A high-contrast mini-led control method, system and medium

Abstract

The application discloses a high-contrast MiniLED control method, system and medium, relates to the technical field of MiniLED display control, and ensures the uniform distribution of luminosity on the screen by subdividing luminosity data and adjusting the luminosity data on the adjacent boundary line; meanwhile, the precision adjustment of the luminosity color difference avoids unnatural color transition or obvious color difference problems; finally, the luminosity regulation model can realize more accurate luminosity coordinated transition, the display effect is more real and delicate, and a plurality of display scenes can be adapted; the whole realizes accurate control and optimization of the luminosity of the MiniLED backlight area; each link solves the problems of luminosity unevenness, color distortion and unnatural brightness transition that may occur in display in different dimensions, especially plays a key role in reducing the halo effect.

Description

Technical Field

[0001] This invention relates to the field of MiniLED display control technology, and in particular to a control method, system and medium for high-contrast MiniLEDs. Background Technology

[0002] MiniLED technology has made continuous breakthroughs in recent years and has become one of the important representatives of new display technologies in the 21st century. With the continuous maturation of this technology, MiniLED displays, with their long lifespan and excellent brightness performance, are gradually becoming widely used information products worldwide. From home appliances, mobile phones, televisions, and computer monitors to automobiles and smart wearable devices, MiniLED displays have demonstrated their superior technological advantages and have gained widespread recognition and praise from consumers and industry professionals.

[0003] MiniLED displays excel in brightness, offering higher brightness and finer image quality than traditional LED technologies. They deliver richer colors and a more immersive visual experience, making them an increasingly popular choice for high-end monitors, televisions, and other devices requiring high-quality images. Furthermore, MiniLED's long lifespan is another significant advantage. Compared to traditional displays, it effectively extends the lifespan of devices, reduces replacement frequency, and improves overall cost-effectiveness.

[0004] With the rapid development of 5G technology both domestically and internationally, many emerging technologies are undergoing accelerated iteration, including MiniLED display technology, which is also ushering in new development opportunities. The high-speed transmission and low latency of 5G greatly improve the speed and stability of image transmission, providing better support and assurance for high-resolution, high-contrast display effects. However, despite this, MiniLED direct-display technology still faces certain challenges in achieving high-contrast and ultra-high-resolution display effects. To achieve truly groundbreaking display effects, greater efforts are needed in technological innovation and process optimization, especially in improving screen contrast, resolution, and color accuracy, where there is still considerable room for improvement. Therefore, although MiniLED technology has broad application prospects, achieving ultimate display effects still requires continuous research and exploration within the industry.

[0005] In summary, how to regulate the luminance of MiniLED displays to coordinate the transition between brightness and darkness and reduce halo effects is an urgent problem to be solved and optimized in the control system of high-contrast MiniLEDs. Summary of the Invention

[0006] This invention provides a control method, system, and medium for high-contrast MiniLEDs, solving the technical problem in related technologies of how to regulate the luminance of MiniLED displays to coordinate the transition between brightness and darkness and reduce halo effects.

[0007] To address the aforementioned technical problems, this invention provides a control method, system, and medium for high-contrast MiniLEDs, with the specific technical solution as follows:

[0008] In a first aspect, a method for controlling a high-contrast MiniLED includes the following steps:

[0009] Collect luminance data from multiple time frames of the MiniLED screen to obtain a luminance data set; based on the luminance data set, divide the MiniLED backlight area into multiple sub-regions to obtain a luminance data subset; based on the luminance data subset, further subdivide the luminance data of the multiple sub-regions into multiple small-sized backlight blocks, and label the multiple backlight blocks to obtain a luminance label dataset.

[0010] Obtain the light intensity data of each data item in the photometric labeled dataset to obtain a light intensity dataset; smooth the boundary lines between adjacent data items in the photometric labeled dataset according to the differences between the light intensity data items to obtain a first control dataset.

[0011] The photometric color level data of each data item in the photometric marker dataset is obtained to obtain the photometric color level dataset; based on the smoothed photometric marker dataset, the boundary lines between adjacent data items in the photometric marker dataset are adjusted according to the color difference between the photometric color level data to obtain the second control dataset.

[0012] The photometric control model is constructed by training the photometric labeled dataset, the first control dataset, and the second control dataset to output a photometric control standard recognition result representing the relationship between adjacent data items in the photometric labeled dataset; at least two other data items in the photometric labeled dataset are input into the photometric control model to output a photometric control prediction result representing the relationship between other adjacent photometric labeled data items.

[0013] Based on the prediction results of the photometric control model, the photometric transition between data items in the photometric labeled dataset is coordinated to reduce the halo effect of the MiniLED screen display.

[0014] As a further optimization of the present invention, based on the photometric data set, the MiniLED backlight area is divided into multiple sub-regions to obtain a subset of photometric data, including:

[0015] The photometric data set is preprocessed to obtain a preprocessed photometric data set; the photometric data set contains the average brightness value of each pixel in the backlight area.

[0016] The photometric data items in the preprocessed photometric data set are divided into regions according to their spatial distribution by setting a photometric threshold, thereby dividing the backlight region into multiple sub-regions; the sub-regions are further divided into bright areas and dark areas according to the photometric values.

[0017] The bright area and the dark area are divided into K initial center points according to the luminance value distribution of each sub-region; each pixel of the sub-region is assigned to the nearest initial center point to obtain the division center point of each cluster; the classification results of the bright area and the dark area are updated based on the division center points until the classification results converge.

[0018] Based on the classification convergence result, the photometric data set is used to finally classify the bright areas and the dark areas to obtain the photometric data subset.

[0019] As a further optimization of the present invention, based on the photometric data subset, the photometric data of multiple sub-regions are subdivided into multiple small-sized backlight blocks, and the multiple backlight blocks are labeled to obtain a photometric labeled dataset, including:

[0020] The photometric data subset is divided into several square regions of equal size using a 1x1 sliding window with a step size of 1.

[0021] Each of the divided square regions is a backlight block; the backlight block covers different photometric data portions of the photometric data subset; the backlight block represents a block-shaped region of uniform size in the bright or dark areas of the sub-region;

[0022] The average illumination intensity of each backlight block is extracted to measure the degree of luminance variation of the backlight block and to capture the texture features of the luminance variation. Based on the extracted features, each backlight block is labeled to obtain a luminance labeled dataset.

[0023] The photometric labeling dataset is assigned multiple labels based on the target characteristics of the label; the multiple labels include highlight, low light, shadow or normal lighting.

[0024] As a further optimization of the present invention, the boundary lines between adjacent data items in the photometric labeled dataset are smoothed according to the differences between the light intensity data items to obtain a first control dataset, including:

[0025] Based on the difference in photometric intensity between adjacent data items in the photometric labeled dataset, the transition anomaly at the photometric intersection between adjacent data items is identified to obtain the boundary line between adjacent data items.

[0026] The adjacent data items represent adjacent backlight blocks after being marked in the photometric label dataset; the boundary line represents the common boundary line between adjacent backlight blocks, that is, the backlight block dividing line in the photometric data subset.

[0027] The boundary line includes two dimensions, vertical and horizontal, to obtain various transition conditions of luminance intensity of the boundary line. When the luminance intensity transition between two adjacent luminance marker data items is drastic and abrupt, the drastic and abrupt luminance intensity transition is not smooth, causing a halo effect on the screen display.

[0028] The light intensity of the transitional abrupt change is smoothed to obtain the first regulation dataset.

[0029] As a further optimization of the present invention, the photometric intensity of the transitional abrupt change is smoothed to obtain the first regulatory dataset, including:

[0030] Let adjacent data items I(x,y) and I(x',y') represent the brightness intensity data on both sides of the boundary line, respectively; through To obtain the difference in luminous intensity; where, Indicates the difference in photometric intensity;

[0031] Set a light intensity threshold for drastic changes ,when To determine whether the difference in photometric intensity values ​​would cause a drastic abrupt change in transition; through To smooth out the abrupt changes in luminous intensity; where, G(i,j,σ) represents the smoothed luminous intensity value, G(i,j,σ) represents the smoothing weight coefficient, σ controls the standard deviation of the smoothing degree, and I(x+i,y+j) represents the luminous intensity value in the neighborhood (i,j).

[0032] After the smoothing process is completed, the drastic changes in photometric intensity in the photometric labeled dataset have been effectively suppressed, and the generated smoothed dataset is the first controlled dataset. By comparing two adjacent photometric labeled data items before and after the processing, it is determined whether a halo effect occurs. If not, proceed to the next step; if so, return to the previous step for reprocessing. This is to evaluate the smoothing effect.

[0033] As a further optimization of the present invention, the boundary lines between adjacent data items in the photometric marker dataset are adjusted with precision based on the color difference between the photometric color scale data to obtain a second control dataset, including:

[0034] According to the color gradation of the color space, there are three hues; based on the three hues, two adjacent data items in the photometric labeled dataset are represented as (R,G,B) and (r,g,b), respectively.

[0035] Based on the three aforementioned hues, through Obtain the hue difference metric value between two adjacent photometric data items; preset color difference threshold. ,when This indicates a significant color difference.

[0036] In the formula, ΔE represents the measure of hue difference, R and r represent the lightness of two adjacent photometric data items, G and g represent the green-red components of the color, and B and b represent the blue-yellow components.

[0037] By analyzing the hue difference metric between two adjacent photometric data items, the light scheduling at the boundary line position is finely adjusted to eliminate abrupt transitions in color difference between adjacent photometric data items.

[0038] As a further optimization of the present invention, by analyzing the hue difference metric between two adjacent photometric data items, the light scheduling at the boundary line position is finely adjusted to eliminate abrupt color difference transitions between two adjacent photometric data items, including:

[0039] when When α is close to 1, it indicates a natural transition in the hue difference measure; when When α is close to 0, it indicates a sharp abrupt change in the hue difference measurement, requiring precision adjustment; where α represents the precision adjustment weighting factor.

[0040] Based on the brightness intensity between two adjacent photometric data items, through

[0041] To adjust the precision of the hue difference measure between two adjacent photometric data items; where, This indicates the adjustment precision value for the hue difference metric. α is adjusted according to the magnitude of the color difference ΔE to eliminate abrupt color difference transitions.

[0042] After precision adjustment, the hue differences between adjacent data items in the photometric labeled dataset are made to transition naturally, ultimately generating a second controlled dataset; through... To evaluate the effect of the accuracy adjustment; where w a ΔE represents the precision-adjusted evaluation value, and β represents the evaluation factor used to assess the sensitivity of weight adjustments. The larger ΔE is, the higher the w... a The smaller the value, the stronger the precision adjustment processing in that area.

[0043] As a further optimization of the present invention, the photometric control model includes:

[0044] The first and second regulation datasets are integrated into a dataset, and the historically acquired dataset is used to generate structured data. The structured data is then encoded into sequence data to train a photometric regulation model.

[0045] The sequence data is input into the photometric control model; the photometric control model includes an input layer, a first hidden layer, a second hidden layer, a third hidden layer and an output layer, and the intermediate representation data of multiple hidden layers are transmitted to the output layer. The output layer outputs the photometric control standard recognition result representing the relationship between two adjacent data items in the photometric labeled dataset.

[0046] Two adjacent data items in the photometric labeled dataset are input into the photometric regulation model to output the newly acquired output layer output, which represents the photometric regulation prediction result between the next pair of adjacent photometric labeled data items.

[0047] Secondly, a control system for a high-contrast MiniLED, the system comprising an electronic device including a memory, a processor, and a control method program for a high-contrast MiniLED stored in the memory and executable on the processor, wherein the control method program for a high-contrast MiniLED, when executed by the processor, implements the steps of a control method for a high-contrast MiniLED, the system comprising:

[0048] Data acquisition module: It is used to acquire luminance data of MiniLED screen display from multiple time frames to obtain a luminance data set; based on the luminance data set, the MiniLED backlight area is divided into multiple sub-regions to obtain a luminance data subset; based on the luminance data subset, the luminance data of multiple sub-regions is further subdivided into multiple small-sized backlight blocks, and the multiple backlight blocks are marked to obtain a luminance labeled dataset;

[0049] The data processing module is used to acquire the luminance intensity data of each data item in the photometric marker dataset to obtain a luminance intensity dataset; to smooth the boundary lines between adjacent data items in the photometric marker dataset according to the differences between the luminance intensity data items to obtain a first control dataset; to acquire the photometric color level data of each data item in the photometric marker dataset to obtain a photometric color level dataset; and to adjust the boundary lines between adjacent data items in the photometric marker dataset according to the color difference between the photometric color level data based on the smoothed photometric marker dataset to obtain a second control dataset.

[0050] Photometric regulation module: It is used to train and construct a photometric regulation model from the photometric labeled dataset, the first regulation dataset, and the second regulation dataset to output a photometric regulation standard recognition result representing the relationship between adjacent data items in the photometric labeled dataset; and to input at least two other data items from the photometric labeled dataset into the photometric regulation model to output a photometric regulation prediction result representing the relationship between other adjacent photometric labeled data items.

[0051] The control and evaluation module is used to coordinate the luminance transition between data items in the luminance-labeled dataset based on the prediction results of the luminance control model, so as to reduce the halo effect of the MiniLED screen display.

[0052] Thirdly, a computer-readable storage medium storing a program that, when executed by a processor, implements the steps of the control method for a high-contrast MiniLED.

[0053] The present invention offers at least the following advantages: By collecting luminance data from multiple time frames of the MiniLED screen, it accurately reflects the luminance changes of the screen at different points in time. This provides a more comprehensive set of luminance data for subsequent luminance adjustment, ensuring the accuracy and timeliness of the adjustment. Dividing the entire backlight area into multiple sub-regions allows for localized handling of luminance issues, rather than uniform adjustment of the entire area. This subdivision effectively identifies luminance changes in different areas, enabling more precise adjustments and reducing display problems caused by localized unevenness.

[0054] By subdividing the luminance data of a sub-region into multiple small backlight blocks and marking them, and further subdividing the backlight blocks and marking each small backlight block, the precision of luminance variation control can be improved, the differences between different regions can be reduced, and the consistency of display effect can be enhanced. This step provides the necessary basic data for subsequent precision adjustments.

[0055] We acquire a brightness intensity dataset and perform boundary smoothing. By acquiring the brightness intensity dataset for each backlight block, we can more intuitively understand the brightness changes in each region. Smoothing the boundary lines reduces the brightness differences between backlight blocks, making the transition of luminance changes more natural, thereby avoiding halo effects and abrupt brightness differences, and improving the visual experience.

[0056] Acquiring a photometric color gradation dataset and performing precision adjustments on the boundary lines helps ensure that photometric control not only considers brightness differences but also finely adjusts color transitions. This process avoids uneven color distribution or significant color differences, ensuring high-quality visual performance of the image during display.

[0057] A photometric regulation model is trained to construct a standard for identifying and predicting photometric regulation results. By training the photometric regulation model and combining the photometric labeled dataset, the first regulation dataset, and the second regulation dataset, a predictive regulation model can be constructed. This model can identify the differences between adjacent photometric data items and predict how to regulate photometric intensity to minimize halo effects and brightness unevenness.

[0058] Photometric coordination transitions are performed based on the prediction results of the photometric control model. By utilizing the prediction results of the photometric control model, the photometric values ​​between various data items in the photometric-labeled dataset can be coordinated, ensuring a smoother transition between photometric values. This step helps reduce unnatural transitions between backlight blocks, optimizes display effects, effectively reduces the halo effect commonly found in MiniLED displays, and improves visual quality. Attached Figure Description

[0059] Figure 1 This is a schematic flowchart of a high-contrast MiniLED control method provided by an embodiment of the present invention;

[0060] Figure 2 This is a schematic diagram of a high-contrast MiniLED control system provided by an embodiment of the present invention. Detailed Implementation

[0061] The present application will now be described in further detail with reference to the accompanying drawings. It should be noted that the following specific embodiments are only used to further illustrate the present application and should not be construed as limiting the scope of protection of the present application. Those skilled in the art can make some non-essential improvements and adjustments to the present application based on the above application content.

[0062] This embodiment provides a control method, system, and medium for high-contrast MiniLEDs, the specific implementation of which is as follows:

[0063] like Figure 1 As shown, a method for controlling a high-contrast MiniLED includes the following steps:

[0064] Step 11: Collect luminance data of MiniLED screen display from multiple time frames to obtain a luminance data set; based on the luminance data set, divide the MiniLED backlight area into multiple sub-regions to obtain a luminance data subset; based on the luminance data subset, further subdivide the luminance data of the multiple sub-regions into multiple small-sized backlight blocks, and label the multiple backlight blocks to obtain a luminance label dataset.

[0065] Step 12: Obtain the luminance intensity data of each data item in the luminance marker dataset to obtain a luminance intensity dataset; smooth the boundary lines between adjacent data items in the luminance marker dataset according to the differences between the luminance intensity data items to obtain a first control dataset; obtain the luminance gradation data of each data item in the luminance marker dataset to obtain a luminance gradation dataset; based on the smoothed luminance marker dataset, adjust the boundary lines between adjacent data items in the luminance marker dataset according to the color difference between the luminance gradation data to obtain a second control dataset;

[0066] Step 13: Train the photometric control model using the photometric labeled dataset, the first control dataset, and the second control dataset to output a photometric control standard recognition result representing the relationship between adjacent data items in the photometric labeled dataset; input at least two other data items from the photometric labeled dataset into the photometric control model to output a photometric control prediction result representing the relationship between other adjacent photometric labeled data items;

[0067] Step 14: Based on the prediction results of the photometric control model, coordinate the photometric transition between data items in the photometric labeled dataset to reduce the halo effect of the MiniLED screen display.

[0068] In this embodiment of the invention, step 11 involves acquiring the different brightness (luminance) of each MiniLED display at different time frames on the MiniLED display. Luminance data of the display is collected through multiple time frames (time periods). This data is typically collected by a photometric sensor or an image sensor, and the data set contains brightness information of different areas of the entire screen at different time points. By acquiring multi-time frame data, the brightness performance of the screen at different time periods can be fully understood, thereby enabling more accurate photometric analysis. Capturing photometric data from multiple time frames helps to discover potential brightness non-uniformity or dynamic changes, thus providing a reliable data foundation for subsequent photometric control models.

[0069] The MiniLED backlight area is divided into multiple sub-regions based on a photometric data set, obtaining a subset of photometric data. Based on this collected photometric data set, the entire backlight area of ​​the display screen is divided into multiple sub-regions, each corresponding to a different subset of photometric data. This division can be based on the physical area of ​​the screen (e.g., a rectangle or grid) or on non-uniform division based on the characteristics of photometric distribution. Dividing the backlight area into multiple sub-regions improves the accuracy of photometric control. Different areas may exhibit different photometric performance; further subdivision allows for more precise adjustment of the brightness of each area. This division ensures that localized photometric problems (e.g., excessively high or low brightness in a certain area) can be accurately identified and optimized.

[0070] Each sub-region is further subdivided into multiple smaller backlight blocks, each representing a smaller area, and each backlight block has corresponding photometric data. By adding labels (e.g., numbers or locations) to each backlight block, a labeled dataset containing the photometric data of each backlight block is formed. By subdividing the backlight blocks, the brightness of each small block can be adjusted more finely, reducing brightness uniformity of the display screen. The labeling of the backlight blocks provides structured information for subsequent data processing and model training, making photometric control in subsequent steps more systematic and efficient.

[0071] In step 12, a brightness intensity dataset is formed based on the brightness intensity information of each data item in the photometric label dataset obtained in step 11. Brightness intensity represents the brightness value of each backlight block. Next, based on the trend of brightness value changes, the boundary lines between adjacent data items are smoothed. The boundary lines refer to the brightness transition areas between adjacent backlight blocks; the smoothing process adjusts these transition areas to be more natural and smooth based on brightness differences. After smoothing, abrupt changes in display brightness are avoided, unnatural brightness jumps are reduced, and viewing comfort is improved. Through the generated first control dataset, the brightness changes of each backlight block can be precisely controlled, making the brightness transition of the display screen smoother and more natural.

[0072] Step 12 extracts the photometric gradation (i.e., the range of color temperature or color gamut) for each data item from the photometric labeled dataset, forming a photometric gradation dataset. Based on the differences between gradations, the boundary lines are precision adjusted. This adjustment, based on the gradation differences between adjacent data items, aims to improve the accuracy of color transitions and avoid abruptness during color transitions. Through precision adjustment, color transitions on the display screen become smoother and more accurate, avoiding obvious color jumps or deviations and improving color coherence. This adjustment helps achieve a high-quality visual experience, especially important when displaying high dynamic range (HDR) content.

[0073] Step 13, based on the photometric labeled dataset, the first regulation dataset, and the second regulation dataset obtained in Steps 11 and 12, trains a photometric regulation model using machine learning methods. During training, the model learns the relationships between photometric data and generates photometric regulation standards between adjacent data items. Then, the photometric regulation model is used to predict other data items in the photometric labeled dataset, thereby obtaining the regulation prediction results between adjacent photometric data items. The trained photometric regulation model can automatically identify and predict photometric regulation needs, greatly improving the efficiency and accuracy of regulation. Through this model, intelligent processing of photometric regulation can be achieved, reducing manual intervention and adapting to different display needs.

[0074] Step 14: In this step, based on the prediction results of the photometric control model constructed in Step 13, the photometric transition between data items in the photometric labeled dataset is coordinated. Specifically, according to the model output, the brightness of adjacent backlight blocks is fine-tuned to eliminate the "halo effect" caused by uneven brightness or unsmooth boundary transitions. The halo effect usually manifests as an unnatural transition of brightness or light overflow, affecting the quality of the display effect; by finely controlling the photometric transition, the halo effect can be effectively reduced or eliminated, making the brightness distribution of the display screen more uniform; this coordinated transition can improve the overall visual effect of the display screen and enhance the user's viewing experience, especially in high-contrast or dark-field displays, where the reduction of the halo effect is particularly important.

[0075] In a preferred embodiment of the present invention, step 11 above, based on the photometric data set, divides the MiniLED backlight area into multiple sub-regions to obtain a subset of photometric data, including:

[0076] Step 111: Preprocess the photometric data set to obtain a preprocessed photometric data set; the photometric data set contains the average brightness value of each pixel in the backlight area.

[0077] Step 112: Based on the spatial distribution of each photometric data item in the preprocessed photometric data set, the region is divided by setting a photometric threshold, thereby dividing the backlight region into multiple sub-regions; the sub-regions are further divided into bright areas and dark areas according to the photometric value.

[0078] Step 113: Select K initial center points for the division of the bright area and the dark area according to the luminance value distribution of each sub-region; assign each pixel of the sub-region to the nearest initial center point to obtain the division center point of each cluster; update the classification results of the bright area and the dark area based on the division center points until the classification results converge.

[0079] Step 114: Based on the classification convergence result, the photometric data set is used to finally classify the bright area and the dark area to obtain the photometric data subset.

[0080] In this embodiment of the invention, in step 111, the photometric data (i.e., the brightness information of each pixel) in the image is first collected and organized. This data can represent the average brightness value of each pixel in the backlit area of ​​the image. The preprocessing process may include operations such as denoising, normalization, and filtering, aiming to reduce errors and noise in the original data, improve the accuracy of subsequent analysis, reduce noise in the data, enhance effective information, provide clearer and more accurate photometric data for subsequent steps, and ensure that the data is balanced across different photometric ranges so that subsequent processing steps are more precise.

[0081] Step 112 sets a threshold for the photometric data items in the photometric dataset preprocessed in Step 111 based on their spatial distribution and photometric values. This threshold is then used to divide the backlit region. In this way, the backlit region is divided into multiple sub-regions, and each sub-region is further divided into "bright areas" and "dark areas" based on its photometric value. The photometric threshold can be set based on a standard, such as average brightness, median, or distribution characteristics. The divided "bright areas" and "dark areas" represent different light intensity regions, more accurately representing different brightness regions in the image. This refines the backlit region, enabling the handling of more complex light distributions. Dividing it into multiple sub-regions facilitates subsequent analysis, especially in distinguishing different brightness regions. It also helps in the separate processing or optimization of high-brightness and low-brightness regions, which is particularly important in tasks such as image enhancement.

[0082] In step 113, K initial center points are selected for photometric data category analysis. These initial center points are typically selected by analyzing the photometric value distribution of each sub-region in bright and dark areas, aiming to represent the center points of different brightness regions. Then, the pixels of each sub-region are assigned to the nearest initial center point, so that each cluster forms a cluster, called a "cluster" or "category". Based on these initial center points, the pixels of each sub-region are iteratively updated until the classification results converge. This is used to find the optimal classification result for each sub-region, accurately identify and divide bright and dark areas in the image, and perform efficient image segmentation. This updates help distinguish similar regions, thus accurately reflecting the spatial distribution of different brightness regions. Through multiple iterations, the final classification result is ensured to be stable, reducing errors.

[0083] Step 114, based on steps 111, 112, and 113, further determines the final region classification result based on the classification convergence result. In this stage, the luminance data set of the backlight region is precisely divided into different bright and dark areas, forming the final luminance data subset; a clear classification result is obtained, providing a foundation for subsequent image processing or display optimization; precise division of bright and dark areas can facilitate more complex image processing, such as enhancing contrast and adjusting brightness; further improving the visual effect of the image, making it more in line with human visual habits, especially in environments with varying brightness.

[0084] The steps described above work together, with step 111 providing a clean and standardized data foundation for subsequent region segmentation and cluster analysis. Without preprocessing, subsequent classification and region segmentation may be affected by noise, leading to decreased accuracy. Step 112 relies on the preprocessing results of step 111 and further subdivides the region by setting a photometric threshold, providing a more refined image region that facilitates subsequent classification processing. Step 113 selects appropriate initial center points for the segmented regions from step 112 by combining photometric value distribution and classifies pixels based on these points, providing a crucial step for the final image segmentation. Step 114 completes the final region segmentation based on the results of steps 111, 112, and 113, ensuring clear definitions of bright and dark areas and providing a useful subset of photometric data for practical applications.

[0085] In a preferred embodiment of the present invention, step 11 above involves subdividing the photometric data of multiple sub-regions into multiple small-sized backlight blocks based on the photometric data subset, and labeling the multiple backlight blocks to obtain a photometric labeled dataset, including:

[0086] Step 115: Divide the photometric data subset into several square regions of equal size using a 1x1 sliding window with a step size of 1;

[0087] Step 116: Each of the square regions after division is a backlight block; the backlight block covers different photometric data portions of the photometric data subset; the backlight block represents a block-shaped region of uniform size in the bright or dark areas of the sub-region.

[0088] Step 117: Extract the average illumination intensity of each backlight block to measure the degree of luminance variation of the backlight block and capture the texture features of luminance variation; based on the extracted features, label each backlight block to obtain a luminance label dataset.

[0089] Step 118: The photometric labeling dataset is assigned multiple labels based on the target features of the label; the multiple labels include highlight, low light, shadow or normal lighting.

[0090] In this embodiment of the invention, step 115 involves traversing an image or photometric data subset using a 1x1 sliding window with a step size of 1 (i.e., sliding one pixel at a time). This sliding window method divides the photometric data subset into several small square regions of equal size. These small square regions are called "backlight blocks." The step size of 1 for the sliding window means that only one pixel is processed each time the window moves, which helps preserve the detail of the photometric data and avoids excessive smoothing. It also helps to accurately divide the image region, ensuring that each backlight block can meticulously cover different parts of the photometric data subset. This fine division provides more information in subsequent feature extraction processes, helping to more accurately capture photometric changes in the image.

[0091] Based on step 115, each square region (backlight block) divided in step 116 covers a different part of the photometric data subset; each backlight block has a uniform size, regardless of the illumination intensity within that region. This means that even in a region with drastic photometric variations, the area of ​​each backlight block remains the same, maintaining structural consistency; this uniform-size division method simplifies subsequent processing because the photometric data features of each block are relatively independent and consistent; thus avoiding complex regional deformations and facilitating the extraction and comparison of illumination intensity changes in subsequent analysis.

[0092] Based on step 116, step 117 calculates the average illumination intensity of pixels in each backlight block, that is, the average illumination intensity of all pixels in that area. By obtaining the average illumination intensity of the backlight block, the illumination information of that area can be obtained, thereby measuring the degree of change in luminance. At the same time, by analyzing the texture of these backlight blocks, the illumination change patterns in the image can be captured, such as shadows, reflections, and other features. The extracted average illumination intensity provides a way to quantify luminance changes, which helps to further analyze the illumination characteristics of the image. Capturing texture features helps to reveal the details of illumination changes, such as the existence of highlight and low-light areas, providing a basis for subsequent labeling and classification.

[0093] Step 118: Based on the photometric features (such as average illumination intensity and texture features) extracted in Step 117, label each backlight block; assign different labels to each backlight block, common labels include "highlight", "lowlight", "shadow" and "normal illumination";

[0094] Highlights: usually refer to areas with strong lighting.

[0095] Low light: refers to areas with weak light intensity.

[0096] Shadows: areas formed by the obstruction of light, which may have lower light intensity or special textures.

[0097] Normal lighting: refers to the normal light intensity in most areas, without excessive changes in lighting.

[0098] This labeling method transforms illumination data into information with clear classifications, providing valuable input for subsequent image analysis (such as enhancement, correction, or generation). Labeled illumination data helps automated systems understand the illumination conditions in images, thereby providing useful feature support for visual tasks of MiniLED displays (such as image enhancement, 3D reconstruction, etc.).

[0099] In a preferred embodiment of the present invention, step 12 above involves smoothing the boundary lines between adjacent data items in the photometric labeled dataset based on the differences between the light intensity data items to obtain a first control dataset, including:

[0100] Step 121: Based on the difference in photometric intensity between adjacent data items in the photometric labeled dataset, identify the transition anomaly at the photometric intersection between adjacent data items to obtain the boundary line between adjacent data items;

[0101] Step 122, the adjacent data item represents the adjacent backlight blocks after being centrally marked in the photometric label dataset; the boundary line represents the common boundary line between adjacent backlight blocks, that is, the backlight block dividing line in the photometric data subset;

[0102] Step 123: The boundary line includes two dimensions, vertical and horizontal, to obtain the transition of luminance intensity of various boundary lines. When the luminance intensity transition between two adjacent luminance marker data items is drastic and sudden; the drastic change in luminance intensity transition is not smooth, causing a halo effect on the screen display.

[0103] Step 124: Smooth the light intensity of the drastic transition to obtain the first regulation dataset.

[0104] In this embodiment of the invention, in step 121, it is first necessary to calculate the photometric difference between adjacent data items in the photometric labeled dataset; adjacent data items refer to backlight blocks that have been labeled with illumination features. The photometric data subset has been divided into many small backlight blocks through the previous steps, and each backlight block has its corresponding illumination intensity label (such as highlight, low light, shadow, etc.); when the photometric difference between adjacent backlight blocks is too large (i.e., the illumination intensity changes abnormally), this photometric difference is considered a "transition anomaly"; the detection of transition anomalies can be achieved by calculating the photometric difference between adjacent backlight blocks, and the places where the photometric change changes abruptly are marked as "adjacent boundaries"; this helps to identify transition anomaly regions in the image caused by abrupt changes in photometric intensity, especially those boundaries that are visually prone to producing unnatural transitions; in image processing, identifying photometric transition anomalies helps with subsequent photometric smoothing operations, thereby reducing unsuitable visual effects (such as unnatural brightness jumps).

[0105] In step 122, adjacent data items refer to the marked backlight blocks in the photometric label dataset obtained in step 121. These backlight blocks are adjacent and have obvious differences in photometric characteristics. By comparing the photometric changes between adjacent backlight blocks, their boundaries are determined and these boundaries are defined as "neighborhood lines." Neighborhood lines are the dividing lines between adjacent backlight blocks, marking the intersection of two regions with different illumination intensities. Typically, these boundary lines appear as dividing lines in the image. Marking neighborhood lines helps to further analyze the boundary regions of illumination changes, especially focusing on places where photometric changes are abrupt. Determining the location of neighborhood lines provides a clear reference for subsequent photometric smoothing processing and visual effect optimization, which helps to reduce hard boundaries and abrupt changes in the image caused by illumination differences.

[0106] Step 123 further analyzes the performance of the boundary line in different dimensions based on step 122. The boundary line may appear horizontally (left-right) or vertically (up-down), meaning the direction of the luminance change can be horizontal or vertical. Analyzing the luminance intensity changes of the boundary line in different dimensions (vertical or horizontal) can help identify areas with drastic luminance transitions. Especially when the luminance change between adjacent backlight blocks is very drastic, it may lead to unsmooth visual effects (such as a halo effect on the screen). This step, by quantifying the transition of luminance changes, can detect areas of luminance change with excessively drastic changes. By considering the luminance transition in different dimensions (vertical and horizontal), a more comprehensive analysis of the luminance change areas in the image can be achieved. If drastic luminance changes are detected, it can help determine whether there are unnatural visual effects (such as halo effects, edge blurring, etc.), providing a basis for subsequent processing.

[0107] Step 124 smooths the regions with abrupt changes in photometric transitions detected in Step 123. The goal of smoothing is to reduce the illumination differences at the abrupt changes, making the illumination transitions in the image more natural and smooth; thus smoothing out areas with overly abrupt changes in photometric values ​​and reducing visual imperfections caused by hard boundaries (such as halo effects); after smoothing, the changes in the image's illumination intensity are smoother and more natural, avoiding overly abrupt transitions; the final result is the "first control dataset," which represents the photometric data after smoothing and can be used for subsequent image optimization or display; through smoothing, visual effect problems caused by overly abrupt changes in photometric intensity (such as hard boundaries, halo effects, etc.) are effectively reduced; this step improves the visual quality of the image, making the illumination transitions in the image smoother and more natural, avoiding abrupt and uncomfortable visual performance; after processing the transition anomalies in the image, it can better adapt to the display device and provide a better user experience, especially in the display effect of high-brightness or low-light areas.

[0108] Steps 121 to 124 describe the process of detecting luminance differences between adjacent backlight blocks and identifying boundary lines, thereby processing regions of abrupt luminance changes and reducing transition anomalies in the image. These steps detect and correct luminance transition anomalies (such as halo effects) to improve image display quality; specifically, by analyzing the luminance transition of boundary lines, regions with drastic luminance changes are identified and processed, and finally, smoothing processes are used to improve illumination transitions in the image, thereby enhancing visual effects.

[0109] In a preferred embodiment of the present invention, step 124 further includes:

[0110] Step 1241: Define adjacent data items I(x,y) and I(x',y') to represent the light intensity data on both sides of the boundary line, respectively; through To obtain the difference in luminous intensity; where, Indicates the difference in photometric intensity;

[0111] Step 1242: Set the light intensity threshold for drastic changes. ,when To determine whether the difference in photometric intensity values ​​would cause a drastic abrupt change in transition; through To smooth out the abrupt changes in luminous intensity; where, G(i,j,σ) represents the smoothed luminous intensity value, G(i,j,σ) represents the smoothing weight coefficient, σ controls the standard deviation of the smoothing degree, and I(x+i,y+j) represents the luminous intensity value in the neighborhood (i,j).

[0112] Step 1243: After the smoothing process is completed, the drastic changes in photometric intensity in the photometric labeled dataset have been effectively suppressed, and the generated smoothed dataset is the first controlled dataset. By comparing two adjacent photometric labeled data items before and after the processing, it is determined whether a halo effect occurs. If not, proceed to the next step; if so, return to the previous step for reprocessing. This is to evaluate the smoothing effect.

[0113] In this embodiment of the invention, step 1241 involves setting adjacent data items I(x,y) and I(x',y'), which represent the luminance intensity data on both sides of the boundary line in the image. This can be understood as selecting two neighboring pixel values ​​as a reference for luminance changes; then, the luminance intensity difference between them is calculated to quantify the change in image luminance. By calculating the luminance intensity difference, it can be determined whether there are large luminance abrupt changes in the image. This difference value serves as a standard for measuring the drasticness of luminance changes, helping us identify possible abnormal changes, such as image noise or edge abrupt changes in the image. When the difference value is too large, it may indicate that the luminance abrupt change is too drastic, thus requiring subsequent smoothing processing.

[0114] In step 1242, a threshold is set to determine whether a sudden change in photometric intensity is "too drastic". When the difference value is greater than this threshold, the change is considered excessively drastic.

[0115] Specifically, based on the human eye's sensitivity to changes in light intensity, people are more sensitive to larger changes (such as 30% or higher) when perceiving changes in light intensity. Therefore, setting a threshold of 20%-30% can effectively avoid unnatural abrupt changes.

[0116] To smooth luminance intensity, a smoothing weight coefficient can be used to control the contribution of each neighboring point to the smoothing. σ represents the standard deviation and determines the degree of smoothing, while I(x+i,y+j) represents the luminance intensity value within the neighborhood. By weighted averaging of the luminance values ​​within the neighborhood, drastic changes in luminance within local areas are smoothed, reducing abrupt transitions. By setting a threshold and performing smoothing, drastic luminance changes in the image can be effectively suppressed, making the image appear smoother and more natural. This is particularly important for eliminating noise and reducing unnecessary halo effects. Appropriate smoothing can avoid overly abrupt luminance changes, which is highly beneficial for tasks such as image processing and edge detection, improving the effectiveness of subsequent processing, such as image segmentation and object detection.

[0117] In step 1243, the presence of a halo effect is first determined by comparing two adjacent photometric data items before and after processing. A halo effect typically refers to an abnormal halo appearing in the transition areas of an image due to over-smoothing or inappropriate smoothing, affecting the visual effect. When a halo effect is detected, it may indicate that the smoothing process is excessive or inappropriate, and the program will return to the previous step for adjustment to optimize the smoothing effect. If no halo effect is detected, the smoothing process is considered successful, and the photometric changes in the image have been effectively controlled. By evaluating the halo effect, damage to image details caused by over-smoothing can be avoided, ensuring that smoothing is only applied to areas of sharp abrupt changes without affecting the smoother parts of the image. This feedback mechanism makes the photometric adjustment process more precise, adaptively adjusting the smoothing intensity to maintain natural transitions in image details and edges.

[0118] In summary, the various steps work together, and the core objective is to identify and eliminate drastic changes or anomalies in image brightness, smoothing brightness intensity to avoid unnatural transitions. The specific working principle can be summarized as follows: determining whether there are drastic brightness changes by calculating the brightness differences between adjacent pixels; mitigating overly drastic brightness changes by setting thresholds and using smoothing algorithms; and evaluating the processing effect to avoid over-smoothing or halo effects, ensuring the visual naturalness and consistency of the image.

[0119] In a preferred embodiment of the present invention, step 12 above involves adjusting the boundary lines between adjacent data items in the photometric marker dataset based on the color difference between the photometric chromaticity data to obtain a second control dataset, including:

[0120] Step 125: According to the color gradation of the color space, there are three hues; based on the three hues, the two adjacent data items in the photometric label dataset are represented as (R,G,B) and (r,g,b) respectively.

[0121] Step 126, based on the three hues, by... Obtain the hue difference metric value between two adjacent photometric data items; preset color difference threshold. ,when This indicates a significant color difference.

[0122] In the formula, ΔE represents the measure of hue difference, R and r represent the lightness of two adjacent photometric data items, G and g represent the green-red components of the color, and B and b represent the blue-yellow components.

[0123] Step 127 involves analyzing the hue difference metric between two adjacent photometric data items to finely adjust the light scheduling at the boundary line position, thereby eliminating abrupt transitions in color difference between adjacent photometric data items.

[0124] In this embodiment of the invention, step 125 first defines three hues to represent the colors in the image. These three hues are typically based on the RGB (Red, Green, Blue) color model, specifically referring to:

[0125] R and r: The "luminance" component of adjacent photometric data items (usually related to brightness and changes in brightness).

[0126] G and g: Green-red components, representing the relative intensity of green and red.

[0127] B and b: Blue-yellow components, representing the relative intensity of blue and yellow.

[0128] The two adjacent photometric data items are represented as (R,G,B) and (r,g,b), respectively, where lowercase letters represent the corresponding hue values ​​of adjacent data items. Through this hue division, the color information of the image can be decomposed into lightness and two color components (green-red, blue-yellow), thereby achieving more detailed color analysis. This provides more accurate color information, which is helpful for subsequent detection and adjustment of color abrupt changes.

[0129] In step 126, the hue difference between two adjacent photometric data items is obtained, typically represented by ΔE in the color space. ΔE is a standardized color difference measure, commonly used to quantify the difference between two colors; ΔE is used to measure the color difference between (R,G,B) and (r,g,b); where R, G, and B represent the lightness and color components of one photometric data item, and r, g, and b represent the corresponding values ​​of another adjacent data item; a preset color difference threshold is set. When the color difference between two adjacent photometric data items exceeds this threshold, it indicates that the hue difference is too large, which may lead to abrupt color changes or unnatural transitions.

[0130] Specifically, ΔE<1: the difference is usually indistinguishable to the naked eye, with almost no difference.

[0131] ΔE=1-2: This is usually considered a very subtle difference, perceptible to the naked eye, but the change is very small.

[0132] ΔE=2-5: This is a change that is easily noticeable to the naked eye and is suitable for general visual applications.

[0133] ΔE=5-10: Obvious color difference, very easy to perceive with the naked eye.

[0134] ΔE>10: The color change is very significant, which usually gives people an unnatural or abrupt feeling.

[0135] By measuring color difference, abrupt color changes or unnatural transitions in an image can be effectively detected, especially in areas with large color differences (such as light and shadow boundaries, color block dividing lines, etc.). The color difference threshold provides an adjustable standard, allowing the system to flexibly decide when to make further tone adjustments, thereby avoiding unnecessary transitions or color distortions.

[0136] In step 127, when the hue difference metric (ΔE) between two adjacent photometric data items exceeds a preset color difference threshold, the system will reduce the color difference at the transition point by finely adjusting the position of the boundary line (i.e., the transition area between adjacent data items). By adjusting the boundary line, the transition area can be smoothed, thus mitigating the abrupt color difference between adjacent photometric data items. The purpose of this is to eliminate the abrupt transition of color difference, making the image's color transition more natural; thereby eliminating the hue abrupt change caused by excessive color difference between photometric data items, making the color transition of the image smoother and more natural, avoiding visual discomfort; fine light scheduling adjustment can improve the detail performance of color transition areas in the image, especially in areas with complex color transitions at the edges. By adjusting the position of the boundary line, the color transition is made softer, avoiding color breaks or unnatural light and shadow changes; by controlling the color difference between adjacent data items, the overall visual effect of the image can be optimized, especially in image processing (such as image fusion and image enhancement), achieving better color balance and aesthetics.

[0137] The common goal of these three steps is to finely adjust the color transitions between photometric data items in the image by analyzing the tonal differences in the color space, thus avoiding overly drastic color abrupt changes.

[0138] In a preferred embodiment of the present invention, step 127 further includes:

[0139] Step 1271, when When α is close to 1, it indicates a natural transition in the hue difference measure; when When α is close to 0, it indicates a sharp abrupt change in the hue difference measurement, requiring precision adjustment; where α represents the precision adjustment weighting factor.

[0140] Step 1272, based on the brightness intensity between two adjacent photometric data items, through

[0141] To adjust the precision of the hue difference measure between two adjacent photometric data items; where, This indicates the adjustment precision value for the hue difference metric. α is adjusted according to the magnitude of the color difference ΔE to eliminate abrupt color difference transitions.

[0142] Step 1273, after precision adjustment, to ensure a natural transition in hue differences between adjacent data items in the photometric labeled dataset, ultimately generating the second control dataset; through To evaluate the effect of the accuracy adjustment; where w a ΔE represents the precision-adjusted evaluation value, and β represents the evaluation factor used to assess the sensitivity of weight adjustments. The larger ΔE is, the higher the w... a The smaller the value, the stronger the precision adjustment processing in that area.

[0143] In this embodiment of the invention, in step 1271, α is a precision adjustment weighting factor, which is used to adjust the transition effect of the hue difference measurement. When α is close to 1, it means that the hue transition process is very natural, that is, the difference between hues is small and the transition is smooth. When α is close to 0, the transition of the hue difference measurement becomes abrupt and drastic, which usually means that the difference between hues is large and the transition is abrupt.

[0144] By adjusting the value of α, flexible control over the measurement of tonal differences can be achieved, especially in controlling the smoothness of the transition. Specifically, the larger the α value, the more natural the transition; the smaller the α value, the more abrupt the transition. By adjusting the value of α, the transition of the tonal difference measurement can be made more natural or more refined as needed. For areas with large tonal differences, the color difference can be refined by adjusting α to a smaller value, making the transition more in line with visual perception, thereby improving the accuracy of the tonal transition.

[0145] In step 1272, the precision is adjusted based on the hue difference metric between two adjacent photometric data items, combined with the α value. Color difference ΔE is a standard for measuring the difference between two colors; generally, a larger ΔE value indicates a more pronounced hue difference. The purpose of the adjustment is to modify the precision of the hue difference metric so that the hue difference between adjacent data items better matches the expected visual effect. For example, when the ΔE value is large, the α value is adjusted to be smaller, thereby reducing abrupt changes in color difference transitions and making the transition smoother. By adjusting the precision of the hue difference metric, the transition of color differences can be effectively controlled, making previously abrupt color differences more natural and reducing visual discomfort. This helps to achieve smoother and more consistent hue transitions in applications such as data processing or image processing, improving the user experience.

[0146] In step 1273, a second control dataset is generated after precision adjustment. During this process, the tonal differences between adjacent data items are effectively adjusted, achieving a natural tonal transition. Simultaneously, by evaluating the precision adjustment effect, the adjustment strategy can be further optimized. The evaluation process relies on the precision adjustment evaluation value (wa) and the evaluation factor (β). The precision adjustment evaluation value wa is inversely proportional to the color difference ΔE; that is, the larger ΔE is, the smaller wa is, indicating that the precision adjustment processing in that area requires stronger intervention. β is a factor used to evaluate the sensitivity of weight adjustment, helping the system determine how the intensity of precision adjustment should be adjusted under different color difference conditions. Through precision evaluation and sensitivity adjustment, the effect of the precision adjustment process is maximized. By applying stronger adjustments to areas with larger ΔE, abrupt visual effects are reduced, improving the overall visual smoothness of the image or dataset. Furthermore, by generating the second control dataset, the precision adjustment strategy can be continuously adjusted and optimized, further improving the quality of tonal transitions and avoiding any unnaturalness or incongruity in the transition.

[0147] By flexibly adjusting the precision adjustment weight factor α, the transition of the hue difference measurement shifts from abrupt to natural and smooth, especially in areas with large color differences; the precision is automatically adjusted according to the magnitude of the color difference ΔE to ensure that the precision adjustment process achieves the best visual effect; through precision adjustment evaluation and sensitivity factor β, the adjustment process is dynamically evaluated and optimized to ensure that ideal hue transition effects can be obtained under various conditions.

[0148] Overall, steps 1271 to 1273, by precisely controlling the adjustment of hue differences, make the hue transition between adjacent photometric marker data items more natural and smooth, and effectively avoid the problems of abrupt color differences or jarring transitions, thereby improving the naturalness and accuracy of the visual effect.

[0149] In a preferred embodiment of the present invention, the photometric control model in step 13 above includes:

[0150] Step 131: Integrate the first regulation dataset and the second regulation dataset into a dataset, generate structured data from the historically acquired dataset, and encode the structured data into sequence data to train a photometric regulation model.

[0151] Step 132: Input the sequence data into the photometric control model; the photometric control model includes an input layer, a first hidden layer, a second hidden layer, a third hidden layer and an output layer, transmit the intermediate representation data of multiple hidden layers to the output layer, and the output layer outputs the photometric control standard recognition result representing the relationship between two adjacent data items in the photometric labeled dataset;

[0152] Step 133: Input two adjacent data items in the photometric labeled dataset into the photometric regulation model to output the newly acquired output layer output representing the photometric regulation prediction result between the next pair of adjacent photometric labeled data items.

[0153] In this embodiment of the invention, step 131 first involves merging the two control datasets (the first control dataset and the second control dataset) to form a more complete dataset. These datasets may contain data from different aspects of photometric control or at different time points. By integrating these data, more comprehensive photometric control information can be obtained. After dataset integration, the data is transformed into structured data. Structured data typically refers to a data format with a clear structure or relationship, which may include photometric values, timestamps, photometric change rates, and other information. This step primarily involves structuring the data, making subsequent processing and modeling more efficient and systematic.

[0154] Encoding as Sequence Data: Converting structured data into sequence data is necessary because photometric regulation models may rely on sequence information (such as time series or dependencies). Sequence data reflects the order and interdependencies between data items, making it suitable for training time-series models (such as recurrent neural networks).

[0155] By integrating two datasets, richer information can be obtained, which helps to improve the generalization ability of the model; structuring and encoding the data into sequences can be conveniently provided to the model for processing, especially suitable for processing sequence or time series tasks, such as the time dependence in photometric regulation.

[0156] In step 132, the encoded sequence data is input into the photometric regulation model. The model may be a multi-layer neural network containing multiple hidden layers (such as a first hidden layer, a second hidden layer, and a third hidden layer) and an output layer.

[0157] Network Structure: The input layer of the photometric modulation model receives sequential data. This data is then passed through multiple hidden layers. Each hidden layer extracts features or patterns from the input data for deeper processing and understanding. Typically, the hidden layers of a neural network use non-linear transformations to enable the model to learn complex patterns.

[0158] Output Layer: Finally, the data processed through multiple hidden layers is passed to the output layer. The output layer generates the final prediction result—the standard recognition result of photometric regulation, that is, the photometric regulation standard between two adjacent data items.

[0159] Multi-layered transmission: The intermediate representation data of the hidden layer is passed to the output layer through fully connected layers or other means. The model extracts and transmits information at multiple levels, gradually optimizing and updating the weights, thereby improving the accuracy of prediction.

[0160] Multiple hidden layers can extract high-order features of data and improve the model's learning ability; temporal dependency modeling: by inputting sequential data, the model can capture the temporal dependency in photometric changes, thereby making more accurate predictions.

[0161] In step 133, the model receives two adjacent data items from the photometric labeled dataset. Each pair of adjacent data items represents the relationship between two photometric points or brightness points, and the model uses these data items as input to make predictions.

[0162] Output prediction results: After the adjacent input data items are processed by various layers (including hidden layers) of the photometric regulation model, the output layer will provide the predicted value for the next photometric point. In other words, the model can not only learn the standard regulation relationship between photometric points, but also predict future photometric values.

[0163] Photometric regulation prediction: The output of the model is the predicted photometric regulation standard, which reflects the adjustment standard or expected change between two adjacent photometric points.

[0164] By predicting adjacent data items, real-time photometric control can be achieved. In practical applications, this is crucial for photometric control systems (such as display brightness and intelligent lighting control); by learning the relationships between adjacent data items through the model, more precise photometric control can be provided, avoiding abrupt or overly drastic brightness changes and improving the user experience.

[0165] like Figure 2 As shown, a control system for a high-contrast MiniLED is provided. The system includes an electronic device comprising a memory, a processor, and a control method program for a high-contrast MiniLED stored in the memory and executable on the processor. When executed by the processor, the high-contrast MiniLED control method program implements the steps of a control method for a high-contrast MiniLED. The system includes:

[0166] Data acquisition module: It is used to acquire luminance data of MiniLED screen display from multiple time frames to obtain a luminance data set; based on the luminance data set, the MiniLED backlight area is divided into multiple sub-regions to obtain a luminance data subset; based on the luminance data subset, the luminance data of multiple sub-regions is further subdivided into multiple small-sized backlight blocks, and the multiple backlight blocks are marked to obtain a luminance labeled dataset;

[0167] The data processing module is used to acquire the luminance intensity data of each data item in the photometric marker dataset to obtain a luminance intensity dataset; to smooth the boundary lines between adjacent data items in the photometric marker dataset according to the differences between the luminance intensity data items to obtain a first control dataset; to acquire the photometric color level data of each data item in the photometric marker dataset to obtain a photometric color level dataset; and to adjust the boundary lines between adjacent data items in the photometric marker dataset according to the color difference between the photometric color level data based on the smoothed photometric marker dataset to obtain a second control dataset.

[0168] Photometric regulation module: It is used to train and construct a photometric regulation model from the photometric labeled dataset, the first regulation dataset, and the second regulation dataset to output a photometric regulation standard recognition result representing the relationship between adjacent data items in the photometric labeled dataset; and to input at least two other data items from the photometric labeled dataset into the photometric regulation model to output a photometric regulation prediction result representing the relationship between other adjacent photometric labeled data items.

[0169] The control and evaluation module is used to coordinate the luminance transition between data items in the luminance-labeled dataset based on the prediction results of the luminance control model, so as to reduce the halo effect of the MiniLED screen display.

[0170] Embodiments of the present invention also provide a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the method described above. All implementations in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0171] When the functions of the above-mentioned modules are implemented as software functional units and used as independent products, they can be stored in a storage device. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.

[0172] Therefore, the object of the present invention can also be achieved by running a program or a set of programs on any computing device. The computing device can be a known general-purpose intelligent device. Therefore, the object of the present invention can also be achieved simply by providing a program product containing program code implementing the method or system. That is, such a program product also constitutes the present invention, and the storage medium storing such a program product also constitutes the present invention. Obviously, the storage medium can be any known storage medium or any storage medium developed in the future. It should also be noted that in the apparatus and method of the present invention, it is obvious that the steps can be decomposed and / or recombined. These decompositions and / or recombinations should be considered equivalent to the present invention. Furthermore, the steps performing the above series of processes can naturally be performed in chronological order as described, but are not necessarily required to be performed in chronological order. Some steps can be performed in parallel or independently of each other.

Claims

1. A control method for a high-contrast MiniLED, characterized in that, include: Collect luminance data from multiple time frames of the MiniLED screen to obtain a luminance data set; Based on the photometric data set, the MiniLED backlight area is divided into multiple sub-regions to obtain a subset of photometric data; Based on the aforementioned photometric data subset, the photometric data of multiple sub-regions are subdivided into multiple small-sized backlight blocks, and the multiple backlight blocks are labeled to obtain a photometric labeled dataset. Obtain the luminance intensity data of each data item in the luminance label dataset to obtain the luminance intensity dataset; The boundary lines between adjacent data items in the photometric labeled dataset are smoothed according to the differences between the light intensity data items to obtain the first control dataset; The photometric color level data of each data item in the photometric marker dataset is obtained to obtain the photometric color level dataset; based on the smoothed photometric marker dataset, the boundary lines between adjacent data items in the photometric marker dataset are adjusted according to the color difference between the photometric color level data to obtain the second control dataset. The photometric control model is constructed by training the photometric labeled dataset, the first control dataset, and the second control dataset to output the photometric control standard recognition result representing the relationship between adjacent data items in the photometric labeled dataset. Input at least two additional data items from the photometric labeled dataset into the photometric regulation model to output a photometric regulation prediction result representing the relationship between other adjacent photometric labeled data items; Based on the prediction results of the photometric control model, the photometric transition between each data item in the photometric labeled dataset is coordinated to reduce the halo effect of the MiniLED screen display. Based on the aforementioned photometric data set, the MiniLED backlight area is divided into multiple sub-regions to obtain photometric data subsets, including: The photometric data set is preprocessed to obtain a preprocessed photometric data set; the photometric data set contains the average brightness value of each pixel in the backlight area. The photometric data items in the preprocessed photometric data set are divided into regions according to their spatial distribution by setting a photometric threshold, thereby dividing the backlight region into multiple sub-regions; the sub-regions are further divided into bright areas and dark areas according to the photometric values. The bright area and the dark area are divided into K initial center points according to the luminance value distribution of each sub-region; each pixel of the sub-region is assigned to the nearest initial center point to obtain the division center point of each cluster; the classification results of the bright area and the dark area are updated based on the division center points until the classification results converge. Based on the classification convergence result, the photometric data set is used to finally classify the bright areas and the dark areas to obtain the photometric data subset.

2. The control method for a high-contrast MiniLED according to claim 1, characterized in that, Based on the aforementioned photometric data subset, the photometric data of multiple sub-regions are further subdivided into multiple small-sized backlight blocks, and these backlight blocks are labeled to obtain a photometric labeled dataset, including: The photometric data subset is divided into several square regions of equal size using a 1x1 sliding window with a step size of 1. Each of the divided square regions is a backlight block; the backlight block covers different photometric data portions of the photometric data subset; the backlight block represents a block-shaped region of uniform size in the bright or dark areas of the sub-region; The average illumination intensity of each backlight block is extracted to measure the degree of luminance variation of the backlight block and to capture the texture features of the luminance variation. Based on the extracted features, each backlight block is labeled to obtain a luminance labeled dataset. The photometric labeling dataset is assigned multiple labels based on the target characteristics of the label; the multiple labels include highlight, low light, shadow, or normal lighting.

3. The control method for a high-contrast MiniLED according to claim 2, characterized in that, The boundary lines between adjacent data items in the photometric labeled dataset are smoothed according to the differences between the light intensity data items to obtain a first controlled dataset, including: Based on the difference in photometric intensity between adjacent data items in the photometric labeled dataset, the transition anomaly at the photometric intersection between adjacent data items is identified to obtain the boundary line between adjacent data items. The adjacent data items represent adjacent backlight blocks after being marked in the photometric label dataset; the boundary line represents the common boundary line between adjacent backlight blocks, that is, the backlight block dividing line in the photometric data subset. The boundary line includes two dimensions, vertical and horizontal, to obtain various transition conditions of luminance intensity of the boundary line. When the luminance intensity transition between two adjacent luminance marker data items is drastic and abrupt, the drastic and abrupt luminance intensity transition is not smooth, causing a halo effect on the screen display. The light intensity of the transitional abrupt change is smoothed to obtain the first regulatory dataset.

4. The control method for a high-contrast MiniLED according to claim 3, characterized in that, The photometric intensity of the abrupt transition is smoothed to obtain the first regulatory dataset, which includes: Let adjacent data items I(x,y) and I(x',y') represent the brightness intensity data on both sides of the boundary line, respectively; through To obtain the difference in luminous intensity; where, Indicates the difference in photometric intensity; Set a light intensity threshold for drastic changes ,when To determine whether the difference in photometric intensity values ​​would cause a drastic abrupt change in transition; through To smooth out the abrupt changes in luminous intensity; where, G(i,j,σ) represents the smoothed luminous intensity value, G(i,j,σ) represents the smoothing weight coefficient, σ controls the standard deviation of the smoothing degree, and I(x+i,y+j) represents the luminous intensity value in the neighborhood (i,j). After the smoothing process is completed, the drastic changes in photometric intensity in the photometric labeled dataset have been effectively suppressed, and the generated smoothed dataset is the first controlled dataset. By comparing two adjacent photometric labeled data items before and after the processing, it is determined whether a halo effect occurs. If not, proceed to the next step; if so, return to the previous step for reprocessing. This is to evaluate the smoothing effect.

5. The control method for a high-contrast MiniLED according to claim 4, characterized in that, The boundary lines between adjacent data items in the photometric labeled dataset are adjusted with precision based on the color difference between the photometric color scale data to obtain a second controlled dataset, including: According to the color gradation of the color space, there are three hues; based on the three hues, two adjacent data items in the photometric labeled dataset are represented as (R,G,B) and (r,g,b), respectively. Based on the three aforementioned hues, through Obtain the hue difference metric value between two adjacent photometric data items; preset color difference threshold. ,when This indicates a significant color difference. In the formula, ΔE represents the measure of hue difference, R and r represent the lightness of two adjacent photometric data items, G and g represent the green-red components of the color, and B and b represent the blue-yellow components. By analyzing the hue difference metric between two adjacent photometric data items, the light scheduling at the boundary line position is finely adjusted to eliminate abrupt transitions in color difference between adjacent photometric data items.

6. The control method for a high-contrast MiniLED according to claim 5, characterized in that, By analyzing the hue difference metric between two adjacent photometric data items, the light scheduling at the boundary line position is finely adjusted to eliminate abrupt color difference transitions between adjacent photometric data items, including: when When α is close to 1, it indicates a natural transition in the hue difference measure; when When α is close to 0, it indicates a sharp abrupt change in the hue difference measurement, requiring precision adjustment; where α represents the precision adjustment weighting factor. Based on the brightness intensity between two adjacent photometric data items, through To adjust the precision of the hue difference measure between two adjacent photometric data items; where, This indicates the adjustment precision value for the hue difference metric. α is adjusted according to the magnitude of the color difference ΔE to eliminate abrupt color difference transitions. After precision adjustment, the hue differences between adjacent data items in the photometric labeled dataset are made to transition naturally, ultimately generating a second controlled dataset; through... To evaluate the effect of the accuracy adjustment; where w a ΔE represents the precision-adjusted evaluation value, and β represents the evaluation factor used to assess the sensitivity of weight adjustments. The larger ΔE is, the higher the w... a The smaller the value, the stronger the precision adjustment processing in that area.

7. The control method for a high-contrast MiniLED according to claim 6, characterized in that, The photometric control model includes: The first and second regulation datasets are integrated into a dataset, and the historically acquired dataset is used to generate structured data. The structured data is then encoded into sequence data to train a photometric regulation model. The sequence data is input into the photometric control model; the photometric control model includes an input layer, a first hidden layer, a second hidden layer, a third hidden layer and an output layer, and the intermediate representation data of multiple hidden layers are transmitted to the output layer. The output layer outputs the photometric control standard recognition result representing the relationship between two adjacent data items in the photometric labeled dataset. Two adjacent data items in the photometric labeled dataset are input into the photometric regulation model to output the newly acquired output layer output, which represents the photometric regulation prediction result between the next pair of adjacent photometric labeled data items.

8. A control system for a high-contrast MiniLED, characterized in that, The system provides an electronic device including a memory, a processor, and a control method program for a high-contrast MiniLED stored in the memory and executable on the processor. When executed by the processor, the high-contrast MiniLED control method program implements the steps of the control method for a high-contrast MiniLED as described in any one of claims 1-7. The system includes: Data acquisition module: It is used to acquire luminance data of MiniLED screen display from multiple time frames to obtain a luminance data set; based on the luminance data set, the MiniLED backlight area is divided into multiple sub-regions to obtain a luminance data subset; based on the luminance data subset, the luminance data of multiple sub-regions is further subdivided into multiple small-sized backlight blocks, and the multiple backlight blocks are marked to obtain a luminance labeled dataset; The data processing module is used to acquire the luminance intensity data of each data item in the photometric marker dataset to obtain a luminance intensity dataset; to smooth the boundary lines between adjacent data items in the photometric marker dataset according to the differences between the luminance intensity data items to obtain a first control dataset; to acquire the photometric color level data of each data item in the photometric marker dataset to obtain a photometric color level dataset; and to adjust the boundary lines between adjacent data items in the photometric marker dataset according to the color difference between the photometric color level data based on the smoothed photometric marker dataset to obtain a second control dataset. Photometric regulation module: It is used to train and construct a photometric regulation model by using the photometric labeled dataset, the first regulation dataset, and the second regulation dataset to output a photometric regulation standard recognition result representing the relationship between adjacent data items in the photometric labeled dataset; and to input at least two other data items in the photometric labeled dataset into the photometric regulation model to output a photometric regulation prediction result representing the relationship between other adjacent photometric labeled data items. The control and evaluation module is used to coordinate the luminance transition between data items in the luminance-labeled dataset based on the prediction results of the luminance control model, so as to reduce the halo effect of the MiniLED screen display.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program that, when executed by a processor, implements the control method steps of a high-contrast MiniLED as described in any one of claims 1-7.

Images