A backlight source zone dimming control method

Abstract

The present invention discloses a backlight light source partition dimming control method, which belongs to the field of data security. The specific steps of the control method are as follows: I. Analyze the input image data characteristics, identify the highlight motion area, and predict and optimize the current backlight partition dimming strategy; II. Adjust the backlight area in real time according to the brightness, contrast and color changes of the input content, and construct a multi-level light diffusion coupling model after the partition adjustment; the present invention enables the backlight partition to be adaptively adjusted according to the image content, improves local contrast and energy efficiency, achieves more uniform light distribution, reduces halo and brightness unevenness problems, improves dimming accuracy and extends the life of the light source, effectively reduces the light spillover effect at the edge of the partition, and enhances the layering of the picture; effectively improves the subjective picture quality experience, improves the light efficiency utilization rate, and can take into account both picture quality and power consumption optimization, so that the brightness adjustment is more in line with the visual characteristics of the human eye, reduces visual fatigue, and ensures the best picture quality at the lowest power consumption.

Description

Technical Field

[0001] The present invention relates to the field of data security, and in particular to a backlight light source partition dimming control method. Background Art

[0002] In recent years, with the rapid development of display technology and the increasing demand for high dynamic range (HDR) displays, ultra-high definition (UHD), and low-power displays, the requirements for backlight control in display devices have become increasingly stringent. Traditional global backlight dimming methods, due to their lack of independent control over local areas, can easily lead to problems such as insufficient black level performance, reduced contrast, and excessive energy consumption. In contrast, local dimming technology, by independently controlling backlight zones, can significantly improve image contrast and reduce backlight power consumption. However, current mainstream local dimming technologies still struggle to adapt to varying image content, resulting in insufficient brightness control accuracy. They also fail to fully consider the human eye's perception of varying brightness and contrast, impacting the subjective visual experience. Using a single LED backlight solution makes it difficult to achieve an optimal balance between energy efficiency and image quality. Furthermore, high-density LED local dimming, while increasing brightness, can also lead to significant heat generation, impacting system stability and lifespan. Therefore, developing a local dimming control method for backlight light sources is crucial.

[0003] After searching, Chinese patent number CN117935743A discloses a backlight light source partition dimming control method, circuit and LED driver chip. Although the invention can realize the adaptive numbering of cascaded LED driver chips; the brightness update command flexibly sets the start / end chip number, so that only the LED driver chips within the number range are updated with brightness, realizing single / batch LED driver chip brightness refresh, and the dimming method is more flexible and efficient, but the backlight partition cannot be adaptively adjusted according to the image content, which reduces local contrast and energy efficiency, cannot evenly distribute light, and is prone to halo and uneven brightness problems, while reducing dimming accuracy and shortening light source life; in addition, the existing backlight light source partition dimming control method cannot improve the subjective picture quality experience, reduces light efficiency utilization, cannot take into account both picture quality and power consumption optimization, and increases visual fatigue; for this reason, we propose a backlight light source partition dimming control method. Summary of the Invention

[0004] The purpose of the present invention is to solve the defects in the prior art and to propose a backlight light source zone dimming control method.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A backlight light source zone dimming control method, the specific steps of the control method are as follows:

[0007] Ⅰ. Analyze the input image data features, identify the highlight motion area, and predict and optimize the current backlight partition dimming strategy;

[0008] II. Adjust the backlight area in real time based on the brightness, contrast, and color changes of the input content, and build a multi-level light diffusion coupling model after the partition adjustment;

[0009] III. Analyze the human eye's sensitivity to different brightness and contrast levels, adjust local brightness weights, and optimize the brightness mapping relationship of backlight partitions;

[0010] IV. Allocate the best light source type and driving mode based on scene characteristics, monitor the backlight module temperature in real time, and dynamically adjust the driving current based on the monitoring data;

[0011] V. Based on the optimized brightness zoning and dimming control strategy, independent dimming control strategies are adopted for different backlight areas;

[0012] VI. Continuously monitor the output image quality parameters, re-predict the backlight zone dimming strategy based on the scene, and form an adaptive closed-loop control.

[0013] As a further solution of the present invention, the specific steps of identifying the highlight motion area in step I are as follows:

[0014] S1.1: Perform brightness normalization on each input image frame, randomly select a set of gamma correction coefficients from the range [0.8, 2.2], adjust the image contrast based on the selected gamma correction coefficients, and then divide the corrected image data into multiple non-overlapping sub-regions;

[0015] S1.2: Calculate the grayscale histogram of each subregion to obtain the grayscale distribution of each subregion, calculate the cumulative distribution value (CDF) of the histogram of each subregion, then clip the CDF value based on a preset contrast limit threshold, and calculate the pixel value of the corresponding subregion after histogram equalization based on each clipped CDF value. Then, smooth the boundaries of each equalized subregion through bilinear interpolation and reconstruct the complete image;

[0016] S1.3: Use a white balance algorithm based on the grayscale world assumption to maintain color consistency between image data frames. Then, use non-local mean filtering to remove noise from the image data. Calculate the optical flow field between the processed image data frames to obtain pixel motion information. Optical flow field information that exceeds the preset highlight brightness threshold and motion amplitude threshold is identified as a highlight motion area and marked.

[0017] As a further solution of the present invention, the specific steps of predicting and optimizing the current backlight zone dimming strategy in step I are as follows:

[0018] S2.1: Initialize the policy adjustment model based on the LSTM model architecture, including the input layer, LSTM layer, and output layer. Use the CNN network to extract the temporal features of each frame of preprocessed image data and associate each temporal feature with its corresponding highlight motion region. Then, divide the generated temporal features into training and test sets, and then input the training set into the policy adjustment model.

[0019] S2.2: The input layer of the strategy adjustment model receives the training set data and divides the training set data into multiple batches of training groups. The training groups are then forward-propagated to the LTSM layer in sequence. Each training group is then processed through the forget gate, input gate, and output gate in the LSTM layer. The final result output by the output gate is nonlinearly processed using the Softmax function of the fully connected layer to obtain the final predicted scene recognition result and the highlight motion area within the preset time interval, which is then output through the output layer.

[0020] S2.3: Use the cross-entropy loss function to calculate the true scene and highlight motion region labels of the current frame and the predicted scene recognition results and highlight motion regions. Input the calculated loss value into the output layer of the policy adjustment model. Backpropagate based on the chain rule and calculate the gradient of the loss value for each network layer of the policy adjustment model in turn. Then, use the Adam optimizer to adjust the parameters of each network layer.

[0021] S2.4: After each round of training, the accuracy, precision, recall, and F1 score of the policy adjustment model are evaluated using the test set data, and a comprehensive score is generated. If the score does not reach the preset score threshold, the policy adjustment model is retrained and tested until the model loss value converges to the preset range, at which point training is stopped.

[0022] S2.5: Use an external memory module to record historical task information of the strategy adjustment model, namely, the features and model parameters of known scenes. Then, use the strategy adjustment model to perform time series modeling on the image data of each frame in the past. Use the forward propagation algorithm to generate the highlight motion area prediction results and scene recognition results within a preset time period. Based on the identified scene category and predicted highlight area, calculate the brightness gain of each partition, and adjust and optimize the original backlight partition dimming strategy based on the calculated brightness gain.

[0023] S2.6: When the loss value between the output result of the policy adjustment model and the actual result exceeds the preset threshold, the cosine similarity between each historical task information stored in the external memory module and the current image data is calculated, and the enhanced hidden state is obtained from the historical task information with the highest cosine similarity. Based on the selected enhanced hidden state and the current loss value, the loss value is recalculated, and the gradient of the current task is calculated based on the latest loss value. The policy adjustment model parameters are re-updated based on the calculated gradient, and the historical task information recorded in the memory module is updated.

[0024] As a further solution of the present invention, the strategy adjustment model described in S2.1 constructs a model of the corresponding level based on the input layer, multi-layer LSTM layer and output layer of the LSTM model, and the processing flow of each layer of the strategy adjustment model is that the input layer receives the extracted feature data, and the input gate, forgetting gate and output gate of the multi-layer LSTM layer process the extracted sets of feature data and generate corresponding hidden states. After that, the fully connected unit in the output layer performs nonlinear processing on the hidden state through the LeakyReLU activation function to generate the brightness coefficient of the backlight partition, dimming direction recommendation and dynamic dimming amplitude, and output them.

[0025] As a further solution of the present invention, the specific steps of adjusting the backlight area in real time in step II are as follows:

[0026] S3.1: After adjusting the brightness of each partition using the optimized backlight partition dimming strategy, calculate the local brightness requirement of the current frame image through weighted averaging. Then, divide the entire image into multiple grid areas and calculate the average brightness of each grid area.

[0027] S3.2: Set a brightness threshold. If the average brightness of a grid is higher than the brightness threshold, use dense honeycomb partitioning to subdivide the area. If the average brightness of a grid is lower than the brightness threshold, merge adjacent areas and use sparse rectangular partitioning. After traversing all grids, obtain the adjusted number of backlight partitions.

[0028] S3.3: After the dynamic partition adjustment is completed, the center points of each partition are extracted, and each center point represents a backlight adjustment area. The entire screen area is divided into multiple polygonal partitions using the Voronoi diagram algorithm;

[0029] S3.4: Calculate the centroid of each partition using the Lloyd iteration algorithm, and use the new centroid as a new Voronoi generation point. Repeat the Voronoi calculation and Lloyd iteration until the position change value of the Voronoi generation point converges to a preset range, and output the final irregular backlight partition.

[0030] As a further solution of the present invention, the specific calculation formula of the weighted average in S3.1 is as follows:

[0031] ;

[0032] Where, Representative pixel The brightness value at ; represents the channel weight; Represents the pixel in the color channel (Red R, Green G, Blue B) values;

[0033] The specific calculation formula of the Voronoi diagram algorithm described in S3.3 is as follows:

[0034] ;

[0035] Where, represents a set of real numbers on a two-dimensional plane; p represents a two-dimensional coordinate point in the two-dimensional plane; Representative point p to the center point distance; Representative point p to the center point distance;

[0036] The specific calculation formula of the Lloyd iterative algorithm described in S3.4 is as follows:

[0037] ;

[0038] Where, Representative The center point of the region; Represents the first segmentation by the Voronoi diagram algorithm regions; Representative A two-dimensional coordinate point in a region.

[0039] As a further solution of the present invention, the specific steps of constructing the multi-level light diffusion coupling model in step II are as follows:

[0040] S4.1: Set the initial radiation intensity distribution of the backlight module's light source. Then, light from the light source enters the light guide plate. Calculate the multiple scattering and absorption behavior of light in the light guide plate using the radiation transfer equation to obtain the light diffusion process.

[0041] S4.2: Utilize the prisms of the light guide plate and the microstructures of the diffuser film to perform additional directional adjustments and intensity redistribution of light, modeled using a transmission function. Then, model the light homogenization process of the diffuser film using a diffusion equation. By integrating the various light diffusion processes, a multi-level light diffusion coupling model is constructed.

[0042] S4.3: Based on a multi-level light diffusion coupling model, randomly sample the light emission angles and positions and discretize them to simulate light propagation in the backlight module. The discretized results are used as the initial ray tracing.

[0043] S4.4: Based on the light propagation equation, simulate the propagation process of light reflection, refraction, and scattering in the backlight module to obtain the light diffusion effect between different partitions. Based on the ray tracing results of each partition of the backlight light source, obtain the light diffusion ratio between different partitions and establish the corresponding light diffusion transfer matrix.

[0044] S4.5: Calculate the final brightness of the corresponding backlight partition through the light diffusion transfer matrix and the partition driving current. According to the set target brightness of each backlight partition, construct the target brightness equation of each partition, and repeatedly solve the target brightness equation through least squares to calculate the optimal compensation current until the difference between the partition brightness and the target brightness converges to within the preset range. Stop solving, and then clip the calculated compensation current according to the minimum and maximum values of the driving current to generate the final compensation current to adjust the brightness of each backlight partition.

[0045] As a further solution of the present invention, the specific calculation formula of the radiation transfer equation in S4.1 is as follows:

[0046] ;

[0047] Where, Represents the light at the position At, along the direction The light intensity, where Represents the angle between the light direction and the normal direction, Represents the angle between the light on the horizontal plane and the preset reference direction; ds represents the infinitesimal element of the light propagation path; Represents the absorption loss of light in the light guide plate; Represents the degree of light scattering in the light guide plate; Represents the probability distribution of the angle change of light when it is scattered; Represents the unit direction space;

[0048] The specific calculation formula of the transmission function described in S4.2 is as follows:

[0049] ;

[0050] Where, Represents the position of the microstructure layer Light intensity on Represents the position of the microstructure layer Transmittance distribution on ; Represents the position of the light guide plate output to the microstructure layer Light intensity on

[0051] The specific calculation formula of the diffusion equation described in S4.2 is as follows:

[0052] ;

[0053] Where, Represents the diffusion membrane position The light intensity on the film; t represents the diffusion time; D represents the diffusion rate of light in the film; The Laplacian operator representing light intensity measures local brightness changes; represents the attenuation coefficient of light;

[0054] The specific calculation formula of the multi-level light diffusion coupling model described in S4.2 is as follows:

[0055] ;

[0056] Where, Represents the final output light intensity distribution; represents the diffusion calculation of the diffusion membrane; Transmission calculation representing the microstructure layer; Represents the calculation of radiation transfer in the light guide; Represents the initial light intensity distribution.

[0057] As a further solution of the present invention, the thickness of the backlight module in S4.1 is within the range of [1.5mm, 5mm], and the horizontal size of the backlight module is within the range of [50mm 2 , 500mm 2 ] range, the wavelength of light is in the range of [380nm, 780nm], and the initial emission angle of light is in the range of [0, 90 ∘ ] range.

[0058] As a further solution of the present invention, the specific steps of adjusting the local brightness weight in step III are as follows:

[0059] S5.1: Convert the input image to a grayscale image based on the red (R), green (G), and blue (B) channel values of each image data. Convert the grayscale image from the spatial domain to the frequency domain using a two-dimensional fast Fourier transform. Center the converted image data to move the zero-frequency component to the center of the image.

[0060] S5.2: Decompose the spatial frequency of the image into low-frequency, medium-frequency, and high-frequency components using a low-pass filter, a band-pass filter, and a high-pass filter, respectively. Then, perform an inverse Fourier transform to convert the extracted frequency components back to the spatial domain and generate a corresponding reconstructed image. The reconstructed image is then normalized to obtain the perceived brightness value of each pixel.

[0061] S5.3: Calculate the local contrast of each image using the Weber contrast model. Calculate the horizontal and vertical spatial frequencies of each image based on the image width, height, and perceived brightness of each pixel, the pixel size of each feature in the image, the distance from the viewer to the screen, and the viewing angle.

[0062] S5.4: Calculate the contrast sensitivity of the human eye to different spatial frequencies using the CSF function. Use the contrast sensitivity to obtain the contrast sensitivity threshold for each frame of the image at the current frequency. Plot a curve for the human eye's visual perception characteristics. Calculate the brightness weight of the corresponding pixel based on the contrast, spatial frequency, and CSF function value. Adjust the local brightness of the image based on the calculated brightness weight.

[0063] As a further solution of the present invention, the specific calculation formula of the CSF function described in S5.4 is as follows:

[0064] ;

[0065] Where, represents the contrast sensitivity at spatial frequency f; f represents the spatial frequency; Represents the average brightness of the image background; as well as represent the empirical fitting parameters.

[0066] As a further solution of the present invention, the specific steps of allocating the optimal light source type and driving mode according to scene characteristics in step IV are as follows:

[0067] S6.1: Analyze the currently displayed image information and extract scene features that affect backlight control, including brightness distribution, contrast, dynamic range, and color information. Then, calculate the lighting requirements for different areas based on the brightness histogram and local contrast.

[0068] S6.2: Calculates the brightness dynamic range based on scene characteristics and selects the corresponding light source type, including Mini-LED, OLED, and Micro-LED, based on the brightness and dynamic range of the currently displayed image. It then optimizes the power consumption and image quality of each backlight zone through local dimming or global dimming drive modes.

[0069] S6.3: Based on the power consumption of each light source under different driving modes, use different light sources for combined driving, use sensors to monitor the screen output brightness, and evaluate the energy efficiency of light source selection and partition driving strategy, and readjust the light source output based on the energy efficiency evaluation results.

[0070] Compared with the prior art, the present invention has the following beneficial effects:

[0071] 1. The backlight source partition dimming control method calculates the brightness requirements of each partition according to the optimized backlight partition dimming strategy, divides the image area by grid, sets the threshold according to the average brightness of the grid, adopts honeycomb dense partitioning for the highlight area, and merges the low-brightness area into sparse rectangular partitioning. After traversing all the grids, the number of dynamically adjusted backlight partitions is obtained, and the partition center of gravity is iteratively optimized until the position change converges to generate the final irregular backlight partition. Subsequently, the initial radiation intensity of the backlight module light source is set, and the optical properties of the light guide plate prism and the diffusion film are combined to model the multi-level light diffusion coupling through the transmission function and diffusion equation. The system performs a combination process and discretizes the light emission angle and position to simulate the propagation of light in the backlight module, and establishes a light diffusion transfer matrix to solve the final brightness of the backlight partition. The optimal compensation current is calculated based on the target brightness equation and the least squares method, and iterative optimization is performed until the brightness error converges. The calculated compensation current is clipped to enable the backlight partition to be adaptively adjusted according to the image content, improve local contrast and energy efficiency, achieve more uniform light distribution, reduce halo and uneven brightness problems, improve dimming accuracy and extend light source life, effectively reduce the light spill effect at the edge of the partition, and enhance the layering of the picture.

[0072] 2. This backlight source zone dimming control method converts the RGB channels of the input image into a grayscale image and transforms it from the spatial domain to the frequency domain using a two-dimensional fast Fourier transform. The transformed image data is then centered and decomposed into low-frequency, medium-frequency, and high-frequency components using low-pass, band-pass, and high-pass filters. Each frequency component is then reconstructed through an inverse Fourier transform and normalized to obtain the pixel-perceived brightness value. Local contrast is calculated based on the Weber contrast model, and the horizontal and vertical spatial frequencies are calculated based on image size, viewing distance, and viewing angle. A contrast sensitivity function is used to calculate the human eye's perceptual sensitivity threshold for different spatial frequencies. A visual perception characteristic curve is then plotted to calculate the brightness weights of the corresponding pixels, thereby optimizing local brightness. The brightness distribution, contrast, dynamic range, and color information of the currently displayed image are analyzed. The lighting requirements of different regions are assessed based on the brightness histogram and local contrast, and the overall brightness dynamic range is calculated. According to the characteristics of the scene, the light source type is selected, and local dimming or global dimming mode is used to optimize the backlight partition power consumption and picture quality, effectively improving the subjective picture quality experience and increasing the utilization rate of light efficiency. At the same time, it can take into account the optimization of picture quality and power consumption, making the brightness adjustment more in line with the visual characteristics of the human eye, reducing visual fatigue, and ensuring the best picture quality at the lowest power consumption. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0074] Figure 1This is a flowchart of a backlight light source zone dimming control method proposed by the present invention. DETAILED DESCRIPTION

[0075] Example 1, reference Figure 1 A backlight source zone dimming control method is provided, wherein the specific steps of the control method are as follows:

[0076] Analyze the input image data features, identify the highlight motion area, and predict and optimize the current backlight zone dimming strategy.

[0077] Specifically, each frame of input image data is brightness normalized. A set of gamma correction coefficients is randomly selected from the range of [0.8, 2.2]. The contrast of the image is adjusted based on the selected gamma correction coefficients. The corrected image data is then divided into multiple non-overlapping sub-regions. The grayscale histogram of each sub-region is calculated to obtain the grayscale distribution of each sub-region. The cumulative distribution value (CDF) of the histogram of each sub-region is calculated. The CDF value is then clipped based on a preset contrast limit threshold. The pixel values of the corresponding sub-region after histogram equalization are calculated based on the clipped CDF values. The boundaries of each equalized sub-region are then smoothed using bilinear interpolation to reconstruct the complete image. A white balance algorithm based on the grayscale world assumption is used to ensure color consistency between different frames of image data. Non-local mean filtering is then used to remove noise from the image data. The optical flow field between the processed image data is calculated to obtain pixel motion information. Optical flow field information that exceeds the preset highlight brightness threshold and motion amplitude threshold is identified as a highlight motion area and marked.

[0078] Specifically, the policy adjustment model is initialized based on the LSTM model architecture, including the input layer, LSTM layer and output layer. The temporal features of each frame image data after preprocessing are extracted through the CNN network, and each temporal feature is associated with its corresponding highlight motion area. The generated temporal features are then divided into a training set and a test set, and the training set is input into the policy adjustment model. The input layer of the policy adjustment model receives the training set data and divides the training set data into multiple batches of training groups. At the same time, the training groups are forward propagated to the LTSM layer in turn, and then each training group is processed through the forget gate, input gate and output gate in the LSTM layer in turn. The final result of the output of the output gate is processed nonlinearly by the Softmax function of the fully connected layer to obtain the final predicted scene recognition result and the highlight motion area within the preset time interval, and output through the output layer. The real scene of the current frame and the highlight motion area label are calculated with the predicted scene recognition result and the highlight motion area through the cross entropy loss function. The calculated loss value is input from the output layer of the policy adjustment model, and back propagation is performed based on the chain rule. The gradient of each network layer of the loss value for the policy adjustment model is calculated in turn. The parameters of each network layer are then adjusted by the Adam optimizer. After each round of training, the gradient of each network layer is calculated by the Adam optimizer. The accuracy, precision, recall rate and F1 score of the strategy adjustment model are evaluated through the test set data, and a comprehensive score is generated. If the score does not reach the preset score threshold, the strategy adjustment model is retrained and tested until the model loss value converges to the preset range. The training is stopped and the external memory module is used to record the historical task information of the strategy adjustment model, that is, the features and model parameters of the known scenes. Then, the strategy adjustment model is used to perform time series modeling on the image data of the past frames, and the forward propagation algorithm is used to generate the highlight motion area prediction results and scene recognition results within the preset time period. According to the recognized scene category and the predicted highlight area , calculate the brightness gain of each partition, and adjust and optimize the original backlight partition dimming strategy based on the calculated brightness gain. When the loss value between the output result of the strategy adjustment model and the actual result exceeds the preset threshold, calculate the cosine similarity between each historical task information stored in the external memory module and the current image data, and obtain the enhanced hidden state from the historical task information with the highest cosine similarity. Based on the selected enhanced hidden state and the current loss value, recalculate the loss value, and calculate the gradient of the current task based on the latest loss value. Based on the calculated gradient, re-update the strategy adjustment model parameters and update the historical task information recorded in the memory module.

[0079] It should be further explained that the policy adjustment model constructs a model of the corresponding level based on the input layer, multi-layer LSTM layer and output layer of the LSTM model, and the processing flow of each layer of the policy adjustment model is that the input layer receives the extracted feature data, and the input gate, forget gate and output gate of the multi-layer LSTM layer process the extracted sets of feature data and generate the corresponding hidden state. After that, the fully connected unit in the output layer performs nonlinear processing on the hidden state through the LeakyReLU activation function to generate the brightness coefficient of the backlight partition, dimming direction recommendation and dynamic dimming amplitude, and output them.

[0080] According to the brightness, contrast and color changes of the input content, the backlight area is adjusted in real time, and after the partition adjustment, a multi-level light diffusion coupling model is constructed.

[0081] Specifically, after adjusting the brightness of each partition through the optimized backlight partition dimming strategy, the local brightness requirement of the current frame image is calculated by weighted average, and then the entire image is divided into multiple grid areas, and the average brightness of each grid is calculated, and the brightness threshold is set. If the average brightness of the grid is higher than the brightness threshold, the area is subdivided by honeycomb dense partitioning. If the grid average brightness is lower than the brightness threshold, the adjacent areas are merged and sparse rectangular partitioning is used. After traversing all grids, the adjusted number of backlight partitions is obtained. After the dynamic partition adjustment is completed, the center point of each partition is extracted, and each center point represents a backlight adjustment area. The entire screen area is divided into multiple polygonal partitions through the Voronoi diagram algorithm, and the center of gravity of each partition is calculated through the Lloyd iterative algorithm. The new center of gravity is used as the new Voronoi generation point, and the Voronoi calculation and Lloyd iteration are repeated until the position change value of the Voronoi generation point converges to the preset range, and the final generated irregular backlight partition is output.

[0082] In addition, in this embodiment, the specific calculation formula of the weighted average is as follows:

[0083] ;

[0084] Where, Representative pixel The brightness value at ; represents the channel weight; Represents the pixel in the color channel (Red R, Green G, Blue B) values;

[0085] The specific calculation formula of the Voronoi diagram algorithm is as follows:

[0086] ;

[0087] Where, represents a set of real numbers on a two-dimensional plane; p represents a two-dimensional coordinate point in the two-dimensional plane; Representative point p to the center point distance; Representative point p to the center point distance;

[0088] The specific calculation formula of the Lloyd iterative algorithm is as follows:

[0089] ;

[0090] Where, Representative The center point of the region; Represents the first segmentation by the Voronoi diagram algorithm regions; Representative A two-dimensional coordinate point in a region.

[0091] Specifically, the initial radiation intensity distribution of the light source of the backlight module is set, and then the light emitted by the light source enters the light guide plate, and the multiple scattering and absorption behavior of the light in the light guide plate is calculated by the radiation transfer equation to obtain the light diffusion process, and the prism of the light guide plate and the microstructures of the diffusion film are used to perform additional direction adjustment and intensity redistribution on the light, and model it through the transmission function, and then the light uniformization process of the diffusion film is modeled through the diffusion equation, and the various light diffusion processes are integrated to construct a multi-level light diffusion coupling model, based on the multi-level light diffusion coupling model, by randomly sampling the light emission angle and position, it is discretized to simulate the propagation of light in the backlight module, and the discretization result is used as the initial ray tracing, according to the light transmission equation, the model The propagation process of light reflection, refraction and scattering in the backlight module is simulated to obtain the light diffusion effect between different partitions. According to the ray tracing results of each partition of the backlight light source, the light diffusion ratio between different partitions is obtained, and the corresponding light diffusion transfer matrix is established. The final brightness of the corresponding backlight partition is calculated through the light diffusion transfer matrix and the partition driving current. According to the set target brightness of each backlight partition, the target brightness equation of each partition is constructed, and the target brightness equation is repeatedly solved by least squares to calculate the optimal compensation current until the difference between the partition brightness and the target brightness converges to the preset range. The solution is stopped, and then the calculated compensation current is clipped according to the minimum and maximum values of the driving current to generate the final compensation current to adjust the brightness of each backlight partition.

[0092] It should be further explained that the specific calculation formula of the radiation transfer equation is as follows:

[0093] ;

[0094] Where, Represents the light at the position At, along the direction The light intensity, where Represents the angle between the light direction and the normal direction, Represents the angle between the light on the horizontal plane and the preset reference direction; ds represents the infinitesimal element of the light propagation path; Represents the absorption loss of light in the light guide plate; Represents the degree of light scattering in the light guide plate; Represents the probability distribution of the angle change of light when it is scattered; Represents the unit direction space;

[0095] The specific calculation formula of the transmission function is as follows:

[0096] ;

[0097] Where, Represents the position of the microstructure layer Light intensity on Represents the position of the microstructure layer Transmittance distribution on ; Represents the position of the light guide plate output to the microstructure layer Light intensity on

[0098] The specific calculation formula of the diffusion equation is as follows:

[0099] ;

[0100] Where, Represents the diffusion membrane position The light intensity on the film; t represents the diffusion time; D represents the diffusion rate of light in the film; The Laplacian operator representing light intensity measures local brightness changes; Represents the attenuation coefficient of light;

[0101] The specific calculation formula of the multi-level light diffusion coupling model described in S4.2 is as follows:

[0102] ;

[0103] Where, Represents the final output light intensity distribution; represents the diffusion calculation of the diffusion membrane; Transmission calculation representing the microstructure layer; Represents the calculation of radiation transfer in the light guide; Represents the initial light intensity distribution.

[0104] In addition, when performing multiple scattering and absorption behaviors, the thickness of the backlight unit is in the range of [1.5mm, 5mm] and the horizontal size of the backlight unit is in the range of [50mm 2 , 500mm 2 ] range, the wavelength of light is in the range of [380nm, 780nm], and the initial emission angle of light is in the range of [0, 90 ∘ ] range.

[0105] Example 2, reference Figure 1 A backlight source zone dimming control method is provided, wherein the specific steps of the control method are as follows:

[0106] Analyze the human eye's sensitivity to different brightness and contrast levels, adjust local brightness weights, and optimize the brightness mapping relationship of backlight partitions.

[0107] Specifically, according to the three groups of channel values of red R, green G and blue B of each image data, the input image is converted into a grayscale image, and the grayscale image is converted from the spatial domain to the frequency domain through a two-dimensional fast Fourier transform. The converted image data is centered to move the zero-frequency component to the center of the image, and the spatial frequency of the image is decomposed into low-frequency, medium-frequency and high-frequency components through a low-pass filter, a band-pass filter and a high-pass filter respectively. The extracted different frequency components are then converted back to the spatial domain through an inverse Fourier transform, and a corresponding reconstructed image is generated. The reconstructed image is then normalized to obtain the perceived brightness value of each pixel. The Weber contrast model calculates the local contrast of each image. Based on the image width, height, and perceived brightness value of each pixel, the horizontal and vertical spatial frequencies of each image are calculated according to the pixel size of each feature information in the image, the distance from the observer to the screen, and the viewing angle. The contrast sensitivity of the human eye to different spatial frequencies is calculated using the CSF function. The contrast sensitivity threshold of the human eye for each frame of image at the current frequency is obtained through the contrast sensitivity, and a curve of the human eye's visual perception characteristics is plotted. The brightness weight of the corresponding pixel is then calculated based on the contrast, spatial frequency, and CSF function value. Based on the calculated brightness weight, the local brightness of the image is adjusted.

[0108] In this embodiment, the specific calculation formula of the CSF function is as follows:

[0109] ;

[0110] Where, represents the contrast sensitivity at spatial frequency f; f represents the spatial frequency; Represents the average brightness of the image background; as well as represent the empirical fitting parameters.

[0111] Assign the best light source type and driving mode according to the scene characteristics, monitor the backlight module temperature in real time, and dynamically adjust the driving current based on the monitoring data.

[0112] Specifically, the currently displayed image information is analyzed, and the scene features of each group of brightness distribution, contrast, dynamic range and color information that affect backlight control are extracted. Then, the lighting requirements of different areas are calculated based on the brightness histogram and local contrast. According to the scene characteristics, the brightness dynamic range is calculated, and the corresponding light source type is selected according to the brightness and dynamic range of the currently displayed image, including Mini-LED, OLED and Micro-LED. Then, the power consumption and image quality of each backlight partition are optimized through local dimming or global dimming driving mode. According to the power consumption of each light source under different driving modes, different light sources are used for combined driving, and sensors are used to monitor the screen output brightness. The energy efficiency of light source selection and partition driving strategy is evaluated, and the light source output is readjusted based on the energy efficiency evaluation results.

[0113] According to the optimized brightness zoning and dimming control strategy, independent dimming control strategies are adopted for different backlight areas.

[0114] Continuously monitor the output image quality parameters, re-predict the backlight zone dimming strategy based on the scene, and form an adaptive closed-loop control.

Claims

1. A backlight source zone dimming control method, characterized in that: The specific steps of this control method are as follows: Ⅰ. Analyze the input image data features, identify the highlight motion area, and predict and optimize the current backlight partition dimming strategy; II. Adjust the backlight area in real time based on the brightness, contrast, and color changes of the input content, and build a multi-level light diffusion coupling model after the partition adjustment; III. Analyze the human eye's sensitivity to different brightness and contrast levels, adjust local brightness weights, and optimize the brightness mapping relationship of backlight partitions; IV. Allocate the best light source type and driving mode based on scene characteristics, monitor the backlight module temperature in real time, and dynamically adjust the driving current based on the monitoring data; V. Based on the optimized brightness zoning and dimming control strategy, independent dimming control strategies are adopted for different backlight areas; VI. Continuously monitor the output image quality parameters, re-predict the backlight zone dimming strategy based on the scene, and form an adaptive closed-loop control; The specific steps for real-time adjustment of the backlight area in step II are as follows: S3.1: After adjusting the brightness of each partition using the optimized backlight partition dimming strategy, calculate the local brightness requirement of the current frame image through weighted averaging. Then, divide the entire image into multiple grid areas and calculate the average brightness of each grid area. S3.2: Set a brightness threshold. If the average brightness of a grid is higher than the brightness threshold, use dense honeycomb partitioning to subdivide the area. If the average brightness of a grid is lower than the brightness threshold, merge adjacent areas and use sparse rectangular partitioning. After traversing all grids, obtain the adjusted number of backlight partitions. S3.3: After the dynamic partition adjustment is completed, the center points of each partition are extracted, and each center point represents a backlight adjustment area. The entire screen area is divided into multiple polygonal partitions using the Voronoi diagram algorithm; S3.4: Calculate the centroid of each partition using the Lloyd iteration algorithm, and use the new centroid as a new Voronoi generating point. Repeat the Voronoi calculation and Lloyd iteration until the position change of the Voronoi generating point converges to a preset range, and output the final irregular backlight partition. The specific steps for constructing the multi-level light diffusion coupling model described in step II are as follows: S4.1: Set the initial radiation intensity distribution of the backlight module's light source. Then, light from the light source enters the light guide plate. Calculate the multiple scattering and absorption behavior of light in the light guide plate using the radiation transfer equation to obtain the light diffusion process. S4.2: Utilize the prisms of the light guide plate and the microstructures of the diffuser film to perform additional directional adjustments and intensity redistribution of light, modeled using a transmission function. Then, model the light homogenization process of the diffuser film using a diffusion equation. By integrating the various light diffusion processes, a multi-level light diffusion coupling model is constructed. S4.3: Based on a multi-level light diffusion coupling model, randomly sample the light emission angles and positions and discretize them to simulate light propagation in the backlight module. The discretized results are used as the initial ray tracing. S4.4: Based on the light propagation equation, simulate the propagation process of light reflection, refraction, and scattering in the backlight module to obtain the light diffusion effect between different partitions. Based on the ray tracing results of each partition of the backlight light source, obtain the light diffusion ratio between different partitions and establish the corresponding light diffusion transfer matrix. S4.5: Calculate the final brightness of the corresponding backlight partition using the light diffusion transfer matrix and the partition drive current. Based on the set target brightness of each backlight partition, construct a target brightness equation for each partition. Repeat the least squares solution to the target brightness equation and calculate the optimal compensation current until the difference between the partition brightness and the target brightness converges to within a preset range. The solution is then stopped. The calculated compensation current is then clipped based on the minimum and maximum drive current values to generate a final compensation current for adjusting the brightness of each backlight partition. The specific steps for adjusting the local brightness weight in step III are as follows: S5.1: Convert the input image to a grayscale image based on the red (R), green (G), and blue (B) channel values of each image data. Convert the grayscale image from the spatial domain to the frequency domain using a two-dimensional fast Fourier transform. Center the converted image data to move the zero-frequency component to the center of the image. S5.2: Decompose the spatial frequency of the image into low-frequency, medium-frequency, and high-frequency components using a low-pass filter, a band-pass filter, and a high-pass filter, respectively. Then, perform an inverse Fourier transform to convert the extracted frequency components back to the spatial domain and generate a corresponding reconstructed image. The reconstructed image is then normalized to obtain the perceived brightness value of each pixel. S5.3: Calculate the local contrast of each image using the Weber contrast model. Calculate the horizontal and vertical spatial frequencies of each image based on the image width, height, and perceived brightness of each pixel, the pixel size of each feature in the image, the distance from the viewer to the screen, and the viewing angle. S5.4: Calculate the contrast sensitivity of the human eye to different spatial frequencies using the CSF function. Use the contrast sensitivity to obtain the contrast sensitivity threshold for each frame of the image at the current frequency. Plot a curve for the human eye's visual perception characteristics. Calculate the brightness weight of the corresponding pixel based on the contrast, spatial frequency, and CSF function value. Adjust the local brightness of the image based on the calculated brightness weight.

2. The backlight source zone dimming control method according to claim 1, characterized in that: The specific steps for identifying the highlight motion area described in step I are as follows: S1.1: Perform brightness normalization on each input image frame, randomly select a set of gamma correction coefficients from the range [0.8, 2.2], adjust the image contrast based on the selected gamma correction coefficients, and then divide the corrected image data into multiple non-overlapping sub-regions; S1.2: Calculate the grayscale histogram of each subregion to obtain the grayscale distribution of each subregion, calculate the cumulative distribution value (CDF) of the histogram of each subregion, then clip the CDF value based on a preset contrast limit threshold, and calculate the pixel value of the corresponding subregion after histogram equalization based on each clipped CDF value. Then, smooth the boundaries of each equalized subregion through bilinear interpolation and reconstruct the complete image; S1.3: Use a white balance algorithm based on the grayscale world assumption to maintain color consistency between image data frames. Then, use non-local mean filtering to remove noise from the image data. Calculate the optical flow field between the processed image data frames to obtain pixel motion information. Optical flow field information that exceeds the preset highlight brightness threshold and motion amplitude threshold is identified as a highlight motion area and marked.

3. The backlight source zone dimming control method according to claim 2, characterized in that: The specific steps for predicting and optimizing the current backlight zone dimming strategy described in step I are as follows: S2.1: Initialize the policy adjustment model based on the LSTM model architecture, extract the temporal features of each preprocessed image frame data, and associate them with the corresponding highlight motion area. Then, divide the generated temporal features into training and test sets, and then input the training set into the policy adjustment model. S2.2: The input layer of the strategy adjustment model receives the training set data and divides the training set data into multiple batches of training groups. The training groups are then forward-propagated to the LTSM layer in sequence. The final result output by the output gate is then nonlinearly processed by the Softmax function of the fully connected layer to obtain the final predicted scene recognition result and the highlight motion area within the preset time interval, which are then output through the output layer. S2.3: Calculate the true scene and highlight motion region labels of the current frame and the predicted scene recognition results and highlight motion regions using the cross-entropy loss function. Input the calculated loss value into the output layer of the policy adjustment model, perform backpropagation based on the chain rule, and adjust the parameters of each network layer using the Adam optimizer. S2.4: After each round of training, the performance indicators of the policy adjustment model are evaluated using the test set data and a comprehensive score is generated. If the score does not reach the preset score threshold, the policy adjustment model is retrained and tested until the model loss value converges to the preset range, at which point training is stopped. S2.5: Use an external memory module to record historical task information of the strategy adjustment model. Then, use the strategy adjustment model to perform time series modeling on the image data of each past frame. Use a forward propagation algorithm to generate highlight motion area prediction results and scene recognition results within a preset time period. Based on the identified scene category and predicted highlight area, calculate the brightness gain of each partition and adjust and optimize the original backlight partition dimming strategy. S2.6: When the loss value between the output result of the policy adjustment model and the actual result exceeds the preset threshold, the cosine similarity between the historical task information stored in the external memory module and the current image data is calculated to obtain the enhanced hidden state. Based on the selected enhanced hidden state and the current loss value, the loss value is recalculated, and the gradient of the current task is calculated based on the latest loss value to re-update the policy adjustment model parameters and update the historical task information recorded in the memory module.

4. The backlight source zone dimming control method according to claim 1, characterized in that: The specific steps for allocating the optimal light source type and driving mode according to scene characteristics in step IV are as follows: S6.1: Analyze the currently displayed image information and extract scene features that affect backlight control, including brightness distribution, contrast, dynamic range, and color information. Then, calculate the lighting requirements for different areas based on the brightness histogram and local contrast. S6.2: Calculates the brightness dynamic range based on scene characteristics and selects the corresponding light source type, including Mini-LED, OLED, and Micro-LED, based on the brightness and dynamic range of the currently displayed image. It then optimizes the power consumption and image quality of each backlight zone through local dimming or global dimming drive modes. S6.3: Based on the power consumption of each light source under different driving modes, use different light sources for combined driving, use sensors to monitor the screen output brightness, and evaluate the energy efficiency of light source selection and partition driving strategy, and readjust the light source output based on the energy efficiency evaluation results.

Images