodiment, the actual light intensity value of the previous frame and the actual light intensity value of the current frame can be determined based on the analysis of all single-frame light intensity values ​​corresponding to the denoised segments, thus realizing the automatic determination of the actual light intensity value of the previous frame and the actual light intensity value of the current frame.

[0075] In one embodiment, prior to step 103 above, the light intensity prediction model training method for the liquid crystal panel further includes the following steps: Starting from the first theoretical grayscale value in the theoretical grayscale sequence, the system iterates through the data, inputting the currently iterated theoretical grayscale value and the initial actual light intensity value into a recursive formula to obtain the first actual light intensity value output by the recursive formula. The first actual light intensity value and the next iterated theoretical grayscale value are then input into the recursive formula to obtain the second actual light intensity value output by the recursive formula. The second actual light intensity value and the next-next iterated theoretical grayscale value are then input into the recursive formula to obtain the third actual light intensity value output by the recursive formula. This process is repeated until the last theoretical grayscale value in the theoretical grayscale sequence is traversed, resulting in the last actual light intensity value. This last actual light intensity value is then used as the new initial actual light intensity value to continue iterating until a preset number of iterations is reached, ultimately obtaining the light intensity sequence label.

[0076] For example, given a sample theoretical grayscale sequence [S1, S2, ..., S...] M The initial actual light intensity is set to the light intensity value of the LCD panel at black level. Typically, the initial actual light intensity is 0. For t=1 to t=M, execute: S t This represents the theoretical grayscale value of the sample in frame t. This represents the actual light intensity value of frame t-1. When traversing frame 1, the theoretical grayscale value of the currently traversed sample is the theoretical grayscale value of the sample in frame 1. , This represents the initial actual light intensity value. and Substituting into the above recursive formula, we obtain the actual light intensity value of the first frame sample output by the recursive formula. That is, the actual light intensity value of the first sample; when traversing the second frame, the theoretical grayscale value of the currently traversed sample is the theoretical grayscale value of the sample in the second frame. , This represents the actual light intensity value of the first sample. and Substituting into the above recursive formula, we obtain the actual light intensity value of the second frame sample output by the recursive formula. That is, the actual light intensity value of the second sample; when traversing the third frame, the theoretical grayscale value of the currently traversed sample is the theoretical grayscale value of the sample in the third frame. , This represents the actual light intensity value of the second sample. and Substituting into the above recursive formula, we obtain the actual light intensity value of the third frame sample output by the recursive formula. This is the actual light intensity value of the third sample. Repeat the above steps until the theoretical grayscale value of the last sample in the theoretical grayscale sequence has been traversed, and the actual light intensity value of the last sample is obtained.

[0077] Upon obtaining the actual light intensity value of the last sample, this value is used as the new initial actual light intensity value to repeat the above traversal process. This process is repeated k times (a preset number of iterations), for example, k > 3, to ensure iterative stability. The final output is the light intensity sequence label [x1, x2, ..., x...]. M ].

[0078] In this embodiment, an iterative approach is used to achieve a complete dynamic description of pixel response under the theoretical grayscale sequence of consecutive multi-frame samples, and finally obtains light intensity sequence labels to improve the accuracy of light intensity sequence labels, and further improve the accuracy of light intensity prediction model trained based on light intensity sequence labels.

[0079] Figure 5 This is a flowchart illustrating a screen display method for a liquid crystal panel provided in an embodiment of the present invention, as shown below. Figure 5 As shown, the method for displaying images on a liquid crystal panel includes the following steps: Step 501: Obtain the target theoretical grayscale sequence for the 3D image.

[0080] For example, relevant personnel can initially set the target theoretical grayscale sequence based on the 3D image to be displayed.

[0081] Step 502: Input the target theoretical grayscale sequence into the light intensity prediction model to obtain the target light intensity sequence output by the light intensity prediction model.

[0082] The light intensity prediction model is trained based on any of the methods described above.

[0083] For example, a pre-trained light intensity prediction model is embedded in a liquid crystal simulation display software. The model parameters of the light intensity prediction model are frozen. When the initially set target theoretical grayscale sequence is obtained, the initially set target theoretical grayscale sequence is input into the pre-trained light intensity prediction model. The light intensity prediction model performs forward simulation on the target theoretical grayscale sequence to obtain the target light intensity sequence corresponding to the target theoretical grayscale sequence output by the light intensity prediction model. This realizes the mapping and conversion from the target theoretical grayscale sequence to the actual light intensity sequence. Experimental results show that the efficiency of multi-frame prediction using the light intensity prediction model is significantly improved compared with multi-round iterative prediction using recursive formulas.

[0084] Step 503: When the target theoretical grayscale sequence is adjusted based on the target light intensity sequence, the liquid crystal panel is controlled to display the three-dimensional image based on the adjusted theoretical grayscale sequence.

[0085] For example, after obtaining the target light intensity sequence output by the light intensity prediction model, the target light intensity sequence is displayed so that relevant personnel can determine whether the target light intensity sequence matches the three-dimensional image to be displayed. If it is determined that the target light intensity sequence does not match the three-dimensional image to be displayed, the initially set target theoretical grayscale sequence needs to be adjusted based on the target light intensity sequence, and the adjusted theoretical grayscale sequence is sent to the LCD panel so that the LCD panel displays the three-dimensional image based on the adjusted theoretical grayscale sequence.

[0086] It should be noted that when it is determined that the target light intensity sequence does not match the required 3D image, the initially set target theoretical grayscale sequence can be initially adjusted based on the target light intensity sequence to obtain the initially adjusted theoretical grayscale sequence. This initially adjusted theoretical grayscale sequence is then input back into the light intensity prediction model to obtain a new light intensity sequence output by the model. If, based on the new light intensity sequence, it is determined that no further adjustment to the initially adjusted theoretical grayscale sequence is needed, the LCD panel is controlled to display the 3D image based on the initially adjusted theoretical grayscale sequence. If, based on the new light intensity sequence, it is determined that further adjustment to the initially adjusted theoretical grayscale sequence is necessary, the LCD panel will display the 3D image based on the initially adjusted theoretical grayscale sequence. When the theoretical grayscale sequence is adjusted, the initially adjusted theoretical grayscale sequence is adjusted again based on the new light intensity sequence until no position adjustment is needed, resulting in the final adjusted theoretical grayscale sequence. Then, the LCD panel is controlled to display a three-dimensional image based on the final adjusted theoretical grayscale sequence, so as to realize the closed loop of target theoretical grayscale sequence → light intensity prediction model prediction → error calculation → gradient descent backpropagation → target theoretical grayscale sequence optimization. This makes the three-dimensional image display shift from passive compensation to active prediction, solving the problem of depth layer aliasing caused by the inability of existing technology to predict multiple light intensity sequences, and realizing precise control of the light field.

[0087] The present invention provides a display method for a liquid crystal panel, which acquires a target theoretical grayscale sequence for a three-dimensional image, inputs the target theoretical grayscale sequence into a light intensity prediction model, obtains a target light intensity sequence output by the light intensity prediction model, and controls the liquid crystal panel to display a three-dimensional image based on the adjusted theoretical grayscale sequence after determining the adjustment based on the target light intensity sequence. It is understood that the present invention measures the actual light intensity value of the previous frame and the actual light intensity value of the current frame using a light measurement module, then determines a recursive relationship based on the triplet data composed of the actual light intensity value of the previous frame, the theoretical grayscale value of the current frame, and the actual light intensity value of the current frame, and obtains the light intensity sequence label required to train the initial light intensity prediction model based on the recursive relationship. Finally, a light intensity prediction model for the liquid crystal panel is trained, and the trained light intensity prediction model can directly fit the mapping relationship between the target theoretical grayscale sequence and the target light intensity sequence. The entire process does not require iterative loops, thereby accelerating the optimization process of the three-dimensional image content, providing pixel-level light intensity prediction support for deep layer calculations, and achieving a true closed loop between display effect and simulation prediction.

[0088] In one embodiment, the image display method for a liquid crystal panel further includes the following steps: If it is determined, based on the target light intensity sequence, that the target theoretical grayscale sequence should not be adjusted, the liquid crystal panel is controlled to display the three-dimensional image based on the target theoretical grayscale sequence.

[0089] For example, when it is determined that the target light intensity sequence matches the three-dimensional image to be displayed, there is no need to adjust the initially set target theoretical grayscale sequence based on the target light intensity sequence. The LCD panel can be directly controlled to display the three-dimensional image based on the target theoretical grayscale sequence.

[0090] In summary, the present invention has the following beneficial effects: (1) It has achieved independent and controllable pixel response modeling capability. This invention completely eliminates the dependence on the internal material parameters of the liquid crystal panel by constructing a recursive relationship through direct external light intensity value measurement. The model can be established using only a conventional light measurement module, which transfers the initiative of modeling from the panel manufacturer to the application end and significantly improves the technological autonomy of the downstream of the industry chain.

[0091] (2) This invention provides the first accurate characterization of the cumulative coupling effect of multiple historical states at high refresh rates. Traditional techniques only consider the state of a single previous frame, while this invention introduces the actual light intensity value of the previous frame as an explicit state variable into the recursive formula, fully describing the joint interference of multiple preceding frames (n-1, n-2, ...) on the current frame through the residual deflection angle of liquid crystal molecules. Experimental data show that in a 240Hz / 6-frame continuous modulation scenario, the mean absolute error (MAE) between the light intensity sequence predicted by this invention and the measured value (normalized value) is less than 0.02. This method has a significant improvement effect on the problems of depth-level aliasing and crosstalk rate degradation caused by inter-frame coupling errors in multi-frame liquid crystal 3D displays.

[0092] (3) This invention uses MLP to transform multi-frame sequence prediction from iterative calculation to single forward propagation. While maintaining an error of less than 0.02, the measured time for single-sequence inference of 6 frames on a single RTX 4090 graphics card is reduced from 7.2ms to 4.2ms. This is particularly advantageous when dealing with high-resolution images with large computational demands. This effect enables efficient optimization and iteration of coding strategies during the 3D image display optimization stage, improving development efficiency.

[0093] (4) This invention has strong generalization ability across panels and operating conditions. Since the model building process only requires external light intensity values, this invention can quickly measure and build a light intensity prediction model for any type of liquid crystal panel, such as twisted nematic (TN) liquid crystal panels, vertical alignment (VA) liquid crystal panels, in-plane switching (IPS) liquid crystal panels, any refresh rate (60Hz-540Hz) and even different operating temperatures (-10℃~60℃).

[0094] (5) This invention improves pixel response error from passive compensation to active prediction. It does not rely entirely on screen response interval, but optimizes the display content by modeling inter-frame correlation, so that the three-dimensional screen display can directly complete the light field pre-calculation and error compensation in the digital domain.

[0095] (6) This invention has the potential for modularization and standardization, and is easy to integrate into industries. The three stages of data acquisition, model fitting and collaborative reasoning in this invention are clearly decoupled and can be encapsulated as a standardized application programming interface (API) or an independent software module, which can be seamlessly integrated into existing 3D display rendering pipelines, image quality adjustment toolchains or automated testing platforms, and has the characteristics of industrial promotion.

[0096] The light intensity prediction model training device for liquid crystal panels provided by the present invention is described below. The light intensity prediction model training device for liquid crystal panels described below can be referred to in correspondence with the light intensity prediction model training method for liquid crystal panels described above.

[0097] Figure 6 This is a schematic diagram of the structure of the light intensity prediction model training device for liquid crystal panels provided in an embodiment of the present invention, as shown below. Figure 6 As shown, the light intensity prediction model training device 600 for liquid crystal panels includes a first acquisition unit 601, a first prediction unit 602, and a training unit 603; wherein: The first acquisition unit 601 is used to acquire the theoretical grayscale sequence of samples for the liquid crystal panel; The first prediction unit 602 is used to input the sample theoretical grayscale sequence into the initial light intensity prediction model to obtain the light intensity prediction sequence output by the initial light intensity prediction model. Training unit 603 is used to iteratively adjust the model parameters of the initial light intensity prediction model based on the light intensity prediction sequence and the light intensity sequence labels corresponding to the theoretical grayscale sequence of the samples, to obtain the light intensity prediction model. The light intensity sequence labels are obtained by inputting the theoretical grayscale sequence of the samples into a recursive formula. The recursive formula is determined based on triplet data. The triplet data includes the actual light intensity value of the previous frame, the theoretical grayscale value of the current frame, and the actual light intensity value of the current frame. The actual light intensity value of the previous frame and the actual light intensity value of the current frame are obtained by measuring the liquid crystal panel based on the light measurement module.

[0098] The present invention provides a method for training a light intensity prediction model for a liquid crystal panel. This method obtains a sample theoretical grayscale sequence for the liquid crystal panel, inputs this sample theoretical grayscale sequence into an initial light intensity prediction model, and obtains the light intensity prediction sequence output by the initial light intensity prediction model. Based on the light intensity sequence labels corresponding to the light intensity prediction sequence and the sample theoretical grayscale sequence, the model parameters of the initial light intensity prediction model are iteratively adjusted to obtain the light intensity prediction model. The light intensity sequence labels are obtained by inputting the sample theoretical grayscale sequence into a recursive formula. The recursive formula is determined based on triplet data, which includes the actual light intensity value of the previous frame, the theoretical grayscale value of the current frame, and the actual light intensity value of the current frame. As can be seen, the present invention measures the actual light intensity value of the previous frame and the actual light intensity value of the current frame by using a light measurement module to measure the liquid crystal panel. Then, based on the triplet data composed of the actual light intensity value of the previous frame, the theoretical grayscale value of the current frame, and the actual light intensity value of the current frame, a recursive relation is determined. Based on the recursive relation, the light intensity sequence label required to train the initial light intensity prediction model is obtained. Finally, a light intensity prediction model for the liquid crystal panel is trained. The whole process does not depend on the internal material parameters of the liquid crystal panel, thereby improving the controllability of the light intensity prediction of the liquid crystal panel.

[0099] Based on any of the above embodiments, the recursive relation is determined as follows: For each of the triplet data, the actual light intensity value of the previous frame and the theoretical grayscale value of the current frame in the triplet data are used as the independent variables of the polynomial, and the actual light intensity value of the current frame in the triplet data is used as the dependent variable of the polynomial. Based on the independent and dependent variables corresponding to each triplet data, fit the coefficient matrix of the polynomial; Based on the coefficient matrix, the recursive relation is determined.

[0100] Based on any of the above embodiments, the light intensity prediction model training device 600 for liquid crystal panels further includes: The first generation unit is used to generate a grayscale transition sequence that includes multiple grayscale values; The second generation unit is used to generate multiple grayscale transition pairs based on the grayscale transition sequence, wherein the grayscale transition pair includes any two target grayscale values ​​in the grayscale transition sequence. The second control unit is used to control the liquid crystal panel to alternately display a preset transfer time based on two target grayscale values ​​in the grayscale transfer pair for each grayscale transfer pair. The acquisition unit is used to acquire the light intensity value of the displayed liquid crystal panel through the light measurement module within the preset transfer time. The first determining unit is used to determine the light intensity response curve based on the light intensity value corresponding to each grayscale transfer pair; The second determining unit is used to determine the triplet data corresponding to each grayscale transition pair based on the light intensity response curve.

[0101] Based on any of the above embodiments, the second determining unit is specifically used for: The light intensity response curve is segmented based on the preset transition time to obtain multiple segments; For each of the segments, the segments are denoised to obtain denoised segments; The light intensity values ​​collected within the same pixel time period in the denoised segment are integrated to obtain the single-frame light intensity value corresponding to each of the same pixel time periods. The same pixel time period is determined based on the sampling resolution of the light measurement module and the refresh rate of the liquid crystal panel. Based on the light intensity values ​​of each single frame corresponding to the denoised segment, the triplet data corresponding to the grayscale transition pair is determined.

[0102] Based on any of the above embodiments, the second determining unit is specifically used for: Based on the light intensity values ​​of all single frames corresponding to the denoised segments, a reference value is determined. Based on the light intensity values ​​of all single frames corresponding to the denoised segments and the reference value, the actual light intensity value of the previous frame and the actual light intensity value of the current frame are determined. The actual light intensity value of the previous frame, the actual light intensity value of the current frame, and the theoretical gray level of the current frame are determined as the triplet data corresponding to the gray level transition pair.

[0103] Based on any of the above embodiments, the second determining unit is specifically used for: If the theoretical grayscale value of the previous frame is less than or equal to the theoretical grayscale value of the current frame, determine the single-frame light intensity values ​​of the previous frame that are less than or equal to the reference value among all single-frame light intensity values ​​corresponding to the denoised segment, and determine the average value of the single-frame light intensity values ​​of the previous frame that are less than or equal to the reference value as the actual light intensity value of the previous frame. If the theoretical grayscale value of the previous frame is greater than the theoretical grayscale value of the current frame, determine the single-frame light intensity values ​​of all previous frames that are greater than the reference value among all single-frame light intensity values ​​corresponding to the denoised segment, and determine the average value of the single-frame light intensity values ​​of all previous frames that are greater than the reference value as the actual light intensity value of the previous frame. If the theoretical grayscale value of the previous frame is less than or equal to the theoretical grayscale value of the current frame, determine the single-frame light intensity values ​​of all current frames that are greater than the reference value from all single-frame light intensity values ​​corresponding to the denoised segment, and determine the average value of the single-frame light intensity values ​​of all current frames that are greater than the reference value as the actual light intensity value of the current frame. If the theoretical grayscale value of the previous frame is greater than the theoretical grayscale value of the current frame, determine the single-frame light intensity value of the current frame that is less than or equal to the reference value among all single-frame light intensity values ​​corresponding to the denoised segment, and determine the average value of the single-frame light intensity values ​​of the current frame that is less than or equal to the reference value as the actual light intensity value of the current frame.

[0104] Based on any of the above embodiments, the light intensity prediction model training device 600 for liquid crystal panels further includes: The third prediction unit is used to traverse from the first sample theoretical gray level value in the sample theoretical gray level sequence, input the currently traversed sample theoretical gray level value and the initial actual light intensity value into the recursive relation to obtain the first sample actual light intensity value output by the recursive relation. The fourth prediction unit is used to input the actual light intensity value of the first sample and the theoretical grayscale value of the sample for the next traversal into the recursive formula to obtain the actual light intensity value of the second sample output by the recursive formula. The fifth prediction unit is used to input the actual light intensity value of the second sample and the theoretical gray level value of the sample in the next iteration into the recursive formula to obtain the actual light intensity value of the third sample output by the recursive formula. The above steps are repeated until the last theoretical gray level value in the theoretical gray level sequence of the samples is traversed to obtain the actual light intensity value of the last sample. An iterative unit is used to take the actual light intensity value of the last sample as the new initial actual light intensity value and continue iterating until a preset number of iterations is reached, and finally the light intensity sequence label is obtained.

[0105] The image display device for a liquid crystal panel provided by the present invention will now be described. The image display device for a liquid crystal panel described below and the image display method for a liquid crystal panel described above can be referred to in correspondence with each other.

[0106] Figure 7 This is a schematic diagram of the structure of a screen display device for a liquid crystal panel provided in an embodiment of the present invention, as shown below. Figure 7 As shown, the image display device 700 for a liquid crystal panel includes a second acquisition unit 701, a second prediction unit 702, and a first control unit 703; wherein: The second acquisition unit 701 is used to acquire the target theoretical grayscale sequence for the three-dimensional image; The second prediction unit 702 is used to input the target theoretical grayscale sequence into the light intensity prediction model to obtain the target light intensity sequence output by the light intensity prediction model. The first control unit 703 is used to control the liquid crystal panel to display the three-dimensional image based on the adjusted theoretical grayscale sequence when the target theoretical grayscale sequence is adjusted based on the target light intensity sequence. The light intensity prediction model is obtained by training based on any of the methods described above.

[0107] The present invention provides a display method for a liquid crystal panel, which acquires a target theoretical grayscale sequence for a three-dimensional image, inputs the target theoretical grayscale sequence into a light intensity prediction model, obtains a target light intensity sequence output by the light intensity prediction model, and controls the liquid crystal panel to display a three-dimensional image based on the adjusted theoretical grayscale sequence after determining the adjustment based on the target light intensity sequence. It is understood that the present invention measures the actual light intensity value of the previous frame and the actual light intensity value of the current frame using a light measurement module, then determines a recursive relationship based on the triplet data composed of the actual light intensity value of the previous frame, the theoretical grayscale value of the current frame, and the actual light intensity value of the current frame, and obtains the light intensity sequence label required to train the initial light intensity prediction model based on the recursive relationship. Finally, a light intensity prediction model for the liquid crystal panel is trained, and the trained light intensity prediction model can directly fit the mapping relationship between the target theoretical grayscale sequence and the target light intensity sequence. The entire process does not require iterative loops, thereby accelerating the optimization process of the three-dimensional image content, providing pixel-level light intensity prediction support for deep layer calculations, and achieving a true closed loop between display effect and simulation prediction.

[0108] Based on any of the above embodiments, the image display device 700 for the liquid crystal panel further includes: The third control unit is used to control the liquid crystal panel to display the three-dimensional image based on the target theoretical grayscale sequence when it is determined, based on the target light intensity sequence, that no adjustment will be made to the target theoretical grayscale sequence.

[0109] Figure 8 This is a schematic diagram of the physical structure of the electronic device provided in the embodiments of the present invention, such as... Figure 8 As shown, the electronic device may include a processor 810, a communications interface 820, a memory 830, and a communication bus 840, wherein the processor 810, the communications interface 820, and the memory 830 communicate with each other via the communication bus 840. The processor 810 can call logic instructions in the memory 830 to execute a light intensity prediction model training method for the liquid crystal panel, the method including: acquiring a sample theoretical grayscale sequence for the liquid crystal panel; The sample theoretical grayscale sequence is input into the initial light intensity prediction model to obtain the light intensity prediction sequence output by the initial light intensity prediction model; Based on the light intensity prediction sequence and the light intensity sequence labels corresponding to the sample theoretical grayscale sequence, the model parameters of the initial light intensity prediction model are iteratively adjusted to obtain the light intensity prediction model. The light intensity sequence labels are obtained by inputting the sample theoretical grayscale sequence into a recursive formula. The recursive formula is determined based on triplet data. The triplet data includes the actual light intensity value of the previous frame, the theoretical grayscale value of the current frame, and the actual light intensity value of the current frame. The actual light intensity value of the previous frame and the actual light intensity value of the current frame are obtained by measuring the liquid crystal panel using a light measurement module.

[0110] Alternatively, you can use the following method: Obtain the target theoretical grayscale sequence for the 3D image; The target theoretical grayscale sequence is input into the light intensity prediction model to obtain the target light intensity sequence output by the light intensity prediction model; When the target theoretical grayscale sequence is adjusted based on the target light intensity sequence, the liquid crystal panel is controlled to display the three-dimensional image based on the adjusted theoretical grayscale sequence. The light intensity prediction model is obtained by training based on any of the methods described above.

[0111] Furthermore, the logical instructions in the aforementioned memory 830 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0112] On the other hand, the present invention also provides a computer program product, the computer program product including a computer program, the computer program being able to be stored on a non-transitory computer-readable storage medium, the computer program being executed by a processor, the computer being able to execute the light intensity prediction model training method for the liquid crystal panel provided by the above methods, the method including: obtaining a sample theoretical grayscale sequence for the liquid crystal panel; The sample theoretical grayscale sequence is input into the initial light intensity prediction model to obtain the light intensity prediction sequence output by the initial light intensity prediction model; Based on the light intensity prediction sequence and the light intensity sequence labels corresponding to the sample theoretical grayscale sequence, the model parameters of the initial light intensity prediction model are iteratively adjusted to obtain the light intensity prediction model. The light intensity sequence labels are obtained by inputting the sample theoretical grayscale sequence into a recursive formula. The recursive formula is determined based on triplet data. The triplet data includes the actual light intensity value of the previous frame, the theoretical grayscale value of the current frame, and the actual light intensity value of the current frame. The actual light intensity value of the previous frame and the actual light intensity value of the current frame are obtained by measuring the liquid crystal panel using a light measurement module.

[0113] Alternatively, when the program instructions are executed by the computer, the computer can implement the following method: Obtain the target theoretical grayscale sequence for the 3D image; The target theoretical grayscale sequence is input into the light intensity prediction model to obtain the target light intensity sequence output by the light intensity prediction model; When the target theoretical grayscale sequence is adjusted based on the target light intensity sequence, the liquid crystal panel is controlled to display the three-dimensional image based on the adjusted theoretical grayscale sequence. The light intensity prediction model is obtained by training based on any of the methods described above.

[0114] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a method for training a light intensity prediction model for a liquid crystal panel provided by the methods described above, the method comprising: obtaining a sample theoretical grayscale sequence for the liquid crystal panel; The sample theoretical grayscale sequence is input into the initial light intensity prediction model to obtain the light intensity prediction sequence output by the initial light intensity prediction model; Based on the light intensity prediction sequence and the light intensity sequence labels corresponding to the sample theoretical grayscale sequence, the model parameters of the initial light intensity prediction model are iteratively adjusted to obtain the light intensity prediction model. The light intensity sequence labels are obtained by inputting the sample theoretical grayscale sequence into a recursive formula. The recursive formula is determined based on triplet data. The triplet data includes the actual light intensity value of the previous frame, the theoretical grayscale value of the current frame, and the actual light intensity value of the current frame. The actual light intensity value of the previous frame and the actual light intensity value of the current frame are obtained by measuring the liquid crystal panel using a light measurement module.

[0115] Alternatively, when the computer program is executed by the processor, it implements the following method: Obtain the target theoretical grayscale sequence for the 3D image; The target theoretical grayscale sequence is input into the light intensity prediction model to obtain the target light intensity sequence output by the light intensity prediction model; When the target theoretical grayscale sequence is adjusted based on the target light intensity sequence, the liquid crystal panel is controlled to display the three-dimensional image based on the adjusted theoretical grayscale sequence. The light intensity prediction model is obtained by training based on any of the methods described above.

[0116] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0117] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0118] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for training a light intensity prediction model for a liquid crystal panel, characterized in that, include: Obtain the theoretical grayscale sequence of samples for the LCD panel; The sample theoretical grayscale sequence is input into the initial light intensity prediction model to obtain the light intensity prediction sequence output by the initial light intensity prediction model; Based on the light intensity prediction sequence and the light intensity sequence labels corresponding to the sample theoretical grayscale sequence, the model parameters of the initial light intensity prediction model are iteratively adjusted to obtain the light intensity prediction model. The light intensity sequence labels are obtained by inputting the sample theoretical grayscale sequence into a recursive formula. The recursive formula is determined based on triplet data. The triplet data includes the actual light intensity value of the previous frame, the theoretical grayscale value of the current frame, and the actual light intensity value of the current frame. The actual light intensity value of the previous frame and the actual light intensity value of the current frame are obtained by measuring the liquid crystal panel using a light measurement module.

2. The method for training a light intensity prediction model for a liquid crystal panel according to claim 1, characterized in that, The recursive formula is determined as follows: For each of the triplet data, the actual light intensity value of the previous frame and the theoretical grayscale value of the current frame in the triplet data are used as the independent variables of the polynomial, and the actual light intensity value of the current frame in the triplet data is used as the dependent variable of the polynomial. Based on the independent and dependent variables corresponding to each triplet data, fit the coefficient matrix of the polynomial; Based on the coefficient matrix, the recursive relation is determined.

3. The method for training a light intensity prediction model for a liquid crystal panel according to claim 1, characterized in that, The method further includes: Generate a grayscale transition sequence that includes multiple grayscale values; Multiple gray-level transition pairs are generated based on the gray-level transition sequence, and each gray-level transition pair includes any two target gray-level values ​​in the gray-level transition sequence; For each grayscale transfer pair, the liquid crystal panel is controlled to alternately display a preset transfer time based on the two target grayscale values ​​in the grayscale transfer pair; Within the preset transfer time, the light intensity value of the displayed liquid crystal panel is collected by the light measurement module; Based on the light intensity values ​​corresponding to each grayscale transfer pair, the light intensity response curve is determined; Based on the light intensity response curve, the triplet data corresponding to each grayscale transition pair is determined.

4. The method for training a light intensity prediction model for a liquid crystal panel according to claim 3, characterized in that, The determination of the triplet data corresponding to each grayscale transition pair based on the light intensity response curve includes: The light intensity response curve is segmented based on the preset transition time to obtain multiple segments; For each of the aforementioned segments, the segments are subjected to denoising processing to obtain denoised segments; The light intensity values ​​collected within the same pixel time period in the denoised segment are integrated to obtain the single-frame light intensity value corresponding to each of the same pixel time periods. The same pixel time period is determined based on the sampling resolution of the light measurement module and the refresh rate of the liquid crystal panel. Based on the light intensity values ​​of each single frame corresponding to the denoised segment, the triplet data corresponding to the grayscale transition pair is determined.

5. The method for training a light intensity prediction model for a liquid crystal panel according to claim 4, characterized in that, The step of determining the triplet data corresponding to the grayscale transition pair based on the light intensity values ​​of each single frame corresponding to the denoised segment includes: Based on the light intensity values ​​of all single frames corresponding to the denoised segments, a reference value is determined. Based on the light intensity values ​​of all single frames corresponding to the denoised segments and the reference value, the actual light intensity value of the previous frame and the actual light intensity value of the current frame are determined. The actual light intensity value of the previous frame, the actual light intensity value of the current frame, and the theoretical gray level of the current frame are determined as the triplet data corresponding to the gray level transition pair.

6. The method for training a light intensity prediction model for a liquid crystal panel according to claim 5, characterized in that, The step of determining the actual light intensity value of the previous frame and the actual light intensity value of the current frame based on all single-frame light intensity values ​​corresponding to the denoised segments and the reference value includes: If the theoretical grayscale value of the previous frame is less than or equal to the theoretical grayscale value of the current frame, determine the single-frame light intensity values ​​of the previous frame that are less than or equal to the reference value among all single-frame light intensity values ​​corresponding to the denoised segment, and determine the average value of the single-frame light intensity values ​​of the previous frame that are less than or equal to the reference value as the actual light intensity value of the previous frame. If the theoretical grayscale value of the previous frame is greater than the theoretical grayscale value of the current frame, determine the single-frame light intensity values ​​of all previous frames that are greater than the reference value among all single-frame light intensity values ​​corresponding to the denoised segment, and determine the average value of the single-frame light intensity values ​​of all previous frames that are greater than the reference value as the actual light intensity value of the previous frame. If the theoretical grayscale value of the previous frame is less than or equal to the theoretical grayscale value of the current frame, determine the single-frame light intensity values ​​of all current frames that are greater than the reference value from all single-frame light intensity values ​​corresponding to the denoised segment, and determine the average value of the single-frame light intensity values ​​of all current frames that are greater than the reference value as the actual light intensity value of the current frame. If the theoretical grayscale value of the previous frame is greater than the theoretical grayscale value of the current frame, determine the single-frame light intensity value of the current frame that is less than or equal to the reference value among all single-frame light intensity values ​​corresponding to the denoised segment, and determine the average value of the single-frame light intensity values ​​of the current frame that is less than or equal to the reference value as the actual light intensity value of the current frame.

7. The method for training a light intensity prediction model for a liquid crystal panel according to any one of claims 1-6, characterized in that, The method further includes: Starting from the first theoretical gray value in the theoretical gray value sequence, the system iterates through the sample, inputting the currently iterated theoretical gray value and the initial actual light intensity value into the recursive formula to obtain the first actual light intensity value of the sample output by the recursive formula. Input the actual light intensity value of the first sample and the theoretical grayscale value of the sample in the next traversal into the recursive formula to obtain the actual light intensity value of the second sample output by the recursive formula. Input the actual light intensity value of the second sample and the theoretical grayscale value of the sample to be traversed in the next iteration into the recursive formula to obtain the actual light intensity value of the third sample output by the recursive formula. Repeat the above steps until the last theoretical grayscale value of the last sample in the theoretical grayscale sequence is traversed to obtain the actual light intensity value of the last sample. The last sample's actual light intensity value is used as the new initial actual light intensity value to continue iterating until the preset number of iterations is reached, and finally the light intensity sequence label is obtained.

8. A method for displaying an image on a liquid crystal panel, characterized in that, include: Obtain the target theoretical grayscale sequence for the 3D image; The target theoretical grayscale sequence is input into the light intensity prediction model to obtain the target light intensity sequence output by the light intensity prediction model; When the target theoretical grayscale sequence is adjusted based on the target light intensity sequence, the liquid crystal panel is controlled to display the three-dimensional image based on the adjusted theoretical grayscale sequence. The light intensity prediction model is trained based on the method described in any one of claims 1-7.

9. A training device for a light intensity prediction model of a liquid crystal panel, characterized in that, include: The first acquisition unit is used to acquire the theoretical grayscale sequence of samples for the liquid crystal panel; The first prediction unit is used to input the theoretical grayscale sequence of the sample into the initial light intensity prediction model to obtain the light intensity prediction sequence output by the initial light intensity prediction model. The training unit is used to iteratively adjust the model parameters of the initial light intensity prediction model based on the light intensity prediction sequence and the light intensity sequence labels corresponding to the theoretical grayscale sequence of the samples, so as to obtain the light intensity prediction model. The light intensity sequence labels are obtained by inputting the theoretical grayscale sequence of the samples into a recursive formula. The recursive formula is determined based on triplet data. The triplet data includes the actual light intensity value of the previous frame, the theoretical grayscale value of the current frame, and the actual light intensity value of the current frame. The actual light intensity value of the previous frame and the actual light intensity value of the current frame are obtained by measuring the liquid crystal panel based on the light measurement module.

10. A display device for a liquid crystal panel, characterized in that, include: The second acquisition unit is used to acquire the target theoretical grayscale sequence for the three-dimensional image; The second prediction unit is used to input the target theoretical grayscale sequence into the light intensity prediction model to obtain the target light intensity sequence output by the light intensity prediction model. The first control unit is configured to control the liquid crystal panel to display the three-dimensional image based on the adjusted theoretical grayscale sequence when the target theoretical grayscale sequence is adjusted based on the target light intensity sequence. The light intensity prediction model is trained based on the method described in any one of claims 1-7.

Images