Display screen backlight brightness adjustment method and system based on Mini QLED technology

Abstract

The present invention relates to a method and system for adjusting the backlight brightness of a display screen based on Mini QLED technology. The method: divides a display image signal into a brightness signal, a chromaticity signal and a saturation signal, and generates image feature data; performs sub-pixel boundary detection to obtain multiple initial backlight sub-regions, and calculates regional brightness distribution data; performs regional clustering operations and regional merging to obtain multiple target backlight sub-regions and calculates backlight partition mapping data; performs nonlinear feature transformation, and calculates the backlight unit control data of the Mini LED backlight unit; uses a deep Q network to output the backlight compensation amount of each of the target backlight sub-regions; performs real-time feedback correction, calculates the deviation value between the actual display parameters and the target parameters of each target backlight sub-region, and outputs a backlight optimization control signal. The implementation of the present invention realizes the precise adjustment of the backlight brightness and improves the consistency of the display effect.

Description

Technical Field

[0001] The present invention relates to the technical field of display screens, and in particular to a method and system for adjusting the backlight brightness of a display screen based on Mini QLED technology. Background Art

[0002] Mini QLED technology has attracted much attention due to its high dynamic range, excellent color performance and local dimming capabilities. The traditional display backlight brightness adjustment method mainly adopts a global dimming strategy, which is difficult to meet the requirements of local contrast and detail performance of different picture contents, resulting in unsatisfactory display effects.

[0003] There are three main problems with the current Mini QLED backlight adjustment technology: the backlight zoning method lacks precision, making it difficult to accurately capture the brightness change characteristics of the screen content; secondly, the driving control scheme of the backlight unit is relatively simple, failing to fully utilize the fine dimming advantages of Mini LED; the backlight compensation strategy lacks intelligence and adaptability, and the display effect is not stable enough in dynamic scenes. In the prior art, the backlight adjustment method based on Mini QLED often adopts a fixed zoning strategy and preset control parameters, which cannot be dynamically optimized for different display contents, and also lacks comprehensive consideration of physical parameters such as backlight unit temperature, current, and light intensity. Summary of the invention

[0004] The main purpose of the present invention is to provide a display screen backlight brightness adjustment method and system based on Mini QLED technology. The present invention realizes precise adjustment of the backlight brightness and improves the consistency of the display effect.

[0005] To achieve the above object, the present invention provides a display backlight brightness adjustment method based on Mini QLED technology, comprising the following steps:

[0006] The display image signal is divided into a brightness signal, a chrominance signal and a saturation signal, and image feature data is generated;

[0007] Performing sub-pixel boundary detection on the image feature data to obtain a plurality of initial backlight sub-regions, and calculating regional brightness distribution data of each of the initial backlight sub-regions;

[0008] Performing regional clustering operation and regional merging on the regional brightness distribution data to obtain a plurality of target backlight sub-regions, and calculating backlight partition mapping data of the target backlight sub-regions;

[0009] Performing nonlinear feature transformation on the backlight partition mapping data, establishing a sub-pixel level backlight mapping model, and calculating backlight unit control data of each Mini LED backlight unit in the target backlight sub-area;

[0010] Inputting the backlight unit control data into a distributed control network, constructing a state space matrix and an action space matrix for each target backlight sub-region, and using a deep Q network to output a backlight compensation amount for each target backlight sub-region;

[0011] The backlight compensation amount is corrected by real-time feedback, the deviation value between the actual display parameter and the target parameter of each target backlight sub-area is calculated, a proportional-integral-differential operation is performed, and a backlight optimization control signal is output.

[0012] The present invention also provides a display screen backlight brightness adjustment system based on Mini QLED technology, comprising:

[0013] A generating module, used for dividing the display image signal into a brightness signal, a chrominance signal and a saturation signal, and generating image feature data;

[0014] A detection module, configured to perform sub-pixel boundary detection on the image feature data to obtain a plurality of initial backlight sub-regions, and calculate regional brightness distribution data of each of the initial backlight sub-regions;

[0015] A merging module, used for performing regional clustering operation and regional merging on the regional brightness distribution data to obtain a plurality of target backlight sub-regions, and calculating backlight partition mapping data of the target backlight sub-regions;

[0016] A transformation module, used to perform nonlinear feature transformation on the backlight partition mapping data, establish a sub-pixel level backlight mapping model, and calculate the backlight unit control data of each Mini LED backlight unit in the target backlight sub-area;

[0017] A construction module, used for inputting the backlight unit control data into a distributed control network, constructing a state space matrix and an action space matrix for each target backlight sub-region, and outputting a backlight compensation amount for each target backlight sub-region using a deep Q network;

[0018] The output module is used to perform real-time feedback correction on the backlight compensation amount, calculate the deviation value between the actual display parameter and the target parameter of each target backlight sub-area, perform proportional-integral-differential operation, and output a backlight optimization control signal.

[0019] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the above methods when executing the computer program.

[0020] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above-mentioned methods are implemented.

[0021] In summary, the technical solution provided by the present invention makes the backlight zoning more accurate and improves the display effect of the picture details by introducing asymmetric Gaussian distribution function and sub-pixel boundary detection technology; adopts regional clustering operation and regional merging strategy to make the boundary transition of the backlight partition more natural and effectively avoid the sudden brightness change of the partition boundary; by establishing a sub-pixel backlight mapping model, the fine control of the Mini LED backlight unit is realized, and the local contrast and picture layering are improved; the introduction of deep Q network and distributed control strategy makes the backlight compensation have self-learning ability and enhances the adaptability of the system to different display contents; the multi-level threshold trigger and real-time feedback correction mechanism are adopted to make the backlight adjustment more timely and accurate and ensure the display stability of dynamic pictures; through the optimization design of PID control algorithm, the precise adjustment of backlight brightness is realized and the consistency of display effect is improved; based on temperature compensation and phase modulation technology, the power consumption of the backlight unit is effectively reduced and the reliability of the system is improved; through parallel sampling and distributed processing architecture, the response speed of the system is improved to meet the needs of real-time display. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is a schematic diagram of the steps of a method for adjusting the backlight brightness of a display screen based on Mini QLED technology in one embodiment of the present invention;

[0023] Figure 2 It is a structural block diagram of a display screen backlight brightness adjustment system based on Mini QLED technology in one embodiment of the present invention;

[0024] Figure 3 It is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0025] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0027] Reference Figure 1 , this embodiment provides a display backlight brightness adjustment method based on Mini QLED technology, comprising the following steps:

[0028] S1, dividing the display image signal into a brightness signal, a chrominance signal and a saturation signal, and generating image feature data;

[0029] The input display image signal is represented as RGB three-channel data, and then the RGB three-channel data is subjected to matrix operation to convert the RGB three-channel data into YUV color space. In the YUV color space, the Y component represents the brightness signal, while the U and V components represent the chromaticity and saturation information respectively. The conversion of the color space is completed through standardized matrix transformation, and the color information is converted from the RGB representation to the YUV representation using linear operations, thereby separating the brightness, chromaticity and saturation signals. The brightness signal is subjected to Gaussian filtering to reduce noise and interference in the image. Gaussian filtering is an image smoothing processing method. By applying a Gaussian template with a kernel function size of 5×5, random noise in the image is effectively removed, while retaining important structural information of the image, and the filtered brightness data is obtained. An asymmetric Gaussian distribution function operation is performed on the filtered brightness data to obtain the characteristic data of the brightness distribution. Through the operation, the local and global change features in the brightness signal are extracted to form descriptive data. At the same time, the chromaticity signal and the saturation signal are standardized to ensure the consistency of the magnitude between different signals and the effectiveness of the calculation. For the chrominance signal, a normalization processing method is used to adjust its value range to the interval of [-1,1] to unify the value range of the chrominance signal. For the saturation signal, its value range is normalized to the interval of [0,1] to obtain a standardized saturation parameter. The distribution feature data of the brightness signal, the standardized chrominance parameter, and the standardized saturation parameter are combined to form an initial feature vector. Sub-pixel interpolation operations are performed on the initial feature vector to sample the image features more finely in space and obtain image feature data. Through interpolation, the ability to describe details in the backlight brightness adjustment process is effectively improved, so that the final backlight adjustment can transition more smoothly, avoiding the phenomenon of halo or abrupt brightness changes between partitions caused by pixel-level discontinuity.

[0030] S2, performing sub-pixel boundary detection on the image feature data to obtain multiple initial backlight sub-regions, and calculating regional brightness distribution data of each initial backlight sub-region;

[0031] Specifically, the image feature data is subjected to sub-pixel gradient calculation to obtain sub-pixel sampling data, so that the edge detection of brightness changes can have higher accuracy and resolution. The sub-pixel gradient calculation generates dense and fine sampling data by accurately calculating the brightness change rate of the pixel point in space. In order to improve the accuracy of boundary detection, the sub-pixel sampling data is subjected to adaptive threshold processing, so that in different image regions, the threshold of boundary judgment is flexibly adjusted according to its local brightness characteristics, and the edge candidate points in the image are obtained, which represent the areas where the brightness in the image changes significantly. Based on the edge candidate points, the region is segmented, and the display area is divided into N×M initial backlight sub-regions according to the brightness uniformity, where N and M represent the number of divisions of the display area in the horizontal and vertical directions, respectively. Through the division, the display area is gridded to obtain the initial partition data. The brightness mean calculation needs to be performed for each initial backlight sub-region. The brightness values ​​of all pixels in each initial backlight sub-region are weighted averaged to obtain the backlight brightness value of the region. The weighted average process is achieved by assigning different weights to pixels at different positions. For example, pixels in the central area have higher weights to ensure that the brightness of the area has a greater impact on the backlight brightness adjustment. This processing method enables the backlight brightness adjustment to more concentratedly reflect the overall brightness change characteristics of the image. In the brightness mean calculation, a standard deviation analysis is performed on each initial backlight sub-area to evaluate the uniformity and stability of the brightness in the area. The results of the standard deviation analysis can reveal the degree of deviation between the brightness of the pixels in the area and the regional mean, and the deviation is normalized to obtain the brightness fluctuation coefficient. The brightness fluctuation coefficient reflects the brightness change amplitude in each initial backlight sub-area. The brightness fluctuation coefficient is threshold-classified to determine the feature level of each area. Through threshold classification, each initial backlight sub-area is divided into different brightness feature levels to better represent the brightness characteristics of different areas. Based on the feature level, a regional brightness correlation matrix is ​​constructed to describe the brightness relationship between each area. The construction of the regional brightness correlation matrix is ​​based on the feature level differences of adjacent areas, and these differences are used as indicators of correlation measurement to describe the brightness correlation between each initial backlight sub-area. The correlation data can indicate the brightness similarity between adjacent areas. The spatial distribution analysis of the regional correlation data is performed to calculate the local connectivity and global distribution characteristics of each initial backlight sub-region. The local connectivity reflects the brightness relationship between each initial backlight sub-region and its surrounding areas, while the global distribution characteristics describe the overall distribution of brightness in the entire display area. Through the spatial distribution analysis, the regional brightness distribution data of each initial backlight sub-region is obtained.

[0032] S3, performing regional clustering operation and regional merging on the regional brightness distribution data to obtain multiple target backlight sub-regions, and calculating backlight partition mapping data of the target backlight sub-regions;

[0033] It should be noted that the regional adjacent relationship analysis is performed on the regional brightness distribution data, and the brightness difference and brightness distribution correlation between each pair of adjacent initial backlight sub-regions are calculated to obtain the correlation index between the regions. The brightness difference reflects the absolute difference in brightness between adjacent regions, while the brightness distribution correlation measures the similarity of brightness changes between adjacent regions. Based on the regional correlation index, the regional merging conditions are judged. Adjacent regions are screened using two preset thresholds T1 and T2. Based on the first preset threshold T1, the brightness difference of adjacent regions is screened to ensure that only regions with brightness differences within the threshold range can be merged. Based on the second preset threshold T2, the brightness distribution correlation of adjacent regions is screened to select adjacent regions with similar brightness distribution characteristics. Double screening ensures that the regions to be merged are not only similar in brightness, but also consistent in brightness change patterns. Through the screening process, the region pairs to be merged, that is, the adjacent regions that meet the merging conditions, are determined. Regional growth calculation is performed on the region pairs to be merged, and the adjacent regions that meet the merging conditions are marked as connected domains. For the regions that meet the conditions, they are connected by connected domain marking, marked as the same group, and the regions with the same mark are merged into a new backlight partition to obtain multiple target backlight sub-regions. The region boundary is extracted for each target backlight sub-region. The spatial boundaries of these regions are determined by locating and marking the boundary pixels of each target backlight sub-region. The brightness gradient direction and amplitude of the boundary pixels are calculated to obtain boundary feature data. The boundary feature data describes the brightness change at the edge of the target backlight sub-region, including the direction and amplitude of the gradient, which is used to guide the brightness transition processing between adjacent backlight regions. Based on the boundary feature data, brightness gradient analysis is performed. Along the direction of the boundary gradient, the brightness transition function between adjacent target backlight sub-regions is calculated, and the width range of the gradient region is determined to ensure smooth brightness transition between adjacent target backlight sub-regions and avoid the boundary halo or abrupt visual effects between partitions caused by brightness discontinuity. Through brightness gradient analysis, the gradient control parameters are obtained, and brightness interpolation operations are performed based on the gradient control parameters. In the boundary area of ​​adjacent target backlight sub-regions, the brightness distribution of the transition area is calculated along the gradient direction to obtain gradient mapping data. The gradient mapping data is used to achieve a gradual transition of brightness between target backlight sub-areas, thereby improving the smoothness and visual comfort of the overall display effect. In order to optimize the backlight control, the gradient mapping data is energy balanced to ensure the reasonable distribution of the total energy and boundary energy of each target backlight sub-area. In this process, the loss of total energy and boundary energy of each target backlight sub-area is calculated to obtain balanced compensation data. The balanced compensation data is used to compensate for the energy imbalance caused by brightness transition to ensure that the brightness of each backlight sub-area can be evenly distributed. The balanced compensation data is merged with the brightness distribution data of each target backlight sub-area to obtain the backlight partition mapping data of each target backlight sub-area.

[0034] S4, performing nonlinear feature transformation on the backlight partition mapping data, establishing a sub-pixel level backlight mapping model, and calculating the backlight unit control data of each Mini LED backlight unit in the target backlight sub-area;

[0035] Specifically, the backlight partition mapping data is sampled at the sub-pixel level. Each backlight partition is sub-pixel sampled at 10 times the density in the horizontal and vertical directions to obtain a sub-pixel brightness matrix containing more detailed brightness information. A nonlinear mapping transformation is performed based on the sub-pixel brightness matrix. A segmented S-shaped curve is used to map the brightness value to the physical driving space of the backlight unit. The slopes of the S-shaped curve in the dark and bright areas are set to 0.6 and 1.2 respectively, thereby achieving nonlinear brightness adjustment. A lower slope in the dark area can better preserve the dark details and prevent the loss of details in the dark area due to too fast brightness increase; while a higher slope in the bright area helps to improve the overall brightness performance and make the bright area more transparent. The application of the segmented S-shaped curve can make the response of the backlight adjustment in different brightness ranges more in line with the visual characteristics of the human eye, improve the contrast and layering of the display effect, and obtain a backlight drive curve. Sub-pixel backlight modeling is performed on the backlight driving curve, and the brightness value of each sub-pixel point is mapped to the Mini LED backlight unit array according to the physical distribution law, and a one-to-many mapping relationship between sub-pixel points and backlight units is established to form a sub-pixel backlight mapping model. The establishment of the mapping model means that a Mini LED backlight unit covers the brightness information of multiple sub-pixels, so that the brightness of multiple sub-pixels is integrated in a weighted manner to achieve control of the backlight unit. Through this step, a fine mapping from the refined brightness information of the displayed image to the control of the Mini LED backlight unit is achieved. Based on the established sub-pixel backlight mapping model, the brightness weighted average calculation is performed on each sub-pixel point within the coverage range of the Mini LED backlight unit, and the weighted result is converted into an 8-bit quantized driving current value to obtain a current control sequence. The weighted average process integrates the brightness characteristics of multiple sub-pixels, so that the backlight unit can reflect the overall performance of the brightness in its coverage area. The 8-bit quantization of the current value means that the brightness adjustment is accurately divided into 256 levels, so that the backlight control has a high degree of precision. The pulse width modulation (PWM) calculation is performed on the current control sequence, and the driving current value is converted into a PWM signal with a duty cycle ranging from 10% to 90% according to the 256-level brightness classification to obtain a pulse width control sequence. The duty cycle of the PWM signal determines the brightness output ratio of the backlight unit, and the brightness of the backlight unit is accurately controlled by controlling the duty cycle. Based on the pulse width control sequence, phase modulation processing is performed to stagger the PWM signal phases of adjacent backlight units by 120 degrees to obtain phase control data. The phase stagger processing is to reduce the current shock and visual flicker caused by simultaneous switching between adjacent backlight units, and improve the smoothness and stability of the display. Temperature compensation is performed on the phase control data. Since the Mini LED backlight unit generates heat during operation, and the temperature change affects its brightness and driving characteristics, the driving current value and duty cycle are corrected according to the temperature coefficient of the backlight unit position.The temperature coefficient is dynamically adjusted within the range of 0.8 to 1.2. Through temperature compensation, it can ensure that the backlight unit can still maintain stable brightness output under different temperature conditions, and obtain the compensated control parameters. The compensated control parameters are encoded in parallel to generate a 32-bit control word containing current value, duty cycle and phase information, which is sent to the drive circuit through the serial interface to obtain the control data of the backlight unit. These control data are used to drive each Mini LED backlight unit to work according to the predetermined brightness and timing requirements, and realize dynamic adjustment of the display backlight.

[0036] S5, inputting the backlight unit control data into the distributed control network, constructing the state space matrix and action space matrix of each target backlight sub-region, and using the deep Q network to output the backlight compensation amount of each target backlight sub-region;

[0037] Among them, the backlight unit control data is sampled in parallel, and the temperature value, current value and light intensity value of each target backlight sub-area are collected to construct a three-dimensional state vector. The three-dimensional state vector is mapped into a state space matrix St through a multi-layer perceptron. The action space is constructed based on the state space matrix. In the process of constructing the action space, the compensation coefficient α is discretized into 8 levels in the range of [0.8, 1.2] to select different compensation intensities. At the same time, the adjustment step β is discretized into 4 levels in the range of [0.05, 0.2] to control the variation of the compensation. Through the discretization operation, the continuous adjustment process is converted into a discrete selection, so that the compensation calculation is performed with lower complexity when the action is selected. Then, through the tensor product operation, the compensation coefficient and the adjustment step are combined to generate the action space matrix At, which represents the set of operations that can be selected for each backlight sub-area under different states. The state space matrix St and the action space matrix At are input into the deep Q network (DQN) to calculate the relationship between state and action. In the structure of DQN, three layers of convolutional layers are used to extract features from the input data. The convolution kernel sizes are 3×3, 2×2, and 1×1, respectively, and the step size is 1. The purpose of the multi-layer convolution operation is to extract spatial features from the input data, so that the network can capture the relationship and change pattern of the backlit sub-areas under different conditions and obtain feature mapping data. The multi-level processing of convolution enables the deep Q network to better understand the complex spatial distribution characteristics and provide high-quality feature expression for subsequent action decisions. The feature mapping data is transformed using a fully connected layer, and the extracted feature vectors are mapped to the state-action value space through two hidden layers. The number of neurons in the hidden layer is 256 and 128 respectively. Through this design, the feature information extracted previously is fully utilized to calculate the value function Q(St,At) between the state and the action. The value function Q(St,At) represents the expected value that can be obtained by performing an action in a specific state. Through the calculation of the value function, the effect of each action is evaluated, providing a quantitative basis for the selection of strategies. Based on the value function Q(St,At), the policy network π(At|St) is constructed. In this process, the Actor-Critic architecture is adopted, in which the Actor network outputs the probability distribution of actions, while the Critic network is used to evaluate the value of each action. In this architecture, the Actor is responsible for exploring different actions, and the Critic is responsible for evaluating the results of the exploration to guide the next decision. This architecture combines the advantages of policy optimization and value evaluation to learn and optimize control strategies more efficiently. The policy model is trained by reinforcement learning. During the training process, the discount factor γ is set to 0.9 to balance the impact of short-term rewards and long-term rewards, ensuring that both the current backlight effect and the future cumulative effect are taken into account in the decision-making.At the same time, the experience replay buffer is used to store the state transition sequence to avoid the negative impact of data correlation on the training process and improve the generalization ability of the network. The network parameters are updated through the time difference algorithm, so that the policy network can gradually approach the optimal policy and obtain the trained policy network. The temperature value, current value and light intensity value of each target backlight sub-area are input into the trained policy network, and the probability distribution of the action is obtained by forward propagation calculation. The ε-greedy strategy is used to select the optimal action combination. In most cases, the current optimal action is selected, and random selection is occasionally performed to maintain a certain exploration ability. Through this step, the compensation control parameters that adapt to the current display environment are found. The compensation control parameters are reversely mapped, and the compensation coefficient and adjustment step size are converted into the backlight compensation amount of each Mini LED backlight unit. Through this process, the optimal control signal of each backlight unit in the current state is obtained, so as to achieve precise adjustment of the backlight.

[0038] S6, performing real-time feedback correction on the backlight compensation amount, calculating the deviation between the actual display parameter and the target parameter of each target backlight sub-area, performing proportional-integral-differential operation, and outputting a backlight optimization control signal.

[0039] Specifically, the backlight compensation amount is separated and processed, and the compensation amount of the entire backlight system is divided into multiple partition compensation parameter groups. Each partition compensation parameter group represents the backlight adjustment information of a backlight sub-area. Real-time data collection is performed based on the partition compensation parameter group, and the collected real-time parameters include the characteristics of brightness, chromaticity and uniformity of each backlight partition in the current display frame. The partition real-time parameters are compared with the target parameters. Through subtraction operation, the brightness deviation, chromaticity deviation and uniformity deviation of each partition are obtained. These deviation values ​​reflect the gap between the current actual display effect and the expected target. The deviation value is standardized to obtain standardized deviation data. The proportional coefficient is calculated based on the standardized deviation data. The brightness deviation, chromaticity deviation and uniformity deviation are multiplied by the corresponding proportional coefficient respectively to obtain the proportional control parameter. The calculation of the proportional control parameter directly reflects the degree of the current deviation, and the proportional coefficient determines the response intensity of the system to the deviation, so that the rapidity and stability of the system are controlled by adjusting the proportional coefficient. At the same time, the standardized deviation data is integrated and accumulated and summed to obtain the integral control parameter. The purpose of integral control is to gradually accumulate the correction amount when the deviation exists for a long time to eliminate the steady-state error and ensure that the system can gradually tend to the target value. The standardized deviation data is differentiated to calculate the parameter change rate between each two adjacent frames to obtain the differential control parameter. The role of differential control is to predict the change trend of the system so as to respond in advance, thereby suppressing the overshoot caused by rapid changes and improving the response smoothness and accuracy of the system. By calculating the change rate of adjacent frames, it is possible to better cope with the adjustment needs of the backlight brightness when the image content changes rapidly and reduce the visual discomfort caused by sudden changes in brightness. The proportional control parameter, the integral control parameter and the differential control parameter are linearly superimposed to obtain the PID control amount. The PID control amount combines the current deviation size (proportional control), the accumulated deviation (integral control) and the change trend of the deviation (differential control), and can comprehensively correct the error of the entire system, making the backlight adjustment more accurate and stable, especially when facing dynamic image content, it can better cope with the changes in brightness, chromaticity and uniformity. The PID control amount is parameter reconstructed, and the PID control amount is converted into the drive parameter adjustment value of the backlight unit to obtain the backlight optimization control signal. The PID control quantity is adapted to a specific physical driving signal, such as a current adjustment value or a duty cycle parameter, to drive the operation of the Mini LED backlight unit.

[0040] In one example, a display image signal is divided into a brightness signal, a chrominance signal, and a saturation signal, and image feature data is generated, including:

[0041] Performing RGB color space conversion on the display image signal to obtain RGB three-channel data, and performing matrix operation on the RGB three-channel data to convert the RGB three-channel data into a YUV color space to obtain a brightness signal, a chrominance signal, and a saturation signal;

[0042] The brightness signal is subjected to Gaussian filtering, and a Gaussian template with a kernel function size of 5×5 is used to remove image noise to obtain filtered brightness data, and an asymmetric Gaussian distribution function operation is performed on the filtered brightness data to obtain brightness distribution feature data;

[0043] Normalizing the chrominance signal, adjusting the value range of the chrominance signal to the interval [-1, 1], and obtaining a standardized chrominance parameter; and normalizing the saturation signal, adjusting the value range of the saturation signal to the interval [0, 1], and obtaining a standardized saturation parameter;

[0044] The brightness distribution feature data, the standardized chromaticity parameters and the standardized saturation parameters are combined into an initial feature vector, and a sub-pixel interpolation operation is performed on the initial feature vector to obtain image feature data.

[0045] In this example, the display image signal is converted from the RGB color space to the YUV color space. The RGB color space is the most common image representation method, in which the color information of each pixel is represented by the values ​​of the three channels of red (R), green (G), and blue (B). The YUV color space is a color representation method that is more in line with the visual perception characteristics of the human eye, in which Y represents the brightness signal, and U and V represent the chromaticity information, that is, the hue and saturation of the color. Matrix operations are performed on the RGB three-channel data to convert the RGB representation into the YUV representation to obtain the brightness signal Y and the chromaticity signals U and V. The RGB to YUV conversion process is implemented through the following linear matrix operations:

[0046] ;

[0047] In this matrix formula, represents the brightness signal, and Represent the chrominance signal respectively. The elements of the matrix are the linear transformation coefficients between RGB and YUV. For the input RGB value, , and Represent the values ​​of the red, green, and blue channels respectively. Through matrix multiplication, we get , and The value of , thus completing the color space conversion. Perform Gaussian filtering. Gaussian filtering is an image smoothing method that removes random noise in the image and makes the brightness signal smoother and more uniform. In the Gaussian filtering process, a kernel function size of 5 is used. 5, which means that the value of each pixel depends on its own and its surrounding The pixel values ​​in the area are weighted averaged, where the weights are determined by the Gaussian function. The expression of the Gaussian function is: ;

[0048] In this formula, represents the value of the Gaussian kernel, and are the horizontal and vertical distances of the pixel relative to the center pixel, is the standard deviation of the Gaussian distribution and determines the strength of the filter. The Gaussian template for the brightness signal Filtering is performed to remove random noise and maintain the edge details of the image to obtain filtered brightness data. An asymmetric Gaussian distribution function is performed on the filtered brightness data to obtain brightness distribution feature data. The use of an asymmetric Gaussian distribution function is to better capture the local brightness change characteristics in the image and enhance the details in the brightness signal, so that the backlight system can more finely control the brightness changes in different areas when adjusting the brightness. For example, the asymmetric Gaussian distribution uses different weighting strategies for areas with high and low brightness, thereby highlighting the brightness characteristics of certain specific areas. At the same time, the chrominance signal and saturation signal , respectively, to ensure that the value ranges of different signals are consistent. Normalization is performed to adjust the value range to the [-1,1] interval to obtain standardized chromaticity parameters. The normalization formula is expressed as: ;

[0049] in, and are the minimum and maximum values ​​of the chrominance signal, respectively. The value range of is mapped to [-1,1]. Similarly, for the saturation signal Normalization is performed to adjust its value range to the [0,1] interval to obtain a standardized saturation parameter. The normalization formula is: ;

[0050] Normalization can eliminate the magnitude differences between different physical quantities, making the subsequent calculation process more stable and ensuring that all input parameters are within the same range for effective feature fusion. The filtered brightness distribution feature data, standardized chromaticity parameters, and standardized saturation parameters are combined to form an initial feature vector. Sub-pixel interpolation operations are performed on the initial feature vector to obtain image feature data. The purpose of sub-pixel interpolation is to increase the resolution of the image in space so that the feature description is more refined. By improving the resolution of image features, backlight control is more accurate and abrupt boundaries of brightness changes in space are avoided. Sub-pixel interpolation operations are implemented through methods such as bilinear interpolation or cubic spline interpolation, and the value of the new pixel is estimated by the values ​​of adjacent pixels. For example, for bilinear interpolation, it is expressed as: ;

[0051] in, represents the interpolated pixel value, , are the coordinates of adjacent known pixels, and Respectively represent the relative position of the target point in the horizontal and vertical directions. Through this interpolation method, the spatial resolution of the image is effectively improved, and more refined image feature data is obtained.

[0052] In one example, sub-pixel boundary detection is performed on the image feature data to obtain multiple initial backlight sub-regions, and regional brightness distribution data of each initial backlight sub-region is calculated, including:

[0053] Perform sub-pixel gradient calculation on image feature data to obtain sub-pixel sampling data, and perform adaptive threshold processing on the sub-pixel sampling data to obtain edge candidate points;

[0054] Perform region segmentation based on the edge candidate points, divide the display area into N×M initial backlight sub-areas according to the brightness uniformity, where N and M are the number of divisions of the display area in the horizontal and vertical directions, respectively, to obtain initial partition data;

[0055] Calculate the brightness mean of each initial backlight sub-region in the initial partition data, perform weighted average of the brightness values ​​of all pixels in each initial backlight sub-region to obtain the brightness value of the backlight sub-region, perform standard deviation analysis based on the brightness value of the backlight sub-region, normalize the deviation of the pixels in each initial backlight sub-region from the regional mean, and obtain the brightness fluctuation coefficient;

[0056] The brightness fluctuation coefficient is threshold-classified to obtain the regional feature level, and the regional brightness correlation matrix is ​​constructed based on the regional feature level. The feature level difference of adjacent regions is used as the correlation measurement index to obtain the regional correlation data;

[0057] The spatial distribution analysis of the regional correlation data is performed, the local connectivity and global distribution characteristics of each initial backlight sub-region are calculated, and the regional brightness distribution data of each initial backlight sub-region is obtained.

[0058] In this example, sub-pixel gradient calculation is performed on the image feature data to obtain more detailed edge information and ensure accurate control of the brightness of the display area. Sub-pixel brightness change information is obtained by calculating the slight difference in brightness changes between pixels. Assume that the image feature data is a matrix , then use the Sobel operator to calculate the gradient of the image. The Sobel operator is an edge detection tool that uses convolution operations to calculate the gradient of the image in the horizontal and vertical directions, defined as: ;

[0059] in, and Respectively represent the gradient calculation results of the image in the horizontal and vertical directions, the symbol * represents the convolution operation, is the input image feature data matrix. By calculating and , and get the magnitude of the gradient : ;

[0060] Gradient Amplitude Indicates the intensity of brightness change of each pixel, reflecting the degree of change in image brightness. When performing sub-pixel gradient calculation, the original gradient value is interpolated to obtain the gradient information of the sub-pixel position, so that the brightness change of the image can be captured at a finer spatial resolution. Adaptive threshold processing is performed on the sub-pixel sampling data to determine the location of the edge candidate point. Different thresholds are adaptively set according to local image features to enhance the robustness of edge detection. Threshold The calculation of is based on the local mean and standard deviation of the image: ;

[0061] in, is the mean value in the local window of the image, is the standard deviation of the local window, is a constant used to control the sensitivity of the threshold. In an adaptive way, edge candidate points whose gradient amplitude exceeds the threshold are obtained. These edge candidate points identify the areas with the most significant brightness changes in the image. Based on the edge candidate points, the display area is segmented and the entire display area is divided into initial backlight sub-areas, where and are the number of divisions of the display area in the horizontal and vertical directions respectively. Using the region growing algorithm based on edge information, starting from the edge candidate points, the adjacent pixels are gradually merged based on these candidate points until the brightness uniformity condition is met, thereby completing the region segmentation. The initial backlight sub-region after segmentation can effectively reflect the brightness characteristics of different parts of the image and provide independent backlight control for each region. After completing the region segmentation, the brightness mean value of each backlight sub-region in the initial partition data is calculated. For each initial backlight sub-region, the brightness mean value Obtained by taking a weighted average of the brightness values ​​of all pixels in the area: ;

[0062] in, Indicates The average brightness of the backlight sub-areas, is the number of pixels in the area, For the The weight of each pixel is determined by the distance between the pixel and the center of the region. For the The brightness value of each pixel. A standard deviation analysis is performed on each initial backlight sub-region. Standard deviation It is used to measure the fluctuation degree of pixel brightness relative to the mean value in the area. The calculation formula is: ;

[0063] The larger the standard deviation, the more drastic the fluctuation of brightness in the area; the smaller the standard deviation, the more uniform the brightness in the area. Normalize the standard deviation of each area to get the brightness fluctuation coefficient : ;

[0064] Normalized brightness fluctuation coefficient Better reflect the brightness fluctuation of each area and eliminate the impact caused by the difference in absolute brightness. Threshold classification is performed on the brightness fluctuation coefficient to determine the characteristic level of each backlight sub-area. Divide into different levels, such as low fluctuation, medium fluctuation and high fluctuation, to obtain the characteristic level of the region. Based on the characteristic level, construct the regional brightness correlation matrix. Regional brightness correlation matrix is a matrix that describes the brightness relationship between adjacent regions. The elements of the matrix It is expressed by the difference in feature levels of adjacent regions: ;

[0065] in, and Respectively and The feature level of adjacent regions. The smaller the feature level difference, the smaller the brightness change between adjacent regions and the higher the correlation; the larger the feature level difference, the greater the brightness change between adjacent regions. The spatial distribution analysis of the regional correlation data is performed to calculate the local connectivity and global distribution characteristics of each initial backlight sub-region. Local connectivity It is an indicator to measure the brightness correlation between a backlight sub-area and its adjacent areas. It is obtained by calculating the sum of the correlation between the area and all its adjacent areas: ;

[0066] in, Indicates The higher the local connectivity, the closer the brightness relationship between the region and its adjacent regions, and the more associated regions need to be considered when adjusting the backlight. Global distribution characteristics It describes the brightness distribution of all areas in the entire display screen and is obtained by calculating the mean or variance of all local connectivity.

[0067] In one example, performing regional clustering operation and regional merging on regional brightness distribution data to obtain multiple target backlight sub-regions, and calculating backlight partition mapping data of the target backlight sub-regions, includes:

[0068] Performing regional adjacent relationship analysis on the regional brightness distribution data, calculating the brightness difference and brightness distribution correlation between each pair of adjacent initial backlight sub-regions, and obtaining a regional correlation index;

[0069] Based on the regional correlation index, the regional merging condition is judged, the brightness difference of adjacent regions is screened according to the first preset threshold value T1, and the brightness distribution correlation of adjacent regions is screened according to the second preset threshold value T2 to obtain the region pair to be merged;

[0070] Perform region growing calculation on the region to be merged, mark the connected domains of adjacent regions that meet the merging conditions, and merge the regions with the same mark into a new backlight partition to obtain multiple target backlight sub-regions;

[0071] Extracting the region boundary based on the target backlight sub-region, positioning and marking the boundary pixels of each target backlight sub-region, and calculating the brightness gradient direction and amplitude of the boundary pixels to obtain boundary feature data;

[0072] Performing brightness gradient analysis on the boundary feature data, calculating the brightness transition function between adjacent target backlight sub-regions along the boundary gradient direction, and determining the width range of the gradient region to obtain the gradient control parameter, and performing brightness interpolation operation based on the gradient control parameter, calculating the brightness distribution of the transition region in the boundary region of adjacent target backlight sub-regions, and obtaining the gradient mapping data;

[0073] The gradient mapping data is energy balanced, the total energy and boundary energy loss of each target backlight sub-area are calculated, and the balanced compensation data is obtained. The balanced compensation data is then fused with the brightness distribution data of each target backlight sub-area to obtain the backlight partition mapping data of each target backlight sub-area.

[0074] In this example, the regional adjacent relationship analysis is performed on the regional brightness distribution data, and the brightness difference and brightness distribution correlation between each pair of adjacent initial backlight sub-regions are calculated to obtain the regional correlation index. Assume that the two adjacent initial backlight sub-regions are region A and region B, and their brightness averages are and Brightness difference Calculated by the following formula: ;

[0075] in, Indicates the brightness difference between area A and area B, which can reflect the degree of brightness change between adjacent areas. , measures the similarity of brightness distribution between two adjacent regions, and calculates the Pearson correlation coefficient of pixel brightness in the two regions: ;

[0076] in, and Represent the brightness values ​​of pixels in area A and area B respectively. and is the average brightness of the two regions. The value range is between [-1,1], where a value close to 1 indicates that the brightness distribution of the two regions is similar, and a value close to -1 indicates that the brightness distribution is inversely correlated. The region merging condition is determined based on the region correlation index. Adjacent regions are screened based on two preset thresholds. The first preset threshold is used. Determine the brightness difference Whether the merging conditions are met, that is, only when When the brightness of adjacent regions is close enough, they can be merged. Similarly, the second preset threshold is used Screening brightness distribution correlation , only when Only when , can it be considered that the brightness distribution similarity of the two regions is high and suitable for merging. Through double screening, the region pair to be merged is obtained. Perform region growing calculation on the region pair to be merged, and mark the connected domains of adjacent regions that meet the merging conditions. Given a seed region, if the region meets the merging conditions, it and the adjacent regions that meet the conditions are marked as the same connected domain until further merging is impossible. Through this process, multiple regions with the same mark are merged into a new backlight partition to obtain multiple target backlight sub-regions. The brightness of each target backlight sub-region is more uniform, which is conducive to improving the stability and consistency of the display effect. The boundary of the target backlight sub-region is extracted. The boundary pixels of each target backlight sub-region are located and marked to determine the physical boundaries of each backlight sub-region. At the same time, the brightness gradient direction and amplitude of the boundary pixels are calculated. The gradient direction is determined by calculating the brightness change of each boundary pixel in the horizontal and vertical directions, and is expressed by the following formula: ;

[0077] in, represents the gradient direction, and are the brightness gradients of the boundary pixels in the horizontal and vertical directions, respectively, calculated using the Sobel operator. It represents the intensity of brightness change, and the calculation formula is: ;

[0078] The calculation of gradient direction and amplitude helps to understand the brightness change trend between adjacent areas, thereby providing necessary data support for subsequent brightness gradient analysis. Brightness gradient analysis is performed based on boundary feature data. Along the direction of the boundary gradient, the brightness transition function between adjacent target backlight sub-areas is calculated to ensure smooth brightness changes between adjacent areas and avoid abrupt brightness boundaries. The brightness transition function is defined as a linear or nonlinear interpolation function used to smoothly transition brightness between adjacent areas. Assume that the brightness transition function between adjacent areas A and B is , then it is calculated by the following linear interpolation formula:

[0079] ;

[0080] in, and are the brightness values ​​of regions A and B respectively, is the distance along the gradient direction, is the width range of the transition area. By determining the width range of the gradient area, the gradient control parameters are obtained, and the brightness is interpolated based on these parameters, so as to calculate the brightness distribution of the transition area in the boundary area of ​​the adjacent target backlight sub-area, and obtain the gradient mapping data. The gradient mapping data can ensure that the brightness transition between different backlight sub-areas is more natural, and improve the uniformity of the display effect and visual comfort. The gradient mapping data is energy balanced. By calculating the total energy of each target backlight sub-area and the boundary energy loss, the energy distribution of each area during the brightness transition process is evaluated. Assume that the total energy of a backlight sub-area is , and its calculation formula is: ;

[0081] in, represents the total energy of the backlight sub-area, For the region The brightness value of a pixel, is the area of ​​a pixel, is the total number of pixels in the region. Similarly, the boundary energy loss By calculating the energy of the boundary pixels: ;

[0082] By calculating the energy loss, the balanced compensation data is obtained to compensate for the energy imbalance that occurs during the brightness transition process. The balanced compensation data is merged with the brightness distribution data of each target backlight sub-area to obtain the backlight partition mapping data of each target backlight sub-area. The backlight partition mapping data is used to control the backlight unit of the display screen so that it outputs appropriate brightness in different areas, thereby realizing dynamic backlight adjustment of the displayed content.

[0083] In one example, nonlinear feature transformation is performed on the backlight partition mapping data, a sub-pixel level backlight mapping model is established, and backlight unit control data of each Mini LED backlight unit in the target backlight sub-area is calculated, including:

[0084] Sub-pixel sampling is performed on the backlight partition mapping data, and each backlight partition is sampled into sub-pixel points with 10 times the density in the horizontal and vertical directions to obtain a sub-pixel brightness matrix;

[0085] Based on the sub-pixel brightness matrix, a nonlinear mapping transformation is performed, and a piecewise S-shaped curve is used to map the brightness value to the physical driving space of the backlight unit, where the slopes of the S-shaped curve in the dark and bright areas are 0.6 and 1.2 respectively, to obtain the backlight driving curve;

[0086] Perform sub-pixel backlight modeling on the backlight driving curve, map the brightness value of the sub-pixel point to the Mini LED backlight unit array according to the physical distribution law, establish a one-to-many mapping relationship from the sub-pixel point to the backlight unit, and obtain a sub-pixel backlight mapping model;

[0087] The driving current is calculated based on the sub-pixel backlight mapping model, and the brightness of the sub-pixels within the coverage area of ​​each Mini LED backlight unit is weighted averaged. The weighted result is converted into an 8-bit quantized driving current value to obtain the current control sequence.

[0088] Perform pulse width modulation calculation on the current control sequence, convert the driving current value into a PWM signal with a duty cycle ranging from 10% to 90% according to 256 levels of brightness grading, obtain a pulse width control sequence, and perform phase modulation based on the pulse width control sequence, stagger the phases of the PWM signals of adjacent backlight units by 120 degrees, and obtain phase control data;

[0089] Perform temperature compensation on the phase control data, and correct the driving current value and duty cycle according to the temperature coefficient of the backlight unit position. The temperature coefficient is dynamically adjusted within the range of 0.8 to 1.2 to obtain the compensated control parameters.

[0090] The compensated control parameters are encoded in parallel to generate a 32-bit control word containing current value, duty cycle and phase information, which is sent to the drive circuit through a serial interface to obtain backlight unit control data.

[0091] In this example, the backlight partition mapping data is sampled at the sub-pixel level. Each backlight partition is sub-pixel sampled at 10 times the density in the horizontal and vertical directions to obtain higher-precision brightness information. Assume that the brightness matrix of the original backlight partition is , after sub-pixel sampling, the new brightness matrix is ​​expressed as ,in and They are one tenth of the original coordinates respectively. Through sub-pixel sampling, a high-resolution brightness matrix, namely the sub-pixel brightness matrix, is obtained, and a nonlinear mapping transformation is performed on the sub-pixel brightness matrix. The brightness value is mapped to the physical driving space of the Mini LED backlight unit, and a segmented S-shaped curve is used to achieve this mapping. The S-shaped curve is a brightness adjustment curve used to enhance the contrast and dynamic range of the image. The formula of the S-shaped curve is expressed as: ;

[0092] In this formula, is the brightness value after nonlinear mapping, is the brightness obtained by sub-pixel sampling, and are the slopes of the dark and bright areas, respectively, where , , used to adjust the mapping speed of dark and bright areas; It is the segmentation point of the S-shaped curve, indicating the transition area in the middle of brightness; is a constant term, ensuring that the curve is The segmented S-shaped curve has a smaller slope in the dark area, which better preserves the details of the dark area, and a larger slope in the bright area, which enhances the brightness of the bright area, thereby improving the contrast of the overall picture. Through nonlinear mapping transformation, the backlight driving curve is obtained to control the brightness output of the Min LED backlight unit. The backlight driving curve is modeled at the sub-pixel level, and the brightness value of the sub-pixel point is mapped to the Mini LED backlight unit array according to the physical distribution law. A one-to-many mapping relationship between sub-pixel points and backlight units is established. Since the area covered by each Mini LED backlight unit contains multiple sub-pixel points, the brightness of each backlight unit is calculated by weighted average. Assume that a Mini LED backlight unit covers sub-pixel points, and the brightness values ​​of the sub-pixel points are , the corresponding weight is , then the brightness value of the backlight unit Calculated by the following formula: ;

[0093] in, It is determined by the distance between the sub-pixel point and the center of the backlight unit. The closer the distance, the greater the weight. In this way, the brightness value of each Mini LED backlight unit is obtained to form a sub-pixel backlight mapping model. Based on the sub-pixel backlight mapping model, the driving current is calculated. For each Mini LED backlight unit, the brightness values ​​of the sub-pixels within the coverage range are used for brightness weighted averaging, and then the weighted result is converted into an 8-bit quantized driving current value. The quantization process is achieved through a linear quantization formula: ;

[0094] in, is the quantized driving current value, is the average brightness of the backlight unit, is the maximum brightness value defined by the system, and 255 represents the maximum level of 8-bit quantization. Through linear quantization, continuous brightness values ​​are mapped to discrete current control values. Pulse width modulation (PWM) calculation is performed on the current control sequence. PWM is a technology that controls current by adjusting the duty cycle of the signal, which is used to accurately adjust the brightness of the LED. The drive current value is converted into a PWM signal according to the 256-level brightness grading, and the duty cycle range is set between 10% and 90% to avoid flickering caused by too low a duty cycle, or too high a duty cycle causing overbrightness. The duty cycle of the PWM signal is calculated by the following formula: ;

[0095] in, represents the duty cycle, is the quantized driving current value. Through calculation, the pulse width control sequence used to drive the backlight unit is obtained. In order to reduce the impact of the current mutation of adjacent backlight units on the overall system, phase modulation is performed based on the pulse width control sequence. The PWM signal phases of adjacent backlight units are staggered by 120 degrees to reduce the current peak caused by the simultaneous switching of adjacent units, improve the stability of the system, and obtain phase control data. The phase control data is temperature compensated. Since the Mini LED backlight unit generates heat when working, the temperature change will affect the brightness and current characteristics of the LED. The driving current value and duty cycle are corrected according to the temperature coefficient of the backlight unit position. Temperature coefficient Dynamically adjusted within the range of 0.8 to 1.2 to compensate for brightness drift caused by temperature changes. Corrected drive current value and duty cycle Calculated by the following formula: ;

[0096] Temperature compensation ensures that the backlight unit can still work stably under different temperature conditions and maintain brightness consistency. The compensated control parameters are encoded in parallel to generate a 32-bit control word containing current value, duty cycle and phase information, which is sent to the drive circuit through the serial interface. The format of the control word is designed as follows: the first 8 bits are used to store the current value The next 8 bits are used to store the duty cycle , the next 8 bits are used to store phase information, and the remaining 8 bits are used to store other control flags or check information. Through encoding, all control information is effectively encapsulated in a control word and transmitted to the backlight drive circuit through a serial interface to obtain the control data of the backlight unit.

[0097] In one example, the backlight unit control data is input into the distributed control network, the state space matrix and the action space matrix of each target backlight sub-area are constructed, and the backlight compensation amount of each target backlight sub-area is outputted using a deep Q network, including:

[0098] The backlight unit control data is sampled in parallel, the temperature value, current value and light intensity value of each target backlight sub-area are collected, a three-dimensional state vector is constructed, and the state space matrix St is obtained through multi-layer perceptron mapping;

[0099] The action space is constructed based on the state space matrix. The compensation coefficient α is discretized into 8 levels in the range of [0.8, 1.2], and the adjustment step length β is discretized into 4 levels in the range of [0.05, 0.2]. The action space matrix At is obtained through tensor product operation.

[0100] The state space matrix St and the action space matrix At are input into the deep Q network, and features are extracted through three convolutional layers. The convolution kernel sizes are 3×3, 2×2, and 1×1, respectively, and the step size is 1 to obtain feature mapping data;

[0101] Perform a fully connected layer transformation on the feature map data, map the feature vector to the state-action value space through two hidden layers, with the number of neurons in the hidden layers being 256 and 128 respectively, and obtain the value function Q(St,At);

[0102] Based on the value function Q(St,At), a policy network π(At|St) is constructed using an Actor-Critic architecture, where the Actor network outputs the action probability distribution and the Critic network evaluates the action value to obtain a policy model.

[0103] The policy model is trained through reinforcement learning, the discount factor γ is set to 0.9, the state transition sequence is stored in the experience replay buffer, and the network parameters are updated through the temporal difference algorithm to obtain the trained policy network.

[0104] The temperature value, current value and light intensity value of each target backlight sub-area are input into the trained strategy network, and the action probability distribution is obtained by forward propagation calculation. The ε-greedy strategy is used to select the optimal action combination to obtain the compensation control parameters. The compensation control parameters are reverse mapped to convert the compensation coefficient and adjustment step into the backlight compensation amount of each Mini LED backlight unit.

[0105] In this example, data is collected for each target backlight sub-area. The temperature, current, and light intensity values ​​of each target backlight sub-area are collected and integrated into a three-dimensional state vector. Assume that the temperature of the backlight sub-area is , the current value is , the light intensity is , then the three-dimensional state vector is expressed as By collecting these characteristic information of each target backlight sub-area, the state space matrix is ​​constructed , each row or column of the matrix represents the current state of a target backlight sub-region. Based on the state space matrix Construct an action space to optimize different control strategies in reinforcement learning. In the construction of the action space, the compensation coefficient Discretized into 8 levels, ranging from [0.8, 1.2]. Compensation coefficient It is a parameter used to adjust the backlight output and indicates the intensity of backlight compensation. The discretized compensation coefficient is expressed as: ;

[0106] Similarly, adjust the step size Discretize into 4 levels in the range [0.05, 0.2] and adjust the step size Represents the step size of backlight adjustment, which is used to control the speed of backlight change. The discretized adjustment step size is expressed as: ;

[0107] By discretizing the compensation coefficient and the adjustment step size, we get the action set. Then, we combine the compensation coefficient and the adjustment step size through tensor product operation to get the action space matrix: , represents all possible combinations of operations available in each state. and the action space matrix As input, it is input into the deep Q network (DQN) for feature extraction. The deep Q network uses three convolutional layers to extract features from the input data, and the convolution kernel sizes are , 2 and , the step size is set to 1. The purpose of the convolution operation is to extract the spatial features in the input data, especially the complex nonlinear relationship between state and action. Assume that the input state-action matrix is , after the After the layer convolution operation, the feature map data is represented as: ;

[0108] in, For the The input feature map of the layer, For the The convolution kernel of the layer, the symbol * represents the convolution operation, is the bias term, is the activation function (e.g. ReLU function). Through the processing of three convolutional layers, feature mapping data is obtained, which contains complex relationship information between states and actions. The feature mapping data extracted by the convolutional layer is transformed into a fully connected layer, and the feature vector is mapped to the state-action value space through two hidden layers. Assuming that the number of neurons in the hidden layer is 256 and 128 respectively, the output of the fully connected layer is calculated by the following formula: ;

[0109] in, and are the weight matrices of the first and second hidden layers, respectively. and is the bias term, is the activation function. The final output value function Indicates that in a specific state Take an action The expected value that can be obtained is used to evaluate the quality of the action. Based on the value function , build a policy network The Actor-Critic architecture is used, where the Actor network outputs the probability distribution of actions, and the Critic network is used to evaluate the value of each action. Calculate the probability distribution of each action: ;

[0110] in, is the weight matrix of the Actor network, The softmax function is used to convert the output into a probability distribution. The Critic network is based on the value function To evaluate the expected value of each action. By combining the advantages of Actor and Critic, the control strategy is effectively optimized and the effect of backlight adjustment is improved. In order to obtain the optimal policy model, the policy network is trained by reinforcement learning. During the training process, the discount factor is set 0.9 is used to balance the impact of short-term rewards and long-term rewards, ensuring that the system can respond to the current state while considering the future cumulative benefits. The experience replay buffer is used to store the state transition sequence to break the correlation between training samples and improve the generalization ability of the model. Each time a small batch is randomly selected from the experience replay buffer for training, the temporal difference (TD) algorithm is used to update the parameters of the network so that the value function The policy network is trained after multiple iterations of training. After the trained policy network is obtained, the temperature value, current value and light intensity value of each target backlight sub-area are used as input, and the probability distribution of the action is obtained through forward propagation calculation. In order to find the optimal action combination in execution, - Greedy strategy for action selection. In most cases (with probability ) selects the current optimal action, and with a certain probability ( ) to maintain a certain exploration capability. In the optimization process, we can not only keep the optimal strategy, but also explore new possibilities to avoid falling into the local optimum. -After the greedy strategy selects the optimal action combination, the compensation control parameter is obtained. The compensation control parameter includes the compensation coefficient and adjustment step length The compensation control parameters are reverse mapped to convert the compensation coefficient and adjustment step size into the backlight compensation amount of each Mini LED backlight unit. For example, for a certain target backlight sub-area, the selected compensation coefficient is , the adjustment step length is , then the corresponding backlight compensation amount is calculated by the following formula:

[0111] ;

[0112] in, is the original backlight brightness value, is the brightness value after compensation. Through reverse mapping, the final brightness adjustment value of each backlight unit is obtained, thereby achieving precise control of the Mini LED backlight.

[0113] In one example, real-time feedback correction is performed on the backlight compensation amount, the deviation between the actual display parameter and the target parameter of each backlight partition is calculated, a proportional-integral-differential operation is performed, and a backlight optimization control signal is output, including:

[0114] Separate the backlight compensation amount to obtain a partition compensation parameter group, and perform real-time data collection based on the partition compensation parameter group to obtain partition real-time parameters;

[0115] The real-time parameters of the partitions are compared with the target parameters, and the brightness deviation, chromaticity deviation and uniformity deviation are obtained by subtraction operation to obtain the standardized deviation data;

[0116] Performing proportional coefficient calculation based on the standardized deviation data, multiplying the brightness deviation, chromaticity deviation and uniformity deviation by the proportional coefficients respectively to obtain proportional control parameters, and performing integral calculation and cumulative summation on the standardized deviation data to obtain integral control parameters;

[0117] Based on the standardized deviation data, differential operation is performed to calculate the parameter change rate between each two adjacent frames to obtain the differential control parameter, and the proportional control parameter, integral control parameter and differential control parameter are linearly superimposed to obtain the PID control amount;

[0118] The PID control amount is reconstructed and converted into a driving parameter adjustment value of the backlight unit to obtain a backlight optimization control signal.

[0119] In this example, the backlight compensation amount is separated and the overall compensation data is divided into multiple partition compensation parameter groups. Each partition compensation parameter group contains the backlight adjustment information of the area, such as brightness compensation coefficient, chromaticity compensation coefficient, etc. These partition compensation parameter groups are used to guide the specific backlight adjustment of each target backlight sub-area. Assume that the backlight sub-area has a total of partitions, the compensation parameters of each partition are vector Indicates that is the brightness compensation amount, is the chromaticity compensation amount, is the uniformity compensation amount, Real-time data collection is performed on each partition compensation parameter group to obtain the actual state parameters of the partition at the current moment, such as real-time brightness value, current value, chromaticity information, etc., collectively referred to as partition real-time parameters. The partition real-time parameters are compared with the target parameters, and the brightness deviation, chromaticity deviation and uniformity deviation are obtained by subtraction. Assuming the partition The target brightness is The actual brightness is , then the brightness deviation It is expressed as: ;

[0120] Similarly, assuming the target chromaticity value is The actual chromaticity value is , then the chromaticity deviation It is expressed as: ;

[0121] For uniformity deviation, assuming the target uniformity is The actual uniformity is , then the uniformity deviation for: ;

[0122] Through calculation, the brightness, chromaticity and uniformity deviations of each backlight sub-area are obtained. These deviations are normalized to obtain standardized deviation data. The normalization process is achieved by dividing the deviation by the target parameter value, such as the standardized brightness deviation for: ;

[0123] Similarly, the chromaticity deviation and uniformity deviation are normalized to obtain and These standardized deviation data are used to guide the PID operation of backlight control. Based on the standardized deviation data, the proportional coefficient operation is performed. For each deviation, a proportional coefficient is multiplied to calculate the proportional control amount. Assume that the proportional coefficient of the brightness deviation is , the proportional coefficient of chromaticity deviation is , the proportional coefficient of uniformity deviation is , the proportional control parameters are: ;

[0124] These proportional control parameters directly reflect the size of the current deviation, the proportional coefficient The choice of determines the system's response speed to the deviation. Perform an integral operation on the standardized deviation data to obtain the integral control parameter. The purpose of integral control is to eliminate the steady-state error by accumulating the correction amount when the deviation exists for a long time, so that the system can better approach the target value. For each deviation, the integral control parameter is calculated by the following formula: ;

[0125] in, are the integral coefficients of brightness, chromaticity and uniformity, and the integral control parameters Help eliminate the residual deviation in the system. At the same time, the standardized deviation data is differentiated to calculate the parameter change rate between each two adjacent frames to obtain the differential control parameter. The role of differential control is to predict the change trend of the deviation, so as to make adjustments in advance to prevent over-adjustment. Assume that the time interval of each frame is , then the differential control parameter of the brightness deviation is It is expressed as: ;

[0126] Similarly, the differential control parameters for chromaticity deviation and uniformity deviation are and is also calculated in a similar way, where are the differential coefficients of brightness, chromaticity and uniformity respectively. Linearly superimpose the proportional control parameter, integral control parameter and differential control parameter to obtain the PID control quantity. The PID control quantity for each deviation is expressed as: ;

[0127] PID control amount Combining the advantages of proportion, integration and differentiation, it is possible to comprehensively handle deviations in different situations. Proportional control enables the system to respond quickly to deviations, integral control eliminates long-term deviations, and differential control is used to predict and suppress rapid changes, making the system more stable and accurate. The PID control quantity is reconstructed and converted into the drive parameter adjustment value of the backlight unit to obtain the backlight optimization control signal. Assume that the brightness control signal is , the chroma control signal is , the uniformity control signal is , the PID control quantity is converted into the corresponding drive parameter adjustment value through the following formula:

[0128] ;

[0129] In this formula, are the basic driving values ​​of the system, and the PID control amount is the adjustment increment. In this way, the final driving parameters of each backlight unit are obtained, and the real-time brightness, chromaticity and uniformity of the display screen are optimized.

[0130] Reference Figure 2 , this embodiment provides a display backlight brightness adjustment system based on Mini QLED technology, including:

[0131] Generating module 1, used for dividing the display image signal into brightness signal, chrominance signal and saturation signal, and generating image feature data;

[0132] Detection module 2, used for performing sub-pixel boundary detection on the image feature data, obtaining multiple initial backlight sub-regions, and calculating regional brightness distribution data of each initial backlight sub-region;

[0133] A merging module 3 is used to perform regional clustering operations and regional merging on the regional brightness distribution data to obtain multiple target backlight sub-regions, and calculate backlight partition mapping data of the target backlight sub-regions;

[0134] Transformation module 4, used to perform nonlinear feature transformation on the backlight partition mapping data, establish a sub-pixel backlight mapping model, and calculate the backlight unit control data of each Mini LED backlight unit in the target backlight sub-area;

[0135] A construction module 5 is used to input the backlight unit control data into a distributed control network, construct a state space matrix and an action space matrix for each target backlight sub-region, and output a backlight compensation amount for each target backlight sub-region using a deep Q network;

[0136] The output module 6 is used to perform real-time feedback correction on the backlight compensation amount, calculate the deviation value between the actual display parameter and the target parameter of each target backlight sub-area, perform proportional-integral-differential operation, and output a backlight optimization control signal.

[0137] In this embodiment, for the specific implementation of each unit in the above system embodiment, please refer to the description in the above method embodiment, which will not be repeated here.

[0138] Reference Figure 3 In an embodiment of the present invention, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Figure 3 As shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0139] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0140] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0141] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided by the present invention and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM.

[0142] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0143] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A display backlight brightness adjustment method based on Mini QLED technology, characterized in that: The following steps are involved: The display image signal is divided into a brightness signal, a chrominance signal and a saturation signal, and image feature data is generated; Performing sub-pixel boundary detection on the image feature data to obtain a plurality of initial backlight sub-regions, and calculating regional brightness distribution data of each of the initial backlight sub-regions; Performing regional clustering operation and regional merging on the regional brightness distribution data to obtain a plurality of target backlight sub-regions, and calculating backlight partition mapping data of the target backlight sub-regions; Specifically, it includes: performing regional adjacent relationship analysis on the regional brightness distribution data, calculating the brightness difference and brightness distribution correlation between each pair of adjacent initial backlight sub-regions, and obtaining a regional correlation index; judging regional merging conditions based on the regional correlation index, screening the brightness difference of adjacent regions according to a first preset threshold value T1, and screening the brightness distribution correlation of adjacent regions according to a second preset threshold value T2, and obtaining a pair of regions to be merged; performing regional growth calculation on the pair of regions to be merged, marking the connected domains of adjacent regions that meet the merging conditions, and merging regions with the same markings into a new backlight partition to obtain multiple target backlight sub-regions; performing regional boundary extraction based on the target backlight sub-regions, and filtering the boundary pixels of each target backlight sub-region. The method comprises the following steps: marking the row positioning marks, calculating the brightness gradient direction and amplitude of the boundary pixels, and obtaining the boundary feature data; performing brightness gradient analysis on the boundary feature data, calculating the brightness transition function between adjacent target backlight sub-regions along the boundary gradient direction, and determining the width range of the gradient region, obtaining the gradient control parameters, and performing brightness interpolation operation based on the gradient control parameters, calculating the brightness distribution of the transition region in the boundary region of the adjacent target backlight sub-regions, and obtaining the gradient mapping data; performing energy balancing processing on the gradient mapping data, calculating the total energy and boundary energy loss of each target backlight sub-region, obtaining the balanced compensation data, and fusing the balanced compensation data with the brightness distribution data of each target backlight sub-region, and obtaining the backlight partition mapping data of each target backlight sub-region; Performing nonlinear feature transformation on the backlight partition mapping data, establishing a sub-pixel level backlight mapping model, and calculating the backlight unit control data of each Mini QLED backlight unit in the target backlight sub-area; Inputting the backlight unit control data into a distributed control network, constructing a state space matrix and an action space matrix for each target backlight sub-region, and using a deep Q network to output a backlight compensation amount for each target backlight sub-region; The backlight compensation amount is corrected by real-time feedback, the deviation value between the actual display parameter and the target parameter of each target backlight sub-area is calculated, a proportional-integral-differential operation is performed, and a backlight optimization control signal is output.

2. The display backlight brightness adjustment method based on Mini QLED technology according to claim 1, characterized in that: The display image signal is divided into a brightness signal, a chrominance signal and a saturation signal, and image feature data is generated, including: Performing RGB color space conversion on the display image signal to obtain RGB three-channel data, and performing matrix operation on the RGB three-channel data to convert the RGB three-channel data into a YUV color space to obtain a brightness signal, a chrominance signal, and a saturation signal; Performing Gaussian filtering on the brightness signal, using a Gaussian template with a kernel function size of 5×5 to remove image noise, obtaining filtered brightness data, and performing an asymmetric Gaussian distribution function operation on the filtered brightness data to obtain brightness distribution feature data; Normalizing the chrominance signal, adjusting the value range of the chrominance signal to the interval [-1, 1], and obtaining a standardized chrominance parameter; and normalizing the saturation signal, adjusting the value range of the saturation signal to the interval [0, 1], and obtaining a standardized saturation parameter; The brightness distribution characteristic data, the standardized chromaticity parameter and the standardized saturation parameter are combined into an initial characteristic vector, and a sub-pixel level interpolation operation is performed on the initial characteristic vector to obtain image characteristic data.

3. The display backlight brightness adjustment method based on Mini QLED technology according to claim 2 is characterized in that: The performing sub-pixel level boundary detection on the image feature data to obtain a plurality of initial backlight sub-regions, and calculating the regional brightness distribution data of each of the initial backlight sub-regions, comprises: Performing sub-pixel level gradient calculation on the image feature data to obtain sub-pixel sampling data, and performing adaptive threshold processing on the sub-pixel sampling data to obtain edge candidate points; Performing region segmentation based on the edge candidate points, dividing the display region into N×M initial backlight sub-regions according to brightness uniformity, where N and M are the number of divisions of the display region in the horizontal and vertical directions, respectively, to obtain initial partition data; Calculate the brightness mean of each initial backlight sub-region in the initial partition data, perform weighted average on the brightness values ​​of all pixels in each initial backlight sub-region to obtain the brightness value of the backlight sub-region, perform standard deviation analysis based on the brightness value of the backlight sub-region, normalize the deviation between the pixel points in each initial backlight sub-region and the regional mean, and obtain the brightness fluctuation coefficient; Performing threshold classification on the brightness fluctuation coefficient to obtain regional feature levels, and constructing a regional brightness association matrix based on the regional feature levels, taking the feature level differences of adjacent regions as association measurement indicators, and obtaining regional association degree data; The spatial distribution analysis is performed on the regional correlation data, and the local connectivity and global distribution characteristics of each initial backlight sub-region are calculated to obtain the regional brightness distribution data of each initial backlight sub-region.

4. The display backlight brightness adjustment method based on Mini QLED technology according to claim 1, characterized in that: The performing of nonlinear feature transformation on the backlight partition mapping data, establishing a sub-pixel level backlight mapping model, and calculating the backlight unit control data of each Mini QLED backlight unit in the target backlight sub-area includes: Performing sub-pixel sampling on the backlight partition mapping data, sampling each backlight partition into sub-pixel points with 10 times the density in the horizontal and vertical directions, respectively, to obtain a sub-pixel brightness matrix; Based on the sub-pixel brightness matrix, a nonlinear mapping transformation is performed, and a segmented S-shaped curve is used to map the brightness value to the physical driving space of the backlight unit, wherein the slopes of the S-shaped curve in the dark and bright areas are 0.6 and 1.2 respectively, to obtain a backlight driving curve; Perform sub-pixel backlight modeling on the backlight driving curve, map the brightness value of the sub-pixel point to the Mini QLED backlight unit array according to the physical distribution law, establish a one-to-many mapping relationship from the sub-pixel point to the backlight unit, and obtain a sub-pixel backlight mapping model; The driving current is calculated based on the sub-pixel backlight mapping model, the brightness of the sub-pixel points within the coverage area of ​​each Mini QLED backlight unit is weighted averaged, and the weighted result is converted into an 8-bit quantized driving current value to obtain a current control sequence; Performing pulse width modulation calculation on the current control sequence, converting the driving current value into a PWM signal with a duty cycle in the range of 10% to 90% according to 256 levels of brightness grading, obtaining a pulse width control sequence, and performing phase modulation based on the pulse width control sequence, staggering the phases of the PWM signals of adjacent backlight units by 120 degrees, obtaining phase control data; Performing temperature compensation on the phase control data, correcting the driving current value and the duty cycle according to the temperature coefficient of the backlight unit position, dynamically adjusting the temperature coefficient within the range of 0.8 to 1.2, and obtaining the compensated control parameters; The compensated control parameters are encoded in parallel to generate a 32-bit control word including current value, duty cycle and phase information, which is sent to the driving circuit through a serial interface to obtain backlight unit control data.

5. The display backlight brightness adjustment method based on Mini QLED technology according to claim 4 is characterized in that: The step of inputting the backlight unit control data into a distributed control network, constructing a state space matrix and an action space matrix for each target backlight sub-region, and using a deep Q network to output a backlight compensation amount for each target backlight sub-region comprises: The backlight unit control data is sampled in parallel, the temperature value, current value and light intensity value of each target backlight sub-area are collected, a three-dimensional state vector is constructed, and a state space matrix St is obtained through multi-layer perceptron mapping; Based on the state space matrix, the action space is constructed, the compensation coefficient α is discretized into 8 levels in the range of [0.8, 1.2], the adjustment step length β is discretized into 4 levels in the range of [0.05, 0.2], and the action space matrix At is obtained through tensor product operation; The state space matrix St and the action space matrix At are input into the deep Q network, and features are extracted through three convolution layers, where the convolution kernel sizes are 3×3, 2×2, and 1×1, respectively, and the step size is 1, to obtain feature mapping data; Performing a fully connected layer transformation on the feature mapping data, mapping the feature vector to the state-action value space through two hidden layers, with the number of neurons in the hidden layers being 256 and 128 respectively, to obtain the value function Q(St,At); Based on the value function Q(St,At), a policy network π(At|St) is constructed, using an Actor-Critic architecture, where the Actor network outputs the action probability distribution and the Critic network evaluates the action value to obtain a policy model; Perform reinforcement learning training on the policy model, set the discount factor γ to 0.9, use the experience replay buffer to store the state transition sequence, and update the network parameters through the temporal difference algorithm to obtain the trained policy network; The temperature value, current value and light intensity value of each target backlight sub-area are input into the trained strategy network, the action probability distribution is obtained by forward propagation calculation, the ε-greedy strategy is used to select the optimal action combination, the compensation control parameters are obtained, and the compensation control parameters are reverse mapped to convert the compensation coefficient and adjustment step into the backlight compensation amount of each Mini QLED backlight unit.

6. The method for adjusting the brightness of a display screen backlight based on Mini QLED technology according to claim 5, characterized in that: The real-time feedback correction of the backlight compensation amount, calculating the deviation between the actual display parameter of each backlight partition and the target parameter, performing proportional-integral-differential operation, and outputting a backlight optimization control signal include: Separating the backlight compensation amount to obtain a partition compensation parameter group, and performing real-time data acquisition based on the partition compensation parameter group to obtain a partition real-time parameter; Performing target parameter comparison on the partition real-time parameters, obtaining brightness deviation, chromaticity deviation and uniformity deviation through subtraction operation, and obtaining standardized deviation data; Performing a proportional coefficient operation based on the standardized deviation data, multiplying the brightness deviation, the chromaticity deviation and the uniformity deviation by the proportional coefficients respectively to obtain a proportional control parameter, and performing an integral operation and cumulative summation on the standardized deviation data to obtain an integral control parameter; Performing a differential operation based on the standardized deviation data, calculating the parameter change rate between each two adjacent frames, obtaining a differential control parameter, and linearly superimposing the proportional control parameter, the integral control parameter and the differential control parameter to obtain a PID control amount; The PID control amount is parameter reconstructed, and the PID control amount is converted into a driving parameter adjustment value of the backlight unit to obtain a backlight optimization control signal.

7. A display screen backlight brightness adjustment system based on Mini QLED technology, characterized in that: For implementing the steps of the method according to any one of claims 1 to 6, the system comprises: A generating module, used for dividing the display image signal into a brightness signal, a chrominance signal and a saturation signal, and generating image feature data; A detection module, configured to perform sub-pixel boundary detection on the image feature data to obtain a plurality of initial backlight sub-regions, and calculate regional brightness distribution data of each of the initial backlight sub-regions; A merging module is used to perform regional clustering operations and regional merging on the regional brightness distribution data to obtain multiple target backlight sub-regions, and calculate the backlight partition mapping data of the target backlight sub-regions; specifically comprising: performing regional adjacent relationship analysis on the regional brightness distribution data, calculating the brightness difference and brightness distribution correlation between each pair of adjacent initial backlight sub-regions, and obtaining a regional correlation index; judging regional merging conditions based on the regional correlation index, screening the brightness difference of adjacent regions according to a first preset threshold value T1, and screening the brightness distribution correlation of adjacent regions according to a second preset threshold value T2, to obtain a pair of regions to be merged; performing regional growth calculation on the pair of regions to be merged, marking the connected domains of adjacent regions that meet the merging conditions, and merging regions with the same mark into a new backlight partition to obtain multiple target backlight sub-regions; based on the The target backlight sub-region is subjected to regional boundary extraction, the boundary pixels of each target backlight sub-region are positioned and marked, and the brightness gradient direction and amplitude of the boundary pixels are calculated to obtain boundary feature data; the boundary feature data is subjected to brightness gradient analysis, the brightness transition function between adjacent target backlight sub-regions is calculated along the boundary gradient direction, and the width range of the gradient region is determined to obtain gradient control parameters, and brightness interpolation operation is performed based on the gradient control parameters, and the brightness distribution of the transition region is calculated in the boundary region of adjacent target backlight sub-regions to obtain gradient mapping data; the gradient mapping data is subjected to energy balancing processing, the total energy and boundary energy loss of each target backlight sub-region are calculated to obtain balancing compensation data, and the balancing compensation data is merged with the brightness distribution data of each target backlight sub-region to obtain backlight partition mapping data of each target backlight sub-region; A transformation module, used to perform nonlinear feature transformation on the backlight partition mapping data, establish a sub-pixel backlight mapping model, and calculate the backlight unit control data of each Mini QLED backlight unit in the target backlight sub-area; A construction module, used for inputting the backlight unit control data into a distributed control network, constructing a state space matrix and an action space matrix for each target backlight sub-region, and outputting a backlight compensation amount for each target backlight sub-region using a deep Q network; The output module is used to perform real-time feedback correction on the backlight compensation amount, calculate the deviation value between the actual display parameter and the target parameter of each target backlight sub-area, perform proportional-integral-differential operation, and output a backlight optimization control signal.

8. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images