Micro-led based high dynamic range partition independent driving system

Abstract

This high dynamic range (HDR) partitioned independent driving system based on micro-LEDs includes a main control interface module, a macro partitioning analysis module, and a drive output module. The main control interface module decapsulates, linearizes, and converts the input image signal to a linear gamut, outputting 16-bit linear RGB data. The macro partitioning analysis module extracts multi-dimensional features of gradients, semantics, and motion vectors, calculates partition granularity decision values, adaptively segments and merges sub-partitions using a quadtree approach, completes blurred transitions at partition boundaries, and dynamically reconstructs dimming modes, refresh rates, and overdrive parameters at granular levels. The drive output module establishes address mappings between sub-partitions and macro / micro partitioning drive units, synchronously distributes parameters, and performs execution. This invention dynamically adapts partitioning and driving strategies based on the image content, improving HDR display levels and dynamic clarity, suppressing abrupt partition boundaries and motion blur, while optimizing power consumption and display uniformity, thus meeting the high dynamic range display requirements of micro-LEDs.

Description

Technical Field

[0001] This invention relates to the field of micro LED display technology, specifically to a high dynamic range zoned independent driving system based on micro LEDs. Background Technology

[0002] As a core direction of next-generation display technology, micro-LEDs, with their advantages brought by inorganic materials, such as ultra-long lifespan (over 100,000 hours), extremely high brightness (peak brightness exceeding 400,000 nits), ultra-high contrast ratio (near-infinite contrast ratio), nanosecond-level response speed, and wide color gamut (covering 120% NTSC color gamut), are gradually becoming a research hotspot and development trend in the high-end display field. High dynamic range (HDR) display technology can accurately reproduce the details of light and dark in real-world scenes, perfectly presenting the depth of dark fields and the clarity of bright fields. Combined with the performance advantages of micro-LEDs, it can achieve display effects far exceeding those of traditional LCD, OLED, and MiniLED, and is one of the core development directions for future high-end displays.

[0003] However, current HDR driving systems based on micro-LEDs still face many technical bottlenecks, which severely restrict their commercialization and performance improvement:

[0004] (1) Low partition driving accuracy and severe optical crosstalk: Most existing micro LED driving systems adopt a fixed partition mode, and the partition size cannot be dynamically adjusted according to the image content. This results in insufficient partition accuracy in areas with rich image details, while partitioning in areas with uniform image brightness is too cumbersome. This not only wastes driving resources but also easily generates optical crosstalk between adjacent partitions, resulting in phenomena such as halos and ghosting, which affect the delicacy of HDR display. At the same time, as the size of micro LED chips shrinks to the micrometer level, it is difficult to guarantee the consistency of tens of millions of chips, which further exacerbates the optical crosstalk problem.

[0005] (2) Poor HDR display effect and limited brightness adjustment range: Existing driving systems mostly use a single dimming method (such as simple PWM dimming or DC dimming), which makes it difficult to balance stable output in the high brightness area and fine adjustment in the low brightness area. This results in insufficient HDR dynamic range, loss of details in dark areas and overexposure in bright areas, failing to fully utilize the brightness advantage of micro LEDs. In addition, the red light chip in the RGB three-color chip integration scheme has low efficiency, resulting in high system power consumption and poor brightness uniformity, which further affects the HDR display effect.

[0006] (3) High system power consumption and high heat dissipation pressure: The pixel density of the micro LED array is extremely high. The traditional driving system adopts a unified power supply and driving mode. Even in low brightness display scenarios, all driving channels are still in a high load state, resulting in high system power consumption. At the same time, the driving chip is not designed to withstand high voltage and low power consumption. Combined with the poor driving matching of glass substrate or flexible substrate, it further aggravates the heat dissipation pressure and affects the long-term stability and service life of the system. Summary of the Invention

[0007] To solve the above-mentioned technical problems, the purpose of this invention is to provide a high dynamic range partitioned independent driving system based on micro LEDs, including a main control interface module, which is communicatively connected to a macro partitioning analysis module and a drive output module.

[0008] The main control interface module is used to receive externally input HDR image signals, preprocess the HDR image signals, generate 16-bit linear RGB data, and send it to the macro partitioning analysis module.

[0009] The macro-partitioning analysis module is used to extract multi-dimensional image features from 16-bit linear RGB data, generate partition granularity decision values ​​based on the extracted multi-dimensional image features, perform adaptive partitioning and merging based on the partition granularity decision values, output sub-partition-driver address mapping results, and perform fuzzy transition processing on partition boundaries. Subsequently, dynamic reconstruction of driver parameters is performed based on partition granularity.

[0010] The drive output module is connected to the micro LED display array for mapping partition drive parameters and synchronous execution with hardware based on the output results of the macro partition analysis module.

[0011] Furthermore, the micro LED display array is composed of several RGB three-color micro LED light-emitting units arranged in a matrix. Based on the display resolution requirements, the micro LED display array is divided into several initial macro partitions. Each macro partition includes several micro partitions. Each macro partition corresponds to a macro partition driving unit, and each micro partition corresponds to a micro partition driving unit. The driving output module communicates with all macro partition driving units through a global high-speed bus. Each macro partition driving unit communicates with all micro partition driving units within its corresponding macro partition through a local area bus.

[0012] Furthermore, the preprocessing of HDR image signals includes:

[0013] The externally input HDR image signal is de-encapsulated and decompressed to obtain non-linear coded pixel data, luminance mapping information and color gamut information. Based on the luminance mapping information, the non-linear coded pixel data is linearized to generate 16-bit linear luminance data. The 16-bit linear luminance data is preprocessed for noise suppression. Then, based on the color gamut information, the 16-bit linear luminance data is color gamut converted to generate 16-bit linear RGB data.

[0014] Furthermore, the process of multi-dimensional image feature extraction from 16-bit linear RGB data includes:

[0015] The 16-bit linear RGB data is divided into macro partitions for image segmentation. Convolution analysis is performed on the 16-bit linear RGB data corresponding to each macro partition to obtain the horizontal and vertical gradients of each pixel. The gradient magnitude is obtained based on the horizontal and vertical gradients. An initial gradient threshold combination is randomly set. Based on the gradient threshold combination, the pixels are divided into three categories. The number of pixels at each gray level of the 16-bit linear RGB data is counted to obtain a histogram. The histogram is normalized to obtain a normalized histogram. Pixel proportion analysis is performed based on the normalized histogram to obtain the probability coefficients of each category. Based on the weight coefficients of each category and the gradient magnitude of each pixel, the optimal threshold combination is obtained.

[0016] Based on the optimal threshold combination, all pixels within the macro partition are divided into three gradient levels, and quantization rules corresponding to different gradient levels are set.

[0017] Furthermore, the process of multi-dimensional image feature extraction from 16-bit linear RGB data also includes:

[0018] Construct a semantic recognition model by inputting 16-bit linear RGB data within the macro partition into the semantic recognition model, and setting the priority and quantization rules for different semantic features based on the semantic features output by the semantic recognition model in the 16-bit linear RGB data.

[0019] For the macro range to which a pixel belongs, the inter-frame block matching algorithm is used to obtain the inter-frame horizontal motion vector and vertical motion vector. The motion magnitude is obtained based on the horizontal and vertical motion vectors. The motion magnitude is then adaptively normalized to obtain the motion vector features.

[0020] The partitioning granularity decision value for each pixel is obtained based on the gradient level, semantic features, and motion vector features of each pixel.

[0021] Furthermore, the process of adaptive partitioning and merging based on partition granularity decision values, and outputting sub-partition-driver address mapping results, includes:

[0022] Mark each macro partition as the root node, and perform the following steps in parallel on all macro partitions:

[0023] Step 1: Obtain the maximum and average values ​​of the partition granularity decision values ​​for all pixels within the root node. If the maximum value within the root node is greater than the preset high segmentation threshold, or the average value is greater than the preset medium segmentation threshold, then divide the root node into four identical sub-partitions along the horizontal and vertical midlines in a 2x2 column manner.

[0024] Step 2: Obtain the maximum and average values ​​of the partition granularity decision values ​​of all pixels in the sub-partition. If the maximum value in the sub-partition is greater than the preset high segmentation threshold, or the average value is greater than the preset middle segmentation threshold, then divide the sub-partition into four identical sub-partitions along the horizontal and vertical midlines of the sub-partition in a 2-row × 2-column manner.

[0025] Step 3: Repeat Step 2 for the sub-partition until the sub-partition size reaches 1×1 pixels, or the maximum value in the sub-partition is less than or equal to the preset high segmentation threshold and the average value is less than or equal to the preset middle segmentation threshold, then mark the sub-partition as a leaf node.

[0026] Step 4: Traverse all leaf nodes. If there are four adjacent leaf nodes of the same size, determine whether the decision values ​​of all pixels of the four adjacent leaf nodes of the same size are lower than the preset merging threshold. If they are lower, merge the four leaf nodes of the same size into one leaf node.

[0027] Step 5: Repeat Step 4 for the leaf nodes until the merged leaf nodes have reached the maximum size of the macro partition, or the decision values ​​of all pixels of four adjacent leaf nodes of the same size are lower than the preset merging threshold, and output the sub-partition-drive address mapping result.

[0028] Furthermore, the process of blurring the boundary between partitions includes:

[0029] Identify the boundary lines between adjacent sub-partitions in the macro partition, perform luminance analysis on the 16-bit linear RGB data of two adjacent sub-partitions, obtain the core representative luminance of the two adjacent sub-partitions, obtain the linear light domain luminance ratio R based on the core representative luminance of the two adjacent sub-partitions, obtain the total width of the transition band of the two adjacent sub-partitions based on the linear light domain luminance ratio and the size of the two adjacent sub-partitions, mark the transition band in the two adjacent sub-partitions according to the total width of the transition band, and perform smooth adjustment on the luminance of each pixel in the transition band to generate the target luminance of each pixel in the transition band.

[0030] Furthermore, the process of dynamically reconstructing driving parameters based on partition granularity includes:

[0031] Based on the size range of each sub-partition, all sub-partitions are divided into five granularity levels, including L0, L1, L2, L3, and L4.

[0032] When the granularity level of the sub-partition is L0 / L1, the analog DC dimming mode is used in the full grayscale range of the sub-partition, and the linearity lower limit of the analog dimming mode is set. When the grayscale value of the sub-partition is lower than the linearity lower limit, the SS-PWM dimming mode is used in the sub-partition to lock the refresh rate of the sub-partition to the highest refresh rate of the hardware. Then, the overdrive parameters of the sub-partition are dynamically reconstructed to generate the overdrive parameters of the sub-partition.

[0033] When the granularity level of the sub-partition is L2, the simulated DC dimming mode is used in the full grayscale range of the sub-partition. When the grayscale value of the sub-partition is lower than the linearity lower limit, the PWM dimming mode is used in the sub-partition. The maximum motion vector feature of the sub-partition is obtained, and the refresh rate of the sub-partition is adaptively matched based on the maximum motion vector feature. The relative brightness difference between frames is obtained according to the core representative brightness of the current frame and the previous frame of the sub-partition. The overdrive parameters are dynamically reconstructed for pixels in the sub-partition whose relative brightness difference between frames is greater than a preset threshold, and the overdrive parameters of the sub-partition are generated. For pixels in the sub-partition whose relative brightness difference between frames is less than or equal to the preset threshold, the overdrive function is turned off.

[0034] When the granularity level of the sub-partition is L3 / L4, the simulated DC dimming mode is used in the full grayscale range of the sub-partition, the refresh rate of the sub-partition is locked to the lowest hardware refresh rate, and the overdrive function of the sub-partition is disabled.

[0035] Furthermore, the process of dynamically reconstructing the driving parameters includes:

[0036] Based on the core representative brightness of the current frame and the previous frame of the sub-partition, the inter-frame brightness difference is obtained, and the sub-partition temperature is obtained. Based on the target brightness of each pixel in the current frame sub-partition, the factory-calibrated photoelectric characteristic LUT table, and the sub-partition temperature, the rated driving voltage of each pixel is obtained, and an overdrive fitting model is constructed. The rated driving voltage, target brightness, inter-frame brightness difference, and sub-partition temperature of each pixel are input into the overdrive fitting model, and the overdrive parameters of the sub-partition are generated based on the overdrive fitting model.

[0037] Furthermore, the process of mapping partition drive parameters and synchronizing execution with hardware includes:

[0038] Based on the sub-partition-drive address mapping results, the pixel range corresponding to each sub-partition is mapped to the physical address of the macro-partition drive unit and the micro-partition drive unit to generate an address mapping table. Based on the target brightness, dimming mode, refresh rate and overdrive parameters of each sub-partition, drive parameters for each sub-partition are generated. The drive parameters of each sub-partition and the address mapping table are synchronously sent to the corresponding macro-partition drive unit. The macro-partition drive unit distributes the drive parameters to the corresponding micro-partition drive unit according to the address mapping table.

[0039] Compared with the prior art, the beneficial effects of the present invention are:

[0040] 1. Based on a multi-dimensional image feature-driven adaptive partitioning mechanism, the partitioning granularity can dynamically adjust to perfectly match the image content. For core areas rich in detail and of high visual interest, it automatically adopts pixel-level minimum partitioning granularity, ensuring ultimate driving control precision. Flat and uniform non-focused areas are automatically merged into larger partitions, avoiding the detail blurring and insufficient contrast issues caused by mismatched partitioning granularity with image content in fixed partitioning schemes. Combined with blurred transition processing at partition boundaries, it completely eliminates display defects inherent in traditional partitioning drives, such as hard-edge bright lines, halos, and ghosting, achieving a natural and smooth transition between different brightness areas with no visible partitioning traces. Simultaneously, based on the refined reconstruction of driving parameters according to partitioning granularity, it matches exclusive grayscale mapping and dimming strategies for different areas, achieving seamless detail restoration across the entire brightness range from pure black to high brightness. The image's transparency, layering, and three-dimensionality are fundamentally improved compared to traditional solutions, accurately reproducing the creative intent of HDR content.

[0041] 2. This invention integrates image motion characteristics into the entire process of partitioning decision-making and driving parameter reconstruction. It can accurately identify motion areas and the intensity of motion in the image, matching differentiated refresh timing and overdrive strategies to different areas, thus solving the core pain points of traditional globally fixed refresh and globally unified overdrive solutions. For high-speed motion core detail areas, it automatically matches the hardware's highest refresh rate and a dedicated strong overdrive strategy, compressing the response speed of the light-emitting units to the extreme, completely eliminating ghosting, ghosting, and edge blurring problems in high-speed motion images. For transition areas, overdrive compensation is only enabled for pixels with significant brightness jumps, ensuring no visible ghosting while avoiding image whitening, brightness overshoot, and color distortion caused by global overdrive. Unnecessary overdrive functions are turned off in static, flat areas, eliminating image noise caused by ineffective switching actions. Whether it's high-speed motion game images, rapidly scrolling text content, or high frame rate dynamic videos, it can always maintain the ultimate dynamic display effect of sharp edges, smooth images, and color fidelity.

[0042] 3. This invention solves the problem of wasted power consumption common in traditional fixed-partition and globally unified driving schemes by dynamically segmenting and merging partitions based on content adaptation, fundamentally addressing the issue at the source of the driving architecture. Flat, detail-less background areas are automatically merged into large partitions, matched with a low-power pure DC dimming mode and a low refresh rate strategy. Areas that have not been illuminated for extended periods are placed into deep sleep mode, significantly reducing switching losses, static power consumption, and wasted refresh power consumption in the driving circuit. For core detail areas, image quality accuracy is prioritized, with high-frequency driving and overcompensation functions only activated when necessary, avoiding the additional power consumption caused by global high-frequency driving and global overdriving. Attached Figure Description

[0043] Figure 1 This is a flowchart of a high dynamic range partitioned independent driving system based on micro LEDs, according to an embodiment of this application. Detailed Implementation

[0044] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0045] like Figure 1 As shown, the high dynamic range partitioned independent driving system based on micro LEDs includes a main control interface module, which is communicatively connected to the macro partitioning analysis module and the drive output module.

[0046] The main control interface module is used to receive externally input HDR image signals, preprocess the HDR image signals, generate 16-bit linear RGB data, and send it to the macro partitioning analysis module.

[0047] The macro-partitioning analysis module is used to extract multi-dimensional image features from 16-bit linear RGB data, generate partition granularity decision values ​​based on the extracted multi-dimensional image features, perform adaptive partitioning and merging based on the partition granularity decision values, output sub-partition-driver address mapping results, and perform fuzzy transition processing on partition boundaries. Subsequently, dynamic reconstruction of driver parameters is performed based on partition granularity.

[0048] The drive output module is connected to the micro LED display array for mapping partition drive parameters and synchronous execution with hardware based on the output results of the macro partition analysis module.

[0049] It should be further explained that, in the specific implementation process, the micro LED display array is composed of several RGB three-color micro LED light-emitting units arranged in a matrix. Based on the display resolution requirements of the micro LED display array, the micro LED display array is divided into several initial macro partitions. Each macro partition includes several micro partitions. Each macro partition corresponds to a top-level macro partition driver unit (with a built-in ARM Cortex-M4 core and hardware decoding unit, supporting screen data processing up to 240Hz). Each micro partition corresponds to a bottom-level pixel-level micro partition driver unit (integrated on the TFT backplane of the display array). The driver output module communicates with all macro partition driver units through a global high-speed bus. Each macro partition driver unit communicates with all micro partition driver units within its corresponding macro partition through a local area bus.

[0050] Macro partitioning: The entire micro LED display array is divided into M×N independent macro partitions, where M and N are positive integers and can be flexibly adjusted according to the display resolution; for example, a 4K resolution display array can be divided into 48×27=1296 macro partitions, and an 8K resolution display array can be divided into 96×54=5184 macro partitions.

[0051] Micro-partitioning: Each macro partition contains K×L pixel-level micro-partitions, where K and L are positive integers. The number of micro-partitions corresponding to a single macro partition is K×L≥1000. Each micro-partition corresponds to at least one MicroLED light-emitting unit. The total number of micro-partitions in the system can reach tens of millions, enabling independent driving of the entire pixel array.

[0052] It should be further explained that, in the specific implementation process, the preprocessing of HDR image signals includes:

[0053] The externally input HDR image signal is decapsulated and decompressed to obtain the decapsulated nonlinear coded pixel data, luminance mapping information (EOTF / OTF curves), and color gamut information. Based on the luminance mapping information, the decapsulated nonlinear coded pixel data is linearized to generate 16-bit linear luminance data. Noise suppression preprocessing is then applied to the 16-bit linear luminance data. Since the pixel value range of the 16-bit linear data is 0~65535, which is much larger than the 0~255 range of the 8-bit nonlinear data, the luminance / color differences between pixels are more significant, allowing the filtering algorithm to more accurately identify noisy pixels and true edge pixels. Therefore, an adaptive median filtering algorithm is used to suppress noise in the 16-bit linear data. The process involves removing Gaussian noise and salt-and-pepper noise generated during signal transmission while preserving image edge details to avoid image blurring. Specifically, by detecting differences in grayscale values ​​of image pixels, the size of the filtering window (3×3~7×7) is dynamically adjusted. Noisy pixels are adaptively replaced, while edge pixels are preserved to ensure the integrity of image details. Subsequently, based on color gamut information, the 16-bit linear luminance data undergoes color gamut conversion. Using a 3×3 color gamut conversion matrix and a lookup table method, a precise mapping is completed from the source color gamut (sRGB / DCI-P3 / BT.2020) to the target wide color gamut (≥120%NTSC) of the micro-LED, generating 16-bit linear RGB data with complete color gamut conversion.

[0054] It should be further explained that, in the specific implementation process, the process of multi-dimensional image feature extraction from 16-bit linear RGB data includes:

[0055] The 16-bit linear RGB data is divided into macro partitions to segment the image. Convolutional analysis is performed on the corresponding 16-bit linear RGB data of each macro partition to obtain the horizontal gradient Gx and vertical gradient of each pixel. The gradient magnitude G is then obtained based on the horizontal and vertical gradients. Initial gradient threshold combinations T1 and T2 are randomly set. Based on these combinations, pixels are divided into three categories: low gradient region (G < T1), medium gradient region (T1 ≤ G ≤ T2), and high gradient region (G > T2). The number of pixels at each gray level (0~65535) of the 16-bit linear RGB data is counted to obtain the histogram. Normalize the histogram to obtain a normalized histogram. ;in, This represents the total number of pixels. Given the number of pixels at the k-th grayscale value, pixel proportion analysis is performed based on the normalized histogram to obtain probability coefficients for various intervals, with the probability coefficients of the low gradient region being the most significant factor. For example, Where t is the number of gray levels in the low gradient region, the optimal threshold combination is obtained based on the weight coefficients of each type of interval and the gradient magnitude of each pixel. and Specifically:

[0056] ;

[0057] in, , , These are the mean gradient magnitudes for the three types of intervals. This represents the overall mean of the gradient magnitudes for the macro partition. ;

[0058] Iterate through all possible threshold combinations (T1, T2) and calculate the corresponding... Select the corresponding The largest threshold combination (T1, T2) is marked as the optimal threshold combination. and .

[0059] Based on the optimal threshold combination, all pixels within the macro partition are divided into three gradient levels: high gradient region (G > high threshold). ): Edges, detailed core areas; Medium gradient areas (medium threshold) ≤G≤High threshold ): Transitional texture region, medium partitioning granularity; low gradient region (G < low threshold) For flat and uniform regions, large partitions are merged first, and quantization rules corresponding to different gradient levels are set, specifically:

[0060] When pixel (x, y) belongs to the high gradient region, then ;

[0061] When pixel (x, y) belongs to the medium gradient region, then ;

[0062] When pixel (x, y) belongs to the low gradient region, then ;

[0063] in, Let the gradient magnitude of the pixel (x,y) be divided. Let (x, y) be the gradient magnitude of pixel (x, y). The maximum gradient magnitude within the macro partition. This represents the minimum gradient magnitude within the macro partition.

[0064] It should be further explained that, in the specific implementation process, the multi-dimensional image feature extraction process for 16-bit linear RGB data also includes:

[0065] A lightweight edge-side CNN model (MobileNetV2-YOLO) is used to construct a semantic recognition model. 16-bit linear RGB data within a macro-partition is input into the semantic recognition model. The model performs real-time semantic segmentation on the 16-bit linear RGB data and outputs the semantic features. Priorities and quantization rules are set for different semantic features, specifically as follows:

[0066] ;

[0067] in, Score the semantic features of the quantized pixel (x,y);

[0068] For the macro region to which a pixel belongs, the inter-frame block matching algorithm is used to obtain the inter-frame horizontal motion vector vx and vertical motion vector vy, and the motion magnitude is obtained based on the horizontal motion vector vx and vertical motion vector vy. The motion modulus is adaptively normalized to obtain motion vector features, specifically:

[0069] ;

[0070] in, The motion vector features of pixel (x,y) The preset maximum motion threshold is 10% of the number of pixels on the shorter side of the image.

[0071] Each pixel is obtained based on its gradient level, semantic features, and motion vector features. Partition granularity decision value , By fusing the three types of features according to their weights, a partition granularity decision value of 0 to 100 is generated for each pixel. The higher the value, the smaller the required partition granularity and the higher the driving accuracy. For example: 0: a completely flat, plain background with no details and no motion (such as a blue sky); 100: the highest priority text, human eyes, and edges of fast-moving objects.

[0072] It should be further explained that, in the specific implementation process, the process of adaptive partitioning and merging based on partition granularity decision values, and outputting the sub-partition-driver address mapping results, includes:

[0073] Mark each macro partition as the root node, and perform the following steps in parallel on all macro partitions:

[0074] Step 1: Obtain the maximum and average values ​​of the partition granularity decision values ​​for all pixels within the root node. If the maximum value within the root node is greater than the preset high segmentation threshold (75 points), or the average value is greater than the preset medium segmentation threshold (50 points), then divide the root node into four identical sub-partitions along the horizontal and vertical midlines of the root node in a 2x2 column manner (i.e., if the macro interval size is 8x8 pixels, then four 4x4 pixel sub-partitions are obtained).

[0075] Step 2: Obtain the maximum and average values ​​of the partition granularity decision values ​​of all pixels in the sub-partition. If the maximum value in the sub-partition is greater than the preset high segmentation threshold (75 points), or the average value is greater than the preset medium segmentation threshold (50 points), then divide the sub-partition into four identical sub-partitions along the horizontal and vertical midlines of the sub-partition in a 2-row × 2-column (four equal parts, i.e., if the macro interval size is 8 × 8 pixels, then four 4 × 4 pixel sub-partitions are obtained).

[0076] Step 3: Repeat Step 2 for the sub-partition until the sub-partition size reaches 1×1 pixels, or the maximum value in the sub-partition is less than or equal to the preset high segmentation threshold and the average value is less than or equal to the preset middle segmentation threshold, then mark the sub-partition as a leaf node.

[0077] Step 4: Traverse all leaf nodes. If there are four adjacent leaf nodes of the same size, determine whether the decision values ​​of all pixels of the four adjacent leaf nodes of the same size are lower than the preset merging threshold (20 points). If they are lower, merge the four leaf nodes of the same size into one leaf node to reduce the number of partitions.

[0078] Step 5: Repeat Step 4 for the leaf nodes until the merged leaf nodes have reached the maximum size of the macro partition, or the decision values ​​of all pixels of four adjacent leaf nodes of the same size are lower than the preset merging threshold. Output the sub-partition-drive address mapping result (i.e., the pixel range corresponding to each sub-partition).

[0079] It should be further explained that, in the specific implementation process, the process of performing fuzzy transition processing on the partition boundaries includes:

[0080] Identify the boundary lines between adjacent sub-partitions in the macro-partition, perform luminance analysis on the 16-bit linear RGB data of two adjacent sub-partitions, and obtain the core representative luminance of the two adjacent sub-partitions. Specifically, extract the core region far from the boundary from each partition. The size of the core region is "indented by 2 pixels" of the partition size (if the partition size is ≤4 pixels, then indented by 1 pixel; if the partition size is ≤2 pixels, then the entire partition is taken). For example, if the partition P size is 80×80 pixels, with coordinates (0,0)-(79,79), then the core region is (2,2)-(77,77). The boundary regions of 2 pixels each on the top, bottom, left, and right are excluded to avoid the influence of boundary noise. Then, the median target luminance of the core region is used as the core representative luminance of the partition. Based on the core representative luminance of the two adjacent sub-partitions, obtain the linear light domain luminance ratio R (the ratio of the core representative luminance of the two adjacent sub-partitions). Based on the linear light domain luminance ratio and the size of the two adjacent sub-partitions, obtain the total width of the transition band of the two adjacent sub-partitions. Specifically:

[0081] The preset brightness ratio levels corresponding to different linear light domain brightness ratios are shown in Table 1 below:

[0082]

[0083] The mapping rules for the total width of the transition band corresponding to different brightness ratio levels are set as shown in Table 2 below:

[0084]

[0085] The total width of the transition band obtained through mapping in Table 2 is corrected for the partition size constraint: the minimum side length of two adjacent partitions is calculated, and the total width of the transition band cannot exceed 1 / 2 of the minimum side length. If it does, the total width of the transition band is corrected to floor(minimum side length / 2) (rounded down). For example, if the size of adjacent partition A is 4×4 pixels and the size of partition B is 8×8 pixels, and the minimum side length is 4 pixels, the maximum total width of the transition band is 2 pixels. Even if the brightness ratio level is 5, it still needs to be corrected to 2 pixels to avoid the transition band filling the small partition.

[0086] The transition zone is marked within two adjacent sub-partitions based on the total width of the transition zone, specifically as follows:

[0087] The corrected total width of the transition band is divided equally between the bright area side (representing high-brightness zones) and the dark area side (representing low-brightness zones):

[0088] If the total width is an even number: the bright area side and the dark area side each occupy 2 pixels of the total width;

[0089] If the total width is an odd number: the bright area occupies 1 more pixel (because the brightness change of the bright area is more sensitive to the human eye, and 1 more pixel can further improve the smoothness of the transition), that is, the bright area occupies (total width + 1) / 2 pixels, and the dark area occupies (total width - 1) / 2 pixels.

[0090] The brightness of each pixel within the transition band is then smoothly adjusted to generate the target brightness for each pixel within the transition band. For example:

[0091] The adjacent A zone (bright area, 10000 nits brightness within the transition band) and B zone (dark area, 0.001 nits brightness within the transition band) have a brightness difference of 10^7 times. The total width of the transition band is 8 pixels (4 pixels on the A side and 4 pixels on the B side). The target brightness gradient for each width of the transition band is smoothly adjusted to:

[0092] Side A (bright area towards the boundary): 8000nit→5000nit→2000nit→500nit;

[0093] Side B (from the boundary to the dark area): 10nit→0.1nit→0.01nit→0.001nit.

[0094] Key advantages: Brightness adjustment is made only within the boundary transition zone, while the brightness of the core area of ​​the partition remains at the target value. This eliminates hard boundaries without sacrificing the HDR accuracy and contrast of the core area.

[0095] The size and number of partitions are dynamically adjusted based on the image feature extraction results: For areas with large brightness gradients and rich details (such as human faces, text, and complex scenes), smaller partitions are divided (the smallest partition can be 1×1 pixel level) to improve partition driving accuracy; for areas with uniform brightness and simple details (such as the sky and solid color backgrounds), larger partitions are divided to reduce the occupation of driving channels, reduce system power consumption, and break the limitations of traditional fixed partitions.

[0096] Subsequently, taking the granularity level of dynamic partitioning as the core anchor point, and combining the characteristics of the image content within the partition, a unique, full-link closed-loop set of driving parameters is generated for each independent dynamic partition. This achieves a one-to-one precise adaptation of "partition granularity - image characteristics - driving parameters," allowing each partition to work in a state of "optimal image quality, lowest power consumption, and strongest stability," fully unleashing the full potential of the dynamic partitioning architecture.

[0097] It should be further explained that, in the specific implementation process, the dynamic reconstruction of driving parameters based on partition granularity includes:

[0098] Based on the size range of each sub-partition, all sub-partitions are divided into five granularity levels, including L0 extreme detail level (1×1~2×2 pixels), L1 detail level (2×2~4×4 pixels), L2 transition level (4×4~8×8 pixels), L3 flat level (8×8~16×16 pixels), and L4 extreme flat level (16×16 pixels and above).

[0099] When the granularity level of the sub-partition is L0 / L1, the simulated DC dimming mode is used in the 0~100% full grayscale range of the sub-partition (the entire brightness adjustment range from pure black to the brightest, 0% grayscale = the screen is completely off (black level), 100% grayscale = the screen's maximum brightness (e.g., 20000 nits)). The linearity lower limit (3%) of the simulated dimming mode is set. When the grayscale value (representing the brightness percentage) of the sub-partition is lower than the linearity lower limit, the SS-PWM dimming mode is used in the sub-partition. The PWM equivalent frequency of the SS-PWM dimming mode is always ≥20000Hz. The refresh rate of the sub-partition is locked to the highest hardware refresh rate (240Hz). Then, the overdrive parameters of the sub-partition are dynamically reconstructed to generate the overdrive parameters of the sub-partition.

[0100] When the granularity level of the sub-partition is L2, a simulated DC dimming mode is used across the entire grayscale range of the sub-partition. When the grayscale value of the sub-partition is lower than the linearity lower limit, a PWM dimming mode is used within the sub-partition, with the PWM equivalent frequency fixed at 480kHz. The maximum motion vector feature of the sub-partition is obtained, and the refresh rate of the sub-partition is adaptively matched based on the maximum motion vector feature (120Hz for maximum motion vector feature greater than 75 points; 60Hz for maximum motion vector feature greater than or equal to 30 points and less than or equal to 75 points; and 30Hz for maximum motion vector feature less than 30 points). The relative brightness difference between frames is obtained based on the core representative brightness of the current frame and the previous frame of the sub-partition. , ,in, The core representative brightness of the current frame and the previous frame, For the core representative brightness of the previous frame, for pixels in the sub-partition whose corresponding inter-frame relative brightness difference is greater than a preset threshold (20%), the overdrive parameters are dynamically reconstructed to generate the overdrive parameters of the sub-partition. For pixels in the sub-partition whose corresponding inter-frame relative brightness difference is less than or equal to the preset threshold (20%), the overdrive function is turned off and traditional fixed constant current drive is used.

[0101] When the granularity level of the sub-partition is L3 / L4, the simulated DC dimming mode is used in the full grayscale range of the sub-partition, the refresh rate of the sub-partition is locked to the lowest hardware refresh rate, and the overdrive function of the sub-partition is disabled.

[0102] It should be further explained that, in the specific implementation process, the dynamic reconfiguration of the driving parameters includes:

[0103] Based on the core representative brightness of the current frame and the previous frame of the sub-partition, obtain the inter-frame brightness difference. Obtain the partition temperature T of the sub-partition, based on the target brightness of each pixel in the current frame sub-partition. Based on the factory-calibrated photoelectric characteristic LUT table and the partition temperature T, obtain the rated drive voltage required for each pixel to maintain the target brightness. The rated driving voltage is the reference value that allows the micro-LED to stably output the target brightness without overdrive, and it is the basis for overdrive calculation. The forward voltage drop and carrier mobility of the micro-LED change significantly with temperature (the forward voltage drop decreases by about 2% for every 10°C increase in temperature). Therefore, the rated parameters are updated based on the temperature sensor data of the macro-region in each frame to avoid overshoot / undershoot caused by temperature drift. An overdrive fitting model is constructed, and the rated driving voltage, target brightness, inter-frame brightness difference and regional temperature of each pixel are input into the overdrive fitting model. The overdrive parameters of the sub-region (rising edge overdrive voltage, rising edge conduction time, falling edge overdrive voltage and falling edge conduction time) are generated according to the overdrive fitting model.

[0104] The core variables of the overdrive fitting model are the inter-frame brightness difference. ( The larger the value, the higher the overdrive voltage multiple and the longer the conduction time. For example, for a 0→100% brightness jump, the overdrive multiple is 8~10 times the rated value; for a 0→50% jump, the multiple is 4~6 times (target brightness). (The lower the L_target, the less charge is required and the shorter the overdrive conduction time; the higher the L_target, the longer the conduction time (but always kept within 1μs)) and the partition temperature T (the higher the temperature, the lower the carrier mobility and the smaller the forward voltage drop, requiring an increase in overdrive voltage and a longer conduction time; the lower the temperature, the lower the overdrive intensity should be to avoid overshoot); the quantization formula is:

[0105] Rising edge overdrive voltage : ;

[0106] Rising edge conduction time : ;

[0107] Falling edge overdrive voltage : ;

[0108] Falling edge conduction time : ;

[0109] in, The target brightness of the sub-region in the previous frame; These are the base overdrive multiples for the rising and falling edges, respectively. The brightness difference adaptation coefficient, and Positive correlation, ranging from 0.2 to 1; This is the temperature compensation coefficient, dynamically adjusted based on real-time temperature, ranging from 0.8 to 1.5. The brightness adaptation coefficient is positively correlated with the target brightness and ranges from 0.1 to 1. The base on-time is fixed at 800 ns to ensure that the total response time does not exceed 1 μs.

[0110] The generated overdrive parameters satisfy the following constraints:

[0111] Total rise time < 1μs (a core requirement for eliminating ghosting).

[0112] When the overdrive ends, the brightness reaches 90%~100% of the target value (no undershoot).

[0113] Throughout the entire frame cycle, the actual brightness does not exceed 1% of the target value (no overshoot, no image distortion).

[0114] Based on the calculated parameters, the micro-partitioning driving unit generates a two-stage overdrive waveform, which is executed entirely by high-speed hardware circuitry, achieving timing accuracy up to 10ns.

[0115] Overdrive phase (0~T_od): At the start of frame refresh, a calculated overdrive voltage V_od is immediately applied to the micro LED, and a corresponding large current is output to quickly complete capacitor charging and discharging. At the end of this phase, the brightness of the micro LED must reach more than 90% of the target value;

[0116] Constant current maintenance phase (T_od ~ frame end): After the overdrive time ends, the overdrive circuit is immediately cut off, and the system seamlessly switches to the rated voltage V_norm to maintain the brightness of the micro LED at the target value and avoid brightness overshoot.

[0117] It should be further explained that, in the specific implementation process, the process of partition driver parameter mapping and hardware synchronization includes:

[0118] Based on the sub-partition-drive address mapping results, the pixel range corresponding to each sub-partition is mapped to the physical address of the macro-partition drive unit and the micro-partition drive unit to generate an address mapping table. Based on the target brightness, dimming mode, refresh rate and overdrive parameters of each sub-partition, drive parameters for each sub-partition are generated. The drive parameters of each sub-partition and the address mapping table are synchronously sent to the corresponding macro-partition drive unit. The macro-partition drive unit distributes the drive parameters to the corresponding micro-partition drive unit according to the address mapping table.

[0119] For large merged partitions: the addresses of all micro-partition drive units within the partition are mapped to the same main drive channel, and the signals of the main drive channel are synchronously sent to all micro-partitions, realizing single-channel control of multiple pixels and reducing drive channel occupancy;

[0120] For a 1×1 small partition: each micro partition driver unit has an independent address mapping and receives driver signals independently, achieving pixel-level precise control.

[0121] It should be further noted that the configuration distribution and parameter updates of all macro partitions are completed within the frame blanking period and take effect synchronously when the next frame is refreshed. The timing synchronization error of the entire array is <1μs, and there are no screen tearing or desynchronization issues.

[0122] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A high dynamic range zoned independent driving system based on micro LEDs, characterized in that, This includes a main control interface module, which communicates with the macro partitioning analysis module and the drive output module. The main control interface module is used to receive externally input HDR image signals, preprocess the HDR image signals, generate 16-bit linear RGB data, and send it to the macro partitioning analysis module. The macro-partitioning analysis module is used to extract multi-dimensional image features from 16-bit linear RGB data, generate partition granularity decision values ​​based on the extracted multi-dimensional image features, perform adaptive partitioning and merging based on the partition granularity decision values, output sub-partition-driver address mapping results, and perform fuzzy transition processing on partition boundaries. Subsequently, dynamic reconstruction of driver parameters is performed based on partition granularity. The drive output module is connected to the micro LED display array for mapping partition drive parameters and synchronous execution with hardware based on the output results of the macro partition analysis module.

2. The high dynamic range zoned independent driving system based on micro-LEDs according to claim 1, characterized in that, The micro LED display array is composed of several RGB three-color micro LED light-emitting units arranged in a matrix. Based on the display resolution requirements, the micro LED display array is divided into several initial macro partitions. Each macro partition includes several micro partitions. Each macro partition corresponds to a macro partition driving unit, and each micro partition corresponds to a micro partition driving unit. The driving output module communicates with all macro partition driving units through a global high-speed bus. Each macro partition driving unit communicates with all micro partition driving units within its corresponding macro partition through a local area bus.

3. The high dynamic range zoned independent driving system based on micro LEDs according to claim 2, characterized in that, The preprocessing process for HDR image signals includes: The externally input HDR image signal is de-encapsulated and decompressed to obtain non-linear coded pixel data, luminance mapping information and color gamut information. Based on the luminance mapping information, the non-linear coded pixel data is linearized to generate 16-bit linear luminance data. The 16-bit linear luminance data is preprocessed for noise suppression. Then, based on the color gamut information, the 16-bit linear luminance data is color gamut converted to generate 16-bit linear RGB data.

4. The high dynamic range zoned independent driving system based on micro LEDs according to claim 3, characterized in that, The process of multi-dimensional image feature extraction from 16-bit linear RGB data includes: The 16-bit linear RGB data is divided into macro partitions for image segmentation. Convolution analysis is performed on the 16-bit linear RGB data corresponding to each macro partition to obtain the horizontal and vertical gradients of each pixel. The gradient magnitude is obtained based on the horizontal and vertical gradients. An initial gradient threshold combination is randomly set. Based on the gradient threshold combination, the pixels are divided into three categories. The number of pixels at each gray level of the 16-bit linear RGB data is counted to obtain a histogram. The histogram is normalized to obtain a normalized histogram. Pixel proportion analysis is performed based on the normalized histogram to obtain the probability coefficients of each category. Based on the weight coefficients of each category and the gradient magnitude of each pixel, the optimal threshold combination is obtained. Based on the optimal threshold combination, all pixels within the macro partition are divided into three gradient levels, and quantization rules corresponding to different gradient levels are set.

5. The high dynamic range zoned independent driving system based on micro LEDs according to claim 4, characterized in that, The process of multi-dimensional image feature extraction from 16-bit linear RGB data also includes: Construct a semantic recognition model by inputting 16-bit linear RGB data within the macro partition into the semantic recognition model, and setting the priority and quantization rules for different semantic features based on the semantic features output by the semantic recognition model in the 16-bit linear RGB data. For the macro range to which a pixel belongs, the inter-frame block matching algorithm is used to obtain the inter-frame horizontal motion vector and vertical motion vector. The motion magnitude is obtained based on the horizontal and vertical motion vectors. The motion magnitude is then adaptively normalized to obtain the motion vector features. The partitioning granularity decision value for each pixel is obtained based on the gradient level, semantic features, and motion vector features of each pixel.

6. The high dynamic range zoned independent driving system based on micro-LEDs according to claim 5, characterized in that, The process of adaptive partitioning and merging based on partition granularity decision values, and outputting sub-partition-driver address mapping results, includes: Mark each macro partition as the root node, and perform the following steps in parallel on all macro partitions: Step 1: Obtain the maximum and average values ​​of the partition granularity decision values ​​for all pixels within the root node. If the maximum value within the root node is greater than the preset high segmentation threshold, or the average value is greater than the preset medium segmentation threshold, then divide the root node into four identical sub-partitions along the horizontal and vertical midlines in a 2x2 column manner. Step 2: Obtain the maximum and average values ​​of the partition granularity decision values ​​of all pixels in the sub-partition. If the maximum value in the sub-partition is greater than the preset high segmentation threshold, or the average value is greater than the preset middle segmentation threshold, then divide the sub-partition into four identical sub-partitions along the horizontal and vertical midlines of the sub-partition in a 2-row × 2-column manner. Step 3: Repeat Step 2 for the sub-partition until the sub-partition size reaches 1×1 pixels, or the maximum value in the sub-partition is less than or equal to the preset high segmentation threshold and the average value is less than or equal to the preset middle segmentation threshold, then mark the sub-partition as a leaf node. Step 4: Traverse all leaf nodes. If there are four adjacent leaf nodes of the same size, determine whether the decision values ​​of all pixels of the four adjacent leaf nodes of the same size are lower than the preset merging threshold. If they are lower, merge the four leaf nodes of the same size into one leaf node. Step 5: Repeat Step 4 for the leaf nodes until the merged leaf nodes reach the maximum size of the macro partition, or the decision values ​​of all pixels of four adjacent leaf nodes of the same size are lower than the preset merging threshold, and output the sub-partition-drive address mapping result.

7. The high dynamic range zoned independent driving system based on micro-LEDs according to claim 6, characterized in that, The process of blurring the boundary of a partition includes: Identify the boundary lines between adjacent sub-partitions in the macro partition, perform luminance analysis on the 16-bit linear RGB data of two adjacent sub-partitions, obtain the core representative luminance of the two adjacent sub-partitions, obtain the linear light domain luminance ratio R based on the core representative luminance of the two adjacent sub-partitions, obtain the total width of the transition band of the two adjacent sub-partitions based on the linear light domain luminance ratio and the size of the two adjacent sub-partitions, mark the transition band in the two adjacent sub-partitions according to the total width of the transition band, and perform smooth adjustment on the luminance of each pixel in the transition band to generate the target luminance of each pixel in the transition band.

8. The high dynamic range zoned independent driving system based on micro LEDs according to claim 7, characterized in that, The process of dynamically reconstructing drive parameters based on partition granularity includes: Based on the size range of each sub-partition, all sub-partitions are divided into five granularity levels, including L0, L1, L2, L3, and L4. When the granularity level of the sub-partition is L0 / L1, the analog DC dimming mode is used in the full grayscale range of the sub-partition, and the linearity lower limit of the analog dimming mode is set. When the grayscale value of the sub-partition is lower than the linearity lower limit, the SS-PWM dimming mode is used in the sub-partition to lock the refresh rate of the sub-partition to the highest refresh rate of the hardware. Then, the overdrive parameters of the sub-partition are dynamically reconstructed to generate the overdrive parameters of the sub-partition. When the granularity level of the sub-partition is L2, the simulated DC dimming mode is used in the full grayscale range of the sub-partition. When the grayscale value of the sub-partition is lower than the linearity lower limit, the PWM dimming mode is used in the sub-partition. The maximum motion vector feature of the sub-partition is obtained, and the refresh rate of the sub-partition is adaptively matched based on the maximum motion vector feature. The relative brightness difference between frames is obtained according to the core representative brightness of the current frame and the previous frame of the sub-partition. The overdrive parameters are dynamically reconstructed for pixels in the sub-partition whose relative brightness difference between frames is greater than a preset threshold, and the overdrive parameters of the sub-partition are generated. For pixels in the sub-partition whose relative brightness difference between frames is less than or equal to the preset threshold, the overdrive function is turned off. When the granularity level of the sub-partition is L3 / L4, the simulated DC dimming mode is used in the full grayscale range of the sub-partition, the refresh rate of the sub-partition is locked to the minimum hardware refresh rate, and the overdrive function of the sub-partition is disabled.

9. The high dynamic range zoned independent driving system based on micro-LEDs according to claim 8, characterized in that, The process of dynamically reconstructing the driving parameters includes: Based on the core representative brightness of the current frame and the previous frame of the sub-partition, the inter-frame brightness difference is obtained, and the sub-partition temperature is obtained. Based on the target brightness of each pixel in the current frame sub-partition, the factory-calibrated photoelectric characteristic LUT table, and the sub-partition temperature, the rated driving voltage of each pixel is obtained, and an overdrive fitting model is constructed. The rated driving voltage, target brightness, inter-frame brightness difference, and sub-partition temperature of each pixel are input into the overdrive fitting model, and the overdrive parameters of the sub-partition are generated based on the overdrive fitting model.

10. The high dynamic range zoned independent driving system based on micro LEDs according to claim 9, characterized in that, The process of partition driver parameter mapping and hardware synchronization includes: Based on the sub-partition-drive address mapping results, the pixel range corresponding to each sub-partition is mapped to the physical address of the macro-partition drive unit and the micro-partition drive unit to generate an address mapping table. Based on the target brightness, dimming mode, refresh rate and overdrive parameters of each sub-partition, drive parameters for each sub-partition are generated. The drive parameters of each sub-partition and the address mapping table are synchronously sent to the corresponding macro-partition drive unit. The macro-partition drive unit distributes the drive parameters to the corresponding micro-partition drive unit according to the address mapping table.

Images