LED backlight integrated control system of liquid crystal display screen

Abstract

The invention relates to the technical field of liquid crystal display, and discloses an LED backlight integrated control system of a liquid crystal display screen, which comprises a data processing unit, a prediction analysis unit, a weight calculation unit, a backlight optimization unit and a drive control unit, the LED backlight integrated control method of the liquid crystal display screen comprises the following steps: receiving a video frame and extracting data in a partitioned manner; predicting future visual characteristics in combination with historical and current data; then calculating an attention weight; then target backlight is determined by integrating the weight, the prediction and the optimization target; and finally generating a driving signal. LED backlight intelligent control is achieved through time sequence prediction and a space-time attention model in combination with global optimization. Compared with the prior art, dynamic response lag is remarkably improved, picture details and focuses are improved, multiple display indexes are balanced, visual flaws are reduced, and the picture quality performance and watching comfort of the liquid crystal screen to various contents are enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of liquid crystal display, and specifically to an LED backlight integrated control system for a liquid crystal display screen. Background Art

[0002] A liquid crystal display screen (LCD) does not emit light by itself and relies on a backlight source to display images. The traditional global backlight method has uniform brightness, which limits the contrast and dark field performance, and has relatively high energy consumption. To improve this situation, using light-emitting diodes (LEDs) as the backlight source and combining with local dimming technology has become the mainstream. This technology divides the LED backlight into multiple independently controlled areas and dynamically adjusts the brightness of each area according to the picture content, thereby improving the contrast and reducing the power consumption.

[0003] However, the existing LED local dimming control still has deficiencies. Many control strategies only adjust based on the information of the current frame. When dealing with rapid scene changes or high-speed moving objects in a video stream, it often leads to a lag in backlight response, affecting the smoothness and naturalness of the dynamic picture quality.

[0004] In addition, how to achieve the best balance among multiple objectives such as improving the picture contrast, accurately reproducing image details, avoiding visual defects such as halos or block effects, ensuring smooth transition of brightness in the time dimension, and effectively controlling energy consumption is still a technical challenge. When analyzing the image content, the existing methods may not fully distinguish the visual importance of different areas, or lack effective prediction of short-term future changes in the picture, making the fineness and intelligent level of backlight adjustment to be improved.

[0005] Therefore, the present invention proposes an LED backlight integrated control system for a liquid crystal display screen to solve the deficiencies of the existing technology. Summary of the Invention

[0006] In view of the deficiencies of the existing technology, the present invention provides an LED backlight integrated control system for a liquid crystal display screen, which solves the problem that it is difficult to achieve the optimal balance among improving the picture quality, ensuring visual comfort, and reducing energy consumption in the LED backlight control technology of liquid crystal display screens due to the lack of in-depth analysis and prediction of the spatio-temporal dynamic characteristics of video content.

[0007] To achieve the above objectives, the present invention is realized through the following technical solutions: An LED backlight integrated control system for a liquid crystal display screen, the system includes the following modules:

[0008] A data processing unit, configured to receive the current video frame through a video input interface, and perform image partitioning processing on the current video frame to obtain the image region data of each of the multiple image regions;

[0009] A prediction analysis unit, configured to apply a temporal prediction model to predict the predictive visual characteristics of each of the multiple image regions in the short-term future by combining historical video frame information with the image region data of each of the multiple image regions;

[0010] A weight calculation unit, configured to apply a multi-scale spatio-temporal attention model to calculate the attention weight corresponding to each of the multiple image regions by combining the image region data of each of the multiple image regions with the predictive visual characteristics;

[0011] A backlight optimization unit, configured to execute an optimization algorithm based on the attention weight, the predictive visual characteristics, and a preset global optimization objective function to determine the target backlight brightness of the multiple backlight partitions corresponding to the multiple image regions;

[0012] A drive control unit, configured to generate and output a drive signal for controlling the actual brightness of the multiple backlight partitions based on the target backlight brightness.

[0013] Preferably, the data processing unit is specifically configured to:

[0014] Perform color space conversion processing on the received current video frame;

[0015] Divide the current video frame after color space conversion processing into the multiple image regions according to a preset backlight physical partition layout;

[0016] And extract the corresponding image data of each image region as the image region data.

[0017] Preferably, the prediction analysis unit is specifically configured to:

[0018] For each of the image regions, extract the brightness feature and motion feature of the image region from the historical video frame information and the image region data of the image region;

[0019] Based on the extracted brightness feature and motion feature, update the temporal state corresponding to the image region through the temporal prediction model;

[0020] And based on the updated temporal state, predict the predictive brightness value and predictive brightness change trend of the image region in the short-term future as the predictive visual characteristics.

[0021] Preferably, when the prediction analysis unit is used to update the temporal state, the image region data of each image region includes brightness-related features and motion-related features extracted from the image region; the predictive visual characteristics predicted through the temporal state include the predictive brightness value and predictive brightness change trend of the image region.

[0022] Preferably, the weight calculation unit is specifically configured to:

[0023] For each of the image regions, extract multi-scale spatial features of the image region based on its image region data;

[0024] And combine the multi-scale spatial features with the predictive visual characteristics of the image region, apply the multi-scale spatio-temporal attention model, comprehensively evaluate the visual importance of the image region in the spatial dimension and the temporal dimension, and calculate the corresponding attention weight of the image region.

[0025] Preferably, when applying the multi-scale spatio-temporal attention model, for each of the image regions, the weight calculation unit:

[0026] Extract the spatial saliency features and spatial detail features of the image region at multiple spatial scales based on its image region data;

[0027] Calculate the temporal attention information in combination with the predictive visual characteristics of the image region;

[0028] And fuse the extracted spatial saliency features, the spatial detail features, and the calculated temporal attention information to generate the corresponding attention weight of the image region.

[0029] Preferably, the global optimization objective function adopted by the backlight optimization unit comprehensively evaluates the following aspects in its construction:

[0030] The accuracy of image content reproduction, which is measured based on the image region data of each image region and its corresponding attention weight;

[0031] The smoothness of the brightness transition between adjacent backlight partitions, which is measured based on the brightness relationship between adjacent image regions, their corresponding attention weights, and the image edge information between the image regions;

[0032] The continuity of the backlight brightness in the time series, which is measured based on the backlight brightness relationship between the current frame and the previous frame and the predicted brightness change trend included in the predictive visual characteristics;

[0033] The energy consumption of the overall backlight system, which is measured based on the relationship between the target backlight brightness of each backlight partition and its preset power consumption model, the attention weight, and the overall visual characteristics of the current scene.

[0034] Preferably, in the global optimization objective function:

[0035] The attention weights are used to dynamically adjust the relative importance of each image region when evaluating the accuracy of the reproduction of the image content, and in combination with the image edge information between the image regions, dynamically adjust the constraint strength imposed when evaluating the smoothness of the brightness transition between adjacent backlight partitions;

[0036] The attention weights, in combination with the overall visual characteristics of the current scene, are used to dynamically adjust the weighting ratio when evaluating the power consumption of the overall backlight system;

[0037] The predictive brightness change trend included in the predictive visual characteristics is used to dynamically guide the desired change target when evaluating the continuity of the backlight brightness over time.

[0038] Preferably, the drive control unit is specifically configured to convert the target backlight brightness of each of the multiple backlight partitions into pulse width modulation parameters or analog drive current parameters for respectively controlling the light emitting intensity of the corresponding backlight partition, and output the converted pulse width modulation parameters or the analog drive current parameters as the drive signal.

[0039] The present invention also provides an integrated control method for the LED backlight of a liquid crystal display screen, the method comprising the following steps:

[0040] S1. Receive the current video frame through the video input interface, and perform image partitioning processing on the current video frame to obtain the image region data of each of the multiple image regions;

[0041] S2. Combine the historical video frame information with the image region data of each of the multiple image regions, and apply a time series prediction model to predict the predictive visual characteristics of each of the multiple image regions in the short term in the future;

[0042] S3. Combine the image region data of each of the multiple image regions with the predictive visual characteristics, and apply a multi-scale spatio-temporal attention model to calculate the attention weights corresponding to each of the multiple image regions;

[0043] S4. Execute an optimization algorithm based on the attention weights, the predictive visual characteristics, and a preset global optimization objective function to determine the target backlight brightness of the multiple backlight partitions corresponding to the multiple image regions;

[0044] S5: Generate and output a drive signal for controlling the actual brightness of the multiple backlight partitions based on the target backlight brightness.

[0045] The present invention provides an integrated control system for the LED backlight of a liquid crystal display screen. It has the following beneficial effects:

[0046] 1. The present invention adopts a technical solution that uses a time series prediction model to pre-judge the future visual changes of the image area, which enables the LED backlight system to respond to the dynamics of the picture earlier. Compared with the prior art solution that only relies on the current frame information to adjust the backlight, the present invention effectively improves the lag of backlight adjustment, especially when the scene switches quickly or the object moves at high speed, the picture brightness transition is more natural and smooth, and the dynamic display quality of the LCD screen is improved.

[0047] 2. The present invention introduces a multi-scale spatiotemporal attention model to calculate the visual importance weight of each image area. It can finely identify the key content and details in the picture and weight them in combination with their future dynamic trends. Compared with the methods in the prior art that may use global dimming or simple zone dimming, treat all areas equally or roughly divide the importance, the present invention solves the problem that it cannot effectively highlight the visual focus and may cause the loss of important details, making the picture clear in primary and secondary, and the viewing experience better.

[0048] 3. The present invention determines the final backlight brightness by constructing a comprehensive global optimization objective function. This function cleverly balances the accuracy of image reproduction, the smoothness of backlight space transition, the continuity of time series, and the overall energy consumption. This is different from the prior art solutions that often focus on a single goal and may sacrifice other aspects of performance. The present invention solves the problem of difficulty in taking into account multiple display indicators and easily leading to loss of one while focusing on another, and achieves a synergistic improvement in the display effect and energy efficiency of the LCD screen.

[0049] 4. When optimizing the backlight, the present invention pays special attention to the smooth transition of brightness between adjacent backlight partitions, and dynamically adjusts the constraint strength in combination with the attention weight and image edge information. At the same time, it also guides the continuous change of the backlight in time based on the predicted brightness change trend. Compared with some backlight control strategies in the prior art that are prone to produce halos at the junction of light and dark or brightness jumps and flickers in dynamic scenes, it significantly reduces visual defects and improves the viewing comfort of the LCD screen.

[0050] 5. The LED backlight integrated control system of the present invention can intelligently adjust the brightness of each LED backlight partition according to the real-time characteristics and future trends of the video content through a complete process of data processing, prediction analysis, weight calculation, backlight optimization and drive control. Compared with the existing technology that may use fixed control logic or algorithms with poor adaptability, the present invention solves the limitation that it is difficult to continuously provide the best display effect when facing diversified video content, so that the LCD screen can show excellent adaptability and image quality performance for different types of pictures. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 is a system architecture diagram of the present invention;

[0052] Figure 2 This is the flowchart of the method of the present invention. Specific embodiments

[0053] Next, in conjunction with the accompanying drawings of the present invention specification, the technical solutions in the embodiments of the present invention will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts fall within the protection scope of the present invention.

[0054] Please refer to Figure 1 , the embodiment of the present invention provides an LED backlight integrated control system for a liquid crystal display screen. The system includes the following modules:

[0055] A data processing unit, configured to receive a current video frame through a video input interface, and perform image partitioning processing on the current video frame to obtain image region data of each of a plurality of image regions;

[0056] In this embodiment, the data processing unit undertakes the task of initially processing the input video signal and extracting basic analysis data.

[0057] The data processing unit receives the current video frame via the video input interface. The video input interface can be a High-Definition-Multimedia-Interface (HDMI), a Display-Port, or an internal system bus, etc., for receiving digital video signals from an external video playback device, a graphics processing unit, or other image sources. The current video frame is an independent picture constituting a dynamic video sequence, carrying the image information at that moment.

[0058] After receiving the current video frame, the data processing unit performs color space conversion processing on the current video frame. This processing aims to convert the video frame data from its original color space to a color space more suitable for subsequent brightness analysis and backlight control. Specifically, a commonly used conversion method is to convert the video frame data in the input RGB (Red, Green, Blue) color space to the YCbCr color space. In this YCbCr color space, the Y component represents the brightness information of the image, while the Cb and Cr components represent the chrominance information of the image. Since the adjustment of the LED backlight system directly affects the brightness perception of the display screen, separating the brightness information (Y component) enables subsequent analysis and optimization to be more focused and efficient.

[0059] For example, if the red, green, and blue component values of the input pixel are R, G, and B respectively, the corresponding brightness component Y can be calculated through the following linear transformation formula:

[0060] Y = w R ·R + w G ·G + w B ·B + O offset ;

[0061] Wherein, R, G, and B respectively represent the normalized or non - normalized pixel values of the red, green, and blue channels; w R , w G , w B are preset weight coefficients, and these coefficients vary according to the adopted color standard (such as ITU - R BT.601, ITU - R - BT.709, or ITU - R - BT.2020, etc.). For example, under the ITU - R - BT.709 standard, for R, G, B values normalized to the [0, 1] interval, the commonly used coefficients are w R = 0.2126, w G = 0.7152, w B = 0.0722; O offset is an optional offset, which is used to adjust the black level in some standards (such as when processing digital video components). For example, for 8 - bit video data, the offset may be 16. After the color space conversion, the obtained luminance component map will be used as the main basis for subsequent image partitioning and feature extraction.

[0062] The data processing unit divides the current video frame (especially its luminance component map) that has undergone color space conversion processing into multiple image regions according to the preset backlight physical partition layout. The purpose of this image partitioning process is to spatially divide the continuous image into several region blocks corresponding to the independently controllable LED physical partitions in the backlight module of the liquid crystal display screen. The backlight physical partition layout information describes the number, geometric shape, size, and spatial arrangement of the backlight LED units or unit groups on the screen, and this information is usually predefined by the hardware specifications of the display device and stored in the system. For example, the backlight system may be divided into a regular rectangular grid array of M×N, where M and N respectively represent the number of backlight partitions along the vertical and horizontal directions of the screen;

[0063] Preferably, an irregular partitioning method may also be adopted to better adapt to a specific screen shape or perform differential control according to the importance of different display regions. The partitioning operation ensures that each generated image region is closely associated with one or a group of specific physical backlight partitions in terms of spatial position.

[0064] After completing the image partitioning, the data processing unit extracts the corresponding image data for each of the partitioned image regions, and these data sets constitute the image region data of the respective multiple image regions. Extracting the image region data is to quantitatively describe the image content characteristics within each local region, thereby providing the necessary input information for subsequent prediction analysis, attention calculation, and backlight optimization. Specifically, for the k-th image region A k , the extracted image region data may include, but is not limited to, one or more of the following features:

[0065] Average brightness value Calculate the arithmetic mean of the brightness values of all pixels within the image region A k . If the region A k contains N k pixels, and the brightness value of each pixel p is L(p), then:

[0066] Brightness histogram H k (l): Statistically count the pixel frequency or frequency of each brightness level l within the image region A k .

[0067] Brightness standard deviation σ L,k : Measure the degree of dispersion of the brightness values within the image region A k , reflecting the contrast or complexity of the regional content:

[0068] Other statistical features, such as the maximum brightness value, minimum brightness value, brightness median, etc.

[0069] Texture features: For example, parameters such as energy, contrast, correlation, and entropy calculated based on the gray-level co-occurrence matrix (GLCM) can be used to describe the texture roughness or uniformity of the region.

[0070] Edge information: For example, after applying edge detection operators (such as Sobel, Prewitt, Canny, etc.), count the number of edge pixels, average edge intensity, or edge direction histogram within the region.

[0071] The extracted image region data comprehensively characterizes the static content characteristics of each local image block and is the basis for the subsequent refined analysis and decision-making of the entire backlight integration control system.

[0072] A prediction analysis unit for combining historical video frame information with the image region data of the respective multiple image regions and applying a time series prediction model to predict the predictive visual characteristics of the respective multiple image regions in the short term in the future;

[0073] In this embodiment, the prediction analysis unit aims to endow the system with adjustment capabilities through in-depth insights into the temporal dynamics of video content.

[0074] The core task of the prediction analysis unit is to combine historical video frame information with the respective image region data of the multiple image regions currently obtained from the data processing unit, apply a temporal prediction model, and predict the visual characteristics of each of the multiple image regions in the short-term future. These predicted visual characteristics, called predictive visual characteristics, provide key time-dimensional information for subsequent attention weight calculation and backlight optimization.

[0075] Specifically, for each independent image region, the prediction analysis unit performs the following operations:

[0076] The prediction analysis unit extracts the brightness feature and motion feature of the image region from the historical video frame information and the image region data of the current image region. The historical video frame information may include the image region data of the corresponding image region in the past several frames (for example, the first N h frames, where N h is a preset integer), such as its historical average brightness, historical brightness change situation, etc. The image region data of the current image region, as mentioned above, already contains rich brightness-related features (such as the average brightness of the current region at time t, the brightness histogram H k (t), etc.).

[0077] The extraction of motion features is crucial for understanding scene dynamics. In one possible implementation, the motion feature can be estimated by comparing the content of the current frame t with that of one or more adjacent historical frames (such as t-1, t-2,...) of the corresponding image region. For example, the inter-frame difference (Frame-Difference, FD) measure between the current frame t and the previous frame t-1 of the image region A k can be calculated, such as the mean absolute difference:

[0078]

[0079] where L(p,t) represents the brightness value of pixel p in the kth image region A k at time t, and N k is the number of pixels in region A k . The magnitude of FD k (t) can characterize the intensity of motion of the region.

[0080] As another option, the block matching algorithm (Block-Matching-Algorithm, BMA) can be used for the image region A kEstimate its motion vector MV k (t) = (mv x , mv y ) k (t). This motion vector describes the main displacement direction and amplitude of region A k between frames. Motion features may include the magnitude ||MV k (t)|| and the direction θ k (t). These luminance features and motion features together form the basis for analyzing the temporal behavior of this image region.

[0081] Based on the extracted luminance features (such as time series ) and the motion features (such as time series {FD k (t), FD k (t - 1),...} or {MV k (t), MV k (t - 1),...}), the prediction analysis unit updates the temporal state S k (t) corresponding to this image region through the temporal prediction model. The temporal prediction model is a model that can learn and represent the inherent laws and dynamics of time series data.

[0082] In some embodiments, the temporal prediction model can be a classical statistical model, such as the Kalman Filter. The Kalman Filter can recursively estimate the state of the system based on the noisy observation sequence (i.e., the extracted luminance features and motion features) and make a one-step prediction. The temporal state S k (t) here can be the state vector in the Kalman Filter, which contains the estimates of potential variables such as the current luminance of the region, the luminance change rate, etc.

[0083] In some other embodiments, more complex machine learning models can be adopted, such as the Auto-regressive-Integrated-Moving-Average (ARIMA), or models based on the Recurrent-Neural-Network (RNN) and its variants such as the Long-Short-Term-Memory (LSTM) or the Gated-Recurrent-Unit (GRU). Deep learning models are particularly good at capturing non-linear and long-term dependencies. For the RNN / LSTM / GRU models, the temporal state S k (t) usually corresponds to the state vector of the hidden layer of the network after processing the input at time t.

[0084] According to the updated timing state S k (t), the prediction analysis unit further predicts that the image area will be within the short term in the future (for example, the future frame t+1 to frame t+Δt pred frame, where Δt pred The predicted brightness value and predicted brightness change trend of the image region are a preset prediction time window length, usually a small number of frames. These together constitute the predicted visual characteristics for the image region.

[0085] The predicted brightness value (where 1≤τ≤Δt pred ) is a direct numerical estimate of the average brightness of the image area at the future time t+τ.

[0086] The predicted brightness change trend can be a qualitative description of the future brightness change direction (e.g., significant increase, slight increase, stable, slight decrease, significant decrease), or a quantitative estimate of the brightness change rate or acceleration (e.g., the predicted first-order derivative of brightness). or the second derivative ).

[0087] For example, if the time series prediction model is an ARIMA model, the prediction function of the model can be used directly to output the brightness value of multiple steps in the future. If it is an RNN / LSTM model, the current hidden state S k (t) and possible future inputs (either zero input, or autoregressive input based on previous outputs) are fed into the network to generate a sequence of predictions for future brightness.

[0088] By predicting the future direction of the brightness and dynamics of the image area, the system can adjust the backlight earlier and more smoothly to avoid backlight response lag or overshoot caused by sudden scene changes, thereby improving visual comfort and dynamic image display quality. For example, if it is predicted that the brightness of a certain area will increase significantly, the backlight system can start to increase the brightness of the corresponding area in advance.

[0089] It should be clear that the image area data of each image area does include the brightness-related features and motion-related features extracted from the image area when the prediction analysis unit is used to update the timing state. In addition, the predictive visual characteristics obtained by predicting the timing state do include the predicted brightness value and predicted brightness change trend of the image area.

[0090] A weight calculation unit, configured to combine the image region data of each of the plurality of image regions with the predictive visual characteristics and apply a multi-scale spatiotemporal attention model to calculate the attention weights corresponding to each of the plurality of image regions;

[0091] In this embodiment, the weight calculation unit is responsible for evaluating the visual importance of each image region in the video frame, so as to provide differential guidance for subsequent backlight optimization:

[0092] The core function of the weight calculation unit is to combine the respective image region data of the multiple image regions provided by the data processing unit and the predictive visual characteristics provided by the predictive analysis unit, and apply a preset multi-scale spatio-temporal attention model to calculate the attention weights corresponding to the multiple image regions. This attention weight is intended to quantify the importance of each image region to human visual perception.

[0093] Specifically, for each image region A k At the current moment t, the weight calculation unit performs the following processing:

[0094] The weight calculation unit first extracts the spatial saliency features and spatial detail features of the image region at multiple spatial scales based on the image region data of this image region A k of. The purpose of adopting multi-scale analysis is to be able to capture the structural information and content characteristics of the image from different granularities, because the human eye has different sensitivities to features at different scales.

[0095] Spatial saliency features At scale s, it is intended to quantify the visual prominence of image region A k relative to its neighborhood or the entire scene. Generally, the human visual system will give priority to areas that are significantly different from the surrounding environment in terms of color, brightness, direction, or texture. In one possible implementation, the spatial saliency features can be calculated by known visual saliency detection algorithms, such as methods based on local contrast analysis, methods based on frequency domain analysis, or methods based on graph theory. For example, a saliency value can be calculated for each image region A k at a specific scale s, and this value reflects the possibility of this region becoming a visual focus.

[0096] Spatial detail features At scale s, it is used to characterize the degree of detail richness and structural complexity inside image region A k . These features help to identify regions containing fine textures, clear edges, or other high-frequency information, which usually require higher display fidelity. Specifically, the spatial detail features can include:

[0097] Edge intensity or density: By applying an edge detection operator (such as Sobel, Canny, etc.) to image region A k at scale s, the number of edge pixels or the average value of edge amplitudes is statistically calculated.

[0098] Texture descriptor: For example, calculate the gray-level co-occurrence matrix (GLCM) of region A at scale s and extract texture parameters such as energy, contrast, and entropy from it. k Local variance or standard deviation: Calculate the variance or standard deviation of the luminance values of region A at scale s. High variance usually means richer details. To achieve multi-scale extraction, the current video frame (or its luminance component) can be constructed into an image pyramid (such as a Gaussian pyramid or a Laplacian pyramid). Subsequently, for the corresponding regions of the original image region A

[0099] at each layer of the pyramid (i.e., each scale s), extract the above-mentioned spatial saliency features k and spatial detail features k in parallel. The weight calculation unit combines the predictive visual characteristics of the image region A

[0100] provided by the prediction analysis unit (such as the predicted future luminance value k and the predicted luminance change trend to calculate its temporal attention information I (t). The temporal attention information aims to capture the importance of those regions that are expected to change significantly or exhibit specific dynamic behaviors in the time dimension. Generally, fast-moving objects, scene regions where significant luminance changes are about to occur, or regions with high prediction uncertainty are more likely to attract the user's attention. T,k (t). The temporal attention information aims to capture the importance of those regions that are expected to change significantly or exhibit specific dynamic behaviors in the time dimension. Generally, fast-moving objects, scene regions where significant luminance changes are about to occur, or regions with high prediction uncertainty are more likely to attract the user's attention.

[0101] In a possible implementation, the temporal attention information I T,k (t) can be calculated as a function of the predictive visual characteristics. For example:

[0102]

[0103] where represents the absolute value of the predicted luminance change rate, represents the absolute value of the predicted luminance change acceleration, and uncertainty k (t) represents the uncertainty measure of the prediction result for this region (such as prediction variance). The function f temporal can be a weighted sum or a more complex mapping function such that when the predicted change is more drastic or the uncertainty is higher, the value of the temporal attention information I t,k (t) is larger.

[0104] After obtaining the spatial saliency features spatial detail features and the temporal attention information I t,k (t), the weight calculation unit fuses this information to generate the final attention weight W k corresponding to the image region A k (t). This fusion process is defined by the multi-scale spatio-temporal attention model.

[0105] The fusion process may include the following steps:

[0106] First, the multi-scale spatial features can be integrated to obtain a comprehensive spatial attention score For example, it can be obtained by weighted averaging of spatial saliency features and spatial detail features at different scales or by combining them through a small neural network:

[0107]

[0108] where g spatial is a predefined spatial feature fusion function.

[0109] Subsequently, the obtained comprehensive spatial attention score is fused with the previously calculated temporal attention information I t,k (t) to generate the final attention weight W k (t). This fusion aims to balance the spatial importance of the image content at the current moment with the temporal importance of future expected changes. In a possible implementation, the final attention weight W k (t) can be obtained by weighted combination:

[0110]

[0111] where α and β are preset weight coefficients used to adjust the relative contributions of the spatial and temporal attention factors.

[0112] In another possible implementation, the fusion process can be implemented by a specially designed neural network model that takes the extracted spatial features and temporal attention information as inputs and directly outputs the attention weights.

[0113] The finally generated attention weight W k (t) is usually normalized, for example, to make its sum over all image regions equal to 1, or to constrain each weight value within the interval [0, 1]. This normalized attention weight W k (t) will be passed to the subsequent backlight optimization unit to guide the allocation of backlight brightness. Regions with high attention weights indicate that they are more visually important and their display quality should be prioritized.

[0114] A backlight optimization unit, which is configured to execute an optimization algorithm based on the attention weights, the predictive visual characteristics, and a preset global optimization objective function to determine the target backlight brightness of multiple backlight partitions corresponding to the multiple image regions;

[0115] In this embodiment, the backlight optimization unit is the core module for making backlight brightness decisions. Its purpose is to determine the optimal light-emitting intensity of each backlight partition on the premise of comprehensively considering various visual quality and system performance indicators:

[0116] The backlight optimization unit receives multiple image regions A for the current video frame t provided by the weight calculation unit k The respective corresponding attention weights W k (t), and the predictive visual characteristics for these image regions provided by the prediction analysis unit Based on this input information and in combination with a preset global optimization objective function The backlight optimization unit executes an optimization algorithm to calculate and determine the multiple image regions A k The respective target backlight brightness of the corresponding multiple backlight partitions

[0117] The global optimization objective function is constructed to comprehensively and balancedly evaluate the display effect and system power consumption. This objective function is usually designed as a scalar function composed of a weighted combination of multiple sub-objective terms. The goal of the optimization algorithm is to find a set of target backlight brightness values {B k (t)} k such that the function value reaches the optimal (e.g., minimized). Specifically, the global optimization objective function comprehensively evaluates the following aspects:

[0118] Accuracy of image content reproduction (fidelity term ):

[0119] This item aims to ensure that the adjusted backlight can accurately reproduce the visual brightness that the original image content should present. Its metric is based on the image region data of each of the image regions A k (e.g., the original average brightness extracted from or the target perceived brightness) and its corresponding attention weight W (t). k

[0120] The attention weight W k (t) dynamically adjusts and evaluates the relative importance of each image region here. For regions with high attention weights, the deviation of brightness reproduction will be more severely punished. In a possible implementation, the fidelity term can be expressed as:

[0121]

[0122] Among them, B k (t) is the brightness of the backlight zone corresponding to the image area A to be optimized k ; is the actual perceived brightness presented on the screen after the image area A is modulated by the liquid crystal panel when the backlight brightness is B k (t). This perceived brightness is related to B k (t), the average transmittance T k (t) of the liquid crystal pixels obtained from the image area data , and the gamma characteristic γ k of the display. For example, L′ lcd ≈T k (t)·(B k (t)) k ; L γlod (t) is the target display brightness of area A target,k , usually derived from the original image data k ; d(·,·) is a non - negative error metric function, such as the squared error (x - y) . 2 .

[0123] Smoothness of the transition of adjacent backlight zone brightness (spatial smoothness term ):

[0124] This term aims to avoid overly drastic brightness jumps between different backlight zones, thereby reducing possible halo - effect or blocking - artifact and enhancing the overall uniformity and naturalness of the picture. Its metric is based on the brightness relationship between adjacent image areas A k and A j , their corresponding attention weights W k (t) and W j (t), and the image edge information E k,j (t) between the image areas.

[0125] The attention weights W k (t) and W j (t), combined with the image edge information E k,j (t) between the image areas, dynamically adjust the constraint strength imposed when evaluating the smoothness of the transition of adjacent backlight zone brightness. For example, if there is a strong natural image edge between adjacent areas (E k,j (t) is large), a certain difference in backlight brightness is allowed; conversely, if the content between areas is smooth or it is a high - attention area, stricter brightness consistency is required.

[0126] In a possible implementation, the spatial smoothness term can be expressed as:

[0127]

[0128] where, represents the index set of the image regions adjacent to region A k ; ω s,kj is a dynamic weighting function, and its value is determined according to W k (t), W j (t) and E k,j (t). For example, ω s,kj can be set to be proportional to (W k (t) + W j (t)) and inversely proportional to E k,j (t), indicating that stronger smoothing constraints are required between high-attention regions or regions with low edge intensity.

[0129] Continuity of backlight brightness in the time series (temporal smoothness term ):

[0130] This term aims to ensure a smooth transition of backlight brightness between video frame sequences, avoiding discomfort to the human eye or a sense of screen flicker caused by sharp fluctuations in brightness. Its metric is based on the relationship between the target backlight brightness B k (t) of each backlight partition in the current frame and the actual backlight brightness B k (t - 1) of the corresponding partition in the previous frame, and combines the predicted brightness change trend included in the predictive visual characteristics .

[0131] The predicted brightness change trend included in the predictive visual characteristics is used to dynamically guide the desired change target when evaluating the continuity of the backlight brightness in the time series. That is, not only punishing rapid changes, but more encouraging the backlight to evolve smoothly towards the predicted trend.

[0132] In a possible implementation, the temporal smoothness term can be expressed as:

[0133]

[0134] where, B k (t - 1) is the brightness value of the backlight partition corresponding to image region A k in the previous frame; ΔB target,k (t) is the desired ideal brightness adjustment amount of this backlight partition in the current frame calculated based on the predictive visual characteristics (such as the predicted brightness change rate or change direction). If the predicted region is going to brighten, ΔB target,k(t) is positive, otherwise it is negative.

[0135] Energy consumption of the overall backlight system (energy consumption item ):

[0136] This item aims to reduce the operating power consumption of the entire backlight system and achieve green energy conservation. Its measurement is based on the target backlight brightness B of each backlight zone k (t) and its preset power consumption model P(B k (t)) (this model describes the relationship between backlight brightness and its actual power consumption, usually a non-linear increasing function), and comprehensively considers the attention weight W k (t) and the overall visual characteristics Ω of the current scene scene (t).

[0137] The attention weight W k (t) and combined with the overall visual characteristics Ω of the current scene scene (t) (for example, scenes with low average scene brightness and simple content indicating high energy-saving potential) are used to dynamically adjust the weighting ratio when evaluating the energy consumption of the overall backlight system. For areas with lower visual importance or scenes with high overall energy-saving potential, the backlight brightness can be reduced more actively to save energy.

[0138] In a possible implementation, the energy consumption item can be expressed as:

[0139]

[0140] Where P(B k (t)) is the power consumption corresponding to the backlight brightness B k (t); ω e,k is a dynamic weighting function, and its value will be adjusted according to the attention weight W k (t) (for example, proportional to 1 - W k (t), indicating that low-attention areas focus more on energy conservation) and the overall visual characteristics Ω of the scene scene (t) (for example, when Ω scene (t) indicates that the scene is darker or simpler, ω c,k increases accordingly to strengthen the energy-saving constraint).

[0141] Finally, the global optimization objective function ) can be the weighted sum of the above items:

[0142]

[0143] Where λ fid , λ spat , λ temp , λen is a preset non - negative weight coefficient used to balance the importance between different optimization objectives. These coefficients can be adjusted according to the application scenario or user preferences.

[0144] The backlight optimization unit then uses a suitable optimization algorithm to solve the above - mentioned global optimization objective function to find a set of optimal target backlight brightness values The optimization algorithms that can be used include, but are not limited to, gradient - based optimization methods (such as gradient descent method, conjugate gradient method), quadratic programming (QP) solver (if the objective function and constraints can be transformed into quadratic form), or various heuristic optimization algorithms (such as simulated annealing, genetic algorithms, etc.). During the optimization process, physical constraint conditions usually need to be satisfied, that is, the target brightness B k (t) of each backlight zone must be within its adjustable minimum brightness B min and maximum brightness B max therebetween.

[0145] A drive control unit, configured to generate and output a drive signal for controlling the actual brightness of the plurality of backlight zones from the target backlight brightness;

[0146] In this embodiment, as the final execution link of the LED backlight integrated control system of the liquid crystal display screen, the main responsibility of the drive control unit is to convert the abstract target backlight brightness value determined by the upstream backlight optimization unit into specific physical control instructions that can be recognized and executed by the underlying hardware:

[0147] The drive control unit receives the target backlight brightness of each of the plurality of backlight zones k for the current video frame t calculated by the backlight optimization unit These target backlight brightness values are usually normalized logical values (for example, within the range of [0, 1], where 0 represents the backlight off and 1 represents the maximum brightness) or represent the desired luminous flux or luminance units.

[0148] The core task is that the drive control unit converts the target backlight brightness of each of the plurality of backlight zones into pulse - width - modulation (PWM) parameters or analog drive current parameters for respectively controlling the light - emitting intensity of the corresponding backlight zones (usually composed of one or a group of LEDs). Whether to use PWM control or analog current control usually depends on the specific design and capabilities of the backlight drive hardware.

[0149] In a possible implementation manner, if pulse - width - modulation (PWM) control is adopted, the drive control unit will convert the target backlight brightness Convert to the duty cycle D of the corresponding PWM signal k (t). PWM control adjusts the average luminous intensity of the LED by changing the pulse width of the driving LED (i.e., the proportion of the "on" time in the entire cycle) within a fixed high-frequency cycle. Generally, a higher target backlight brightness corresponds to a larger duty cycle. This conversion process can be achieved by a predefined mapping function f PWM as follows:

[0150]

[0151] where D k (t) generally ranges from 0% to 100% (or 0 to 1). The function f PWM may not be a simple linear relationship because it needs to consider the optoelectronic conversion characteristics of the LED itself (i.e., the relationship between light output and PWM duty cycle may not be linear) and possible visual perception corrections (such as gamma correction to make the perceived brightness linearly related to the target brightness value). In some embodiments, f PWM can be implemented by a Look-Up-Table (LUT), which pre-stores the optimal PWM duty cycles corresponding to different target brightness values. These corresponding relationships can be obtained through detailed calibration and characteristic measurement of a specific LED backlight module.

[0152] As an alternative, if analog drive current control is adopted, the drive control unit will convert the target backlight brightness to the corresponding analog drive current value I k (t). By directly adjusting the magnitude of the current flowing through the LED, the luminous intensity of the LED can be controlled. Generally, the higher the current, the higher the brightness. This conversion process can also be achieved by a predefined mapping function f current as follows:

[0153]

[0154] where I k (t) must be within the safe operating current range of the LED. The function f current needs to accurately reflect the luminous flux-current (L-I) characteristic curve of the LED, which describes the non-linear relationship between the luminous intensity of the LED and the drive current. Similarly, to obtain accurate brightness control, the specific form or its parameters of f current can be determined through calibration tests of the LED module.

[0155] Regardless of the conversion method used, the core objective is to ensure that the converted physical control parameter (PWM duty cycle or analog current) can make the corresponding backlight zone generate the same brightness as the target backlight brightness The actual light output that is consistent

[0156] After completing the conversion from the target backlight brightness to the specific control parameters, the drive control unit will use the converted pulse width modulation parameter D k (t) or the analog drive current parameter I k (t) as the drive signal output. These drive signals are transmitted to each LED drive chip or circuit in the display backlight module. These drive chips or circuits precisely adjust the voltage or current applied to the corresponding LED backlight partition according to the received drive signals, thereby achieving independent and precise control of the actual light emission brightness of each backlight partition.

[0157] Please refer to Figure 2 , the present invention also provides an integrated control method for the LED backlight of a liquid crystal display, and the method includes the following steps:

[0158] S1. Receive the current video frame through the video input interface, and perform image partitioning processing on the current video frame to obtain the image region data of each of the multiple image regions;

[0159] In this step, the system first receives the input current video frame in real time through a standard video input interface, such as HDMI or Display-Port, etc. The received video frame data will then undergo a color space conversion process, for example, converting from a common RGB color space to a color space such as YCbCr that is more conducive to brightness information analysis, so as to separate the brightness component for subsequent processing. Immediately afterwards, according to the physical partition layout information preset by the liquid crystal display backlight hardware, the video frame (especially its brightness information) after color space conversion is spatially divided into multiple independent image regions, so that each image region corresponds to one or a group of independently controllable LED backlight partitions. After the partitioning is completed, the system extracts the key image features of each divided image region, such as the average brightness, brightness distribution, texture complexity, or edge strength of the region, etc. These features together constitute the image region data of the image region, providing a basis for subsequent analysis and decision-making.

[0160] S2. Combine the historical video frame information with the image region data of each of the multiple image regions, and apply a temporal prediction model to predict the predictive visual characteristics of each of the multiple image regions in the short term in the future;

[0161] In this step, the system focuses on the temporal dynamics of the video content. For each image region, the system combines the information of this region in several historical video frames (such as the past luminance change trajectory) and the image region data obtained from the current frame in step S1 to extract the luminance features and motion features that can reflect the content change of this region. Based on these extracted temporal features, the system applies a built-in temporal prediction model (such as a Kalman filter, an ARIMA model, or a model based on a recurrent neural network) to update and maintain an internal temporal state of this image region, which summarizes the current dynamic characteristics of this region. Then, according to this updated temporal state, the temporal prediction model infers and predicts the possible visual characteristics of this image region within a short future time window (such as the next few frames), mainly including its predicted luminance value and the predicted luminance change trend.

[0162] S3. Combine the image region data of each of the multiple image regions with the predicted visual characteristics, and apply a multi-scale spatio-temporal attention model to calculate the attention weights corresponding to each of the multiple image regions;

[0163] In this step, the goal of the system is to evaluate the relative importance of different image regions in the current video frame for human visual perception. For each image region, the system first extracts its spatial features at multiple different spatial scales or resolution levels based on the image region data obtained in step S1. These spatial features can include spatial saliency features reflecting the degree of regional prominence, and spatial detail features describing the internal texture details and edge sharpness of the region. At the same time, the system also combines the predicted visual characteristics of this image region obtained in step S2 (especially the predicted luminance change situation) to calculate its attention information in the time dimension, that is, the degree to which the future dynamic changes of this region may attract human eye attention. Finally, the system effectively fuses the extracted multi-scale spatial features with the calculated temporal attention information through a multi-scale spatio-temporal attention model, comprehensively evaluates the spatial importance of this image region at the current moment and the temporal importance of the future expected changes, and calculates a quantified attention weight value for each image region accordingly.

[0164] S4. Execute an optimization algorithm based on the attention weights, the predicted visual characteristics, and a preset global optimization objective function to determine the target backlight luminance of the multiple backlight partitions corresponding to the multiple image regions;

[0165] In this step, the system makes the core backlight brightness decision. The system comprehensively utilizes the attention weights of each image region calculated in the previous step, the predicted visual characteristics in the short-term future predicted in step S2, and a pre-set global optimization objective function. This global optimization objective function is a comprehensive evaluation criterion, which is designed to balance multiple interrelated and even potentially conflicting performance indicators, such as: ensuring the fidelity of accurately reproducing the brightness of the image content, maintaining the smoothness of the brightness transition between adjacent backlight zones to avoid halos or block effects, ensuring the continuity of the backlight brightness change in the time series to avoid flicker, and minimizing the energy consumption of the entire backlight system as much as possible. The system then executes an optimization algorithm (such as the gradient descent method or quadratic programming, etc.) to calculate an optimal target backlight brightness value for each backlight zone corresponding to an image region under the guidance of this global optimization objective function and considering the upper and lower limit constraints of the physical brightness of the LEDs.

[0166] S5: Generate and output a driving signal for controlling the actual brightness of the multiple backlight zones based on the target backlight brightness;

[0167] In this step, the system converts the calculated target backlight brightness value into actual hardware control instructions. For each backlight zone and its determined target backlight brightness in step S4, the system will convert it into corresponding physical driving parameters. If the backlight hardware is controlled by pulse width modulation (PWM), the target backlight brightness will be converted into PWM signal parameters with a specific duty cycle; if it is controlled by analog current drive, the target backlight brightness will be converted into a specific drive current value. This conversion process takes into account the optoelectronic characteristics of the LEDs themselves to ensure that the generated driving parameters can make the LEDs emit light intensities consistent with the target brightness. Finally, these generated PWM parameters or analog drive current parameters are used as driving signals and output to each LED unit or zone in the liquid crystal display backlight module through the driving circuit, so as to accurately control their respective actual light emission brightness.

[0168] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An LED backlight integrated control system for a liquid crystal display screen, characterized in that, The system includes the following modules: A data processing unit, configured to receive a current video frame via a video input interface, and perform image partitioning processing on the current video frame to obtain image region data for each of a plurality of image regions; A prediction and analysis unit, configured to combine historical video frame information with the image region data for each of the plurality of image regions, and apply a temporal prediction model to predict the predictive visual characteristics of each of the plurality of image regions in the short-term future; A weight calculation unit, configured to combine the image region data for each of the plurality of image regions with the predictive visual characteristics, and apply a multi-scale spatio-temporal attention model to calculate the attention weights corresponding to each of the plurality of image regions; A backlight optimization unit, configured to execute an optimization algorithm based on the attention weights, the predictive visual characteristics, and a preset global optimization objective function to determine the target backlight brightness for a plurality of backlight partitions corresponding to the plurality of image regions; A drive control unit, configured to generate and output a drive signal for controlling the actual brightness of the plurality of backlight partitions based on the target backlight brightness; 2. The LED backlight integrated control system of the liquid crystal display screen according to claim 1, wherein Specifically, the data processing unit is configured to: Perform color space conversion processing on the received current video frame; Divide the current video frame after color space conversion processing into the plurality of image regions according to a preset backlight physical partition layout; And extract the corresponding image data for each image region as the image region data.

3. The LED backlight integrated control system of the liquid crystal display screen according to claim 1, wherein Specifically, the prediction and analysis unit is configured to: For each of the image regions, extract the brightness feature and motion feature of the image region from the historical video frame information and the image region data of the image region; Update the temporal state corresponding to the image region through the temporal prediction model based on the extracted brightness feature and motion feature; And predict the predictive brightness value and predictive brightness change trend of the image region in the short-term future based on the updated temporal state as the predictive visual characteristics.

4. The LED backlight integrated control system of the liquid crystal display screen according to claim 3, wherein When the prediction and analysis unit is used to update the temporal state, the image region data of each image region includes brightness-related features and motion-related features extracted from the image region; the predictive visual characteristics predicted through the temporal state include the predictive brightness value and predictive brightness change trend of the image region.

5. The LED backlight integrated control system of the liquid crystal display screen according to claim 1, characterized in that Specifically, the weight calculation unit is configured to: For each of the image regions, extract the multi-scale spatial features of the image region based on its image region data; And combine the multi-scale spatial features with the predictive visual characteristics of the image region, apply the multi-scale spatio-temporal attention model, comprehensively evaluate the visual importance of the image region in the spatial dimension and the time dimension, and calculate the corresponding attention weight of the image region.

6. The LED backlight integrated control system of the liquid crystal display screen according to claim 5, characterized in that, When applying the multi-scale spatio-temporal attention model, the weight calculation unit, for each of the image regions: Extract the spatial saliency feature and spatial detail feature of the image region at multiple spatial scales based on its image region data; Calculate the temporal attention information by combining the predictive visual characteristics of the image region; Fuse the extracted spatial saliency features, the spatial detail features, and the calculated temporal attention information to generate the attention weights corresponding to the image region.

7. The LED backlight integrated control system of the liquid crystal display screen according to claim 1, characterized in that, The global optimization objective function adopted by the backlight optimization unit constructs a comprehensive evaluation of the following aspects: The accuracy of image content reproduction, which is measured based on the image region data of each image region and its corresponding attention weights; The smoothness of the brightness transition between adjacent backlight partitions, which is measured based on the brightness relationship between adjacent image regions, their corresponding attention weights, and the image edge information between the image regions; The continuity of the backlight brightness in the time series, which is measured based on the backlight brightness relationship between the current frame and the previous frame and the predicted brightness change trend included in the predictive visual characteristics; The energy consumption of the overall backlight system, which is measured based on the relationship between the target backlight brightness of each backlight partition and its preset power consumption model, the attention weights, and the overall visual characteristics of the current scene.

8. The LED backlight integrated control system of the liquid crystal display screen according to claim 7, characterized in that, In the global optimization objective function: The attention weights are used to dynamically adjust the relative importance of each image region when evaluating the accuracy of image content reproduction, and dynamically adjust the constraint strength applied when evaluating the smoothness of the brightness transition between adjacent backlight partitions in combination with the image edge information between the image regions; The attention weights, combined with the overall visual characteristics of the current scene, are used to dynamically adjust the weighting ratio when evaluating the energy consumption of the overall backlight system; The predicted brightness change trend included in the predictive visual characteristics is used to dynamically guide the desired change target when evaluating the continuity of the backlight brightness in the time series.

9. The LED backlight integrated control system of the liquid crystal display screen according to claim 1, characterized in that The drive control unit is specifically configured to convert the target backlight brightness of each of the multiple backlight partitions into pulse width modulation parameters or analog drive current parameters for respectively controlling the light emission intensity of the corresponding backlight partition, and output the converted pulse width modulation parameters or the analog drive current parameters as the drive signal.

10. An LED backlight integrated control method for a liquid crystal display screen, applied to the system described in any one of claims 1-9, characterized in that, The method includes the following steps: S1. Receive the current video frame through the video input interface, and perform image partitioning processing on the current video frame to obtain the image region data of each of the multiple image regions; S2. Combine the historical video frame information with the image region data of each of the multiple image regions, and apply a temporal prediction model to predict the predictive visual characteristics of each of the multiple image regions in the short term in the future; S3. Combine the image region data of each of the multiple image regions and the predictive visual characteristics, and apply a multi-scale spatio-temporal attention model to calculate the attention weights corresponding to each of the multiple image regions; S4. Execute an optimization algorithm based on the attention weights, the predictive visual characteristics, and a preset global optimization objective function to determine the target backlight brightness of the multiple backlight partitions corresponding to the multiple image regions; S5: Generate and output a drive signal for controlling the actual brightness of the multiple backlight partitions based on the target backlight brightness.

Images