Liquid crystal screen power consumption optimization method and system based on picture content analysis

Abstract

The application relates to a liquid crystal screen power consumption optimization method and system based on picture content analysis, and belongs to the technical field of liquid crystal screen power consumption optimization. The application predicts a region with high visual saliency by using an eye fixation prediction model, initializes refresh rates of all regions in a liquid crystal screen according to a motion vector field analysis result, and finally optimizes the refresh rates of all regions in the liquid crystal screen by using the region with high visual saliency. The application can effectively distinguish global motion and local motion, effectively avoid the phenomenon that local small-range motion triggers global high refresh rate and the phenomenon that the refresh rate reduction is insufficient or excessive under a global uniform motion scene, thereby optimizing the power consumption of the liquid crystal screen through the method, breaking through the technical limitation of traditional global refresh rate control, realizing region-level differentiated refresh rate control through fine analysis of a motion vector field, realizing more fine-grained power consumption optimization, and reducing the use power consumption of liquid crystal screen display driving.

Description

Technical Field

[0001] This invention relates to the field of liquid crystal display technology, and in particular to a method and system for optimizing the power consumption of liquid crystal displays based on image content analysis. Background Technology

[0002] Liquid crystal displays (LCDs), as a mainstream flat panel display technology, are widely used in smartphones, televisions, computer monitors, automotive displays, and medical devices. Their basic working principle involves using an external electric field to control the orientation and alignment of liquid crystal molecules, thereby adjusting the intensity of light emitted from the backlight passing through a polarizer and the liquid crystal layer, thus achieving image display with different gray levels and colors.

[0003] Since the advent of the self-contained liquid crystal display (LCD), liquid crystal display technology has undergone a major revolution, evolving from passive matrix driving to active matrix driving. The introduction of thin-film transistor (TFT) active driving technology completely solved the problems of crosstalk and limited scan lines, making dynamic display of high resolution and large capacity information possible. Subsequently, coplanar conversion technology and vertical alignment technology, designed for wide viewing angles, have matured and significantly improved color shift and contrast degradation at different viewing angles, meeting the stringent image quality requirements of professional image processing and high-end consumer electronics.

[0004] Existing refresh rate adjustment strategies typically rely solely on simple time thresholds (such as delaying the refresh rate after detecting a static image), failing to proactively identify the motion characteristics of the screen content and the user's interaction status, resulting in coarse adjustment granularity. Summary of the Invention

[0005] This invention overcomes the shortcomings of the prior art and provides a method and system for optimizing the power consumption of LCD screens based on screen content analysis.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: The first aspect of this invention provides a method for optimizing the power consumption of a liquid crystal display (LCD) screen based on image content analysis, comprising: Image data is continuously extracted from the LCD screen, and motion vector field data in the image data is obtained by extracting the motion vector field from the image data. By performing spatial clustering analysis on the motion vector field data in the image data, and using the motion vector field analysis results, a human eye gaze prediction model is constructed based on deep learning. The human eye gaze prediction model is used to predict areas with high visual saliency, and the refresh rate of each area within the LCD screen is initialized based on the motion vector field analysis results. The refresh rate of each area within the LCD screen is optimized using areas with high visual salience.

[0007] Furthermore, in the LCD screen power consumption optimization method based on image content analysis, image data is continuously extracted from the LCD screen, and motion vector field data in the image data is obtained by extracting the motion vector field from the image data, specifically as follows: Image data of the current frame and the previous frame are simultaneously obtained from the frame buffer of the display driver IC in a dual-port parallel reading mode. The motion complexity index of the previous frame image is obtained, and the read image data is adaptively downsampled according to the motion complexity index of the previous frame image. The downsampled image is converted from the RGB color space to the YCbCr color space, and the luminance component is extracted for motion estimation. The luminance image of the current frame is recursively divided into quadtrees based on the gradient complexity within the block. Calculate the gradient magnitude variance of each candidate block. If the gradient magnitude variance is greater than a preset gradient magnitude threshold, then the corresponding candidate block is further divided into four sub-blocks. This process is repeated recursively until the minimum block size is reached. If the gradient magnitude variance is not greater than the preset gradient magnitude threshold, the partitioning is stopped; using the motion vector field calculated in the previous frame, a non-uniform block grid is obtained according to the partitioning result, and motion vector spatial domain prediction and temporal prediction are performed on each block in the current frame.

[0008] Furthermore, the LCD screen power consumption optimization method based on image content analysis also includes: In the spatial domain prediction, the median of the motion vectors of the adjacent left, top, and right top blocks of the current block is taken as the prediction value. In the temporal domain prediction, linear extrapolation is performed based on the motion vectors of the blocks with the same position as the current block in the previous frame, combined with the global motion trend. Calculate the predicted residual energy. If the residual energy is less than the residual energy threshold, the predicted motion vector is directly used as the motion vector of the current block, and the motion search of the current block is skipped. For blocks that cannot be reused by time prediction, a range search is initiated. During the range search, the motion vector magnitude of the block corresponding to the previous frame and the region where the block is located are used as the target range for the search. After the motion search is completed, if the matching cost is lower than a preset multiple of the optimal cost per integer pixel, subpixel interpolation and thinning are performed to obtain the subpixel precision motion vector. Median filtering and consistency verification are then performed on the motion vector field to obtain the motion vector field data in the image data.

[0009] Furthermore, in the LCD screen power consumption optimization method based on image content analysis, spatial clustering analysis is performed on the motion vector field data in the image data. The specific results of the motion vector field analysis are as follows: Calculate the amplitude features, orientation angle features, amplitude time stability features, orientation change rate features, and local motion consistency features of each block in the motion vector field data of the image data, and construct a multidimensional motion feature vector for each block; Based on the multidimensional motion feature vector of each block, the motion vector amplitude of all blocks in the whole picture is statistically analyzed, a histogram is generated, and the amplitude corresponding to the first local minimum point of the histogram is used as the low speed threshold. The second local minimum point of the histogram is found as the high-speed threshold. At the same time, motion noise basis estimation is introduced, and the amplitude corresponding to the histogram that is lower than the preset quantile threshold is used as the noise threshold. Based on the multidimensional motion feature vector of each block and the noise threshold, spatial clustering analysis is performed on each block to generate static regions, low-speed motion regions and high-speed motion regions. Based on the static region, low-speed motion region, and high-speed motion region, a motion vector field analysis result is generated, and the motion vector field analysis result is output.

[0010] Furthermore, in the LCD screen power consumption optimization method based on image content analysis, a human eye gaze prediction model is constructed based on deep learning, specifically as follows: Historical gaze center point data of static areas, low-speed motion areas, and high-speed motion areas are collected. A human eye gaze prediction model is constructed based on deep learning, and the historical gaze center point data of static areas, low-speed motion areas, and high-speed motion areas are used as model input. The probability of each pixel belonging to the human eye's gaze point is used as the model output, and the human eye gaze prediction model is used to train the training data. The human eye gaze prediction model is trained when its prediction accuracy is greater than a preset prediction accuracy threshold.

[0011] Furthermore, in the LCD screen power consumption optimization method based on image content analysis, the human eye gaze prediction model is used to predict areas with high visual saliency, specifically as follows: Collect image data of the screen content within a preset time period, and input the image data of the screen content within the preset time period into the human eye gaze prediction model for prediction. By prediction, the probability that each pixel in the image data of the LCD screen's content within a preset time period belongs to the human eye's gaze point is obtained, and the probability that each pixel in the image data of the LCD screen's content within the preset time period belongs to the human eye's gaze point is calculated to generate a probability map. The locations of regions in the probability map with a probability greater than a preset probability threshold are obtained, and regions with high visual saliency are generated based on the regions in the probability map with a probability greater than the preset probability threshold.

[0012] Furthermore, in the LCD screen power consumption optimization method based on image content analysis, the refresh rate of each area within the LCD screen is initialized according to the motion vector field analysis results, specifically as follows: Based on the motion vector field analysis results, static region, low-speed motion region, and high-speed motion region are obtained, and refresh rates are configured for static region, low-speed motion region, and high-speed motion region. Specifically, static areas perform intra-frame partial refresh, low-speed motion areas perform frequency reduction refresh, and high-speed motion areas maintain full frame rate refresh.

[0013] Furthermore, in the LCD screen power consumption optimization method based on image content analysis, the refresh rate of each area within the LCD screen is optimized for areas with high visual salience, specifically as follows: Determine whether there are visually salient regions in static, low-speed motion, and high-speed motion regions; When there are areas with high visual saliency in the static area, low-speed motion area, and high-speed motion area, increase the refresh rate of the corresponding area. When there are no visually significant areas in the static area, low-speed motion area, and high-speed motion area, the current refresh rate remains unchanged.

[0014] Furthermore, the LCD screen power consumption optimization method based on image content analysis also includes: The hash fingerprint comparison algorithm is used to calculate two consecutive frames of image data to obtain a perceptual hash fingerprint. Based on the perceptual hash fingerprint, it is determined whether the similarity between the two frames is greater than a preset similarity threshold. When the similarity between two frames determined by the perceptual hash fingerprint is greater than the preset similarity threshold, the reading operation of the frame buffer is stopped, and the driving signal of the previous frame is directly reused. When the change in the two frames of the perceptual hash fingerprint is detected to be significantly lower than the preset range threshold, the driver IC only performs data reading and data line driving on the pixel blocks that have changed, and reuses the driving state of the previous frame for the unchanged areas. If there are discrepancies in the hash fingerprint determination, the frame is further divided into several pixel blocks, the difference metric of each block is calculated, and the location and range of the changed pixel blocks are identified.

[0015] A second aspect of the present invention provides a liquid crystal display (LCD) power consumption optimization system based on screen content analysis, including a memory and a processor. The memory includes a program for a liquid crystal display power consumption optimization method based on screen content analysis. When the processor executes the program for the liquid crystal display power consumption optimization method based on screen content analysis, it implements the steps of any of the liquid crystal display power consumption optimization methods based on screen content analysis.

[0016] This invention addresses the shortcomings of the prior art and has the following beneficial effects: This invention extracts image data continuously from an LCD screen and obtains motion vector field data by extracting motion vector fields from the image data. Then, it performs spatial clustering analysis on the motion vector field data. Based on the motion vector field analysis results, it constructs a human eye gaze prediction model using deep learning. This model predicts regions with high visual salience and initializes the refresh rate of each region within the LCD screen based on the motion vector field analysis results. Finally, it optimizes the refresh rate of each region within the LCD screen using these highly visually salient regions. This invention effectively distinguishes between global and local motion, effectively avoiding the phenomenon of small-scale local motion triggering a global high refresh rate and insufficient or excessive refresh rate reduction in globally uniform motion scenarios. This method optimizes the power consumption of the LCD screen. Furthermore, it overcomes the limitations of traditional global refresh rate control by achieving region-level differentiated refresh rate control through refined motion vector field analysis, enabling finer-grained power consumption optimization and reducing the power consumption of the LCD screen display driver. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained from these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating the overall process of optimizing LCD screen power consumption based on screen content analysis is provided. Figure 2 A system block diagram of an LCD screen power consumption optimization system based on screen content analysis is shown. Detailed Implementation

[0019] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0020] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0021] The first aspect of this invention provides a method for optimizing the power consumption of a liquid crystal display (LCD) screen based on image content analysis, comprising: Image data is continuously extracted from the LCD screen, and motion vector field data in the image data is obtained by extracting the motion vector field from the image data. By performing spatial clustering analysis on motion vector field data in image data, and using the motion vector field analysis results, a human eye gaze prediction model is constructed based on deep learning. The human eye gaze prediction model is used to predict areas with high visual saliency, and the refresh rate of each area in the LCD screen is initialized based on the motion vector field analysis results. The refresh rate of each area within the LCD screen is optimized by utilizing areas with high visual salience.

[0022] Furthermore, in the LCD screen power consumption optimization method based on image content analysis, image data is continuously extracted from the LCD screen, and motion vector field data in the image data is obtained by extracting the motion vector field from the image data, specifically: Image data of the current frame and the previous frame are simultaneously obtained from the frame buffer of the display driver IC in a dual-port parallel reading mode. The motion complexity index of the previous frame image is obtained, and the read image data is adaptively downsampled according to the motion complexity index of the previous frame image. For example, to reduce the computational burden of subsequent motion estimation, adaptive downsampling is performed on the read image data: the downsampling factor α is dynamically determined based on the motion complexity index of the previous frame, such as the average motion vector magnitude. If the motion complexity is low, such as an average magnitude < 1 pixel / frame, then α = 4, downsampling to 1 / 4 of the size; if the motion complexity is medium, α = 2; if the motion complexity is high, α = 1, and no downsampling is performed. Bilinear interpolation is used for downsampling to preserve edge information.

[0023] The downsampled image is converted from the RGB color space to the YCbCr color space, and the luminance component is extracted for motion estimation. The luminance image of the current frame is recursively divided into quadtrees based on the gradient complexity within the block. The downsampled image is converted from the RGB color space to the YCbCr color space, and only the Y (luminance) component is extracted for motion estimation. The luminance component contains the main components of motion information, and ignoring the chrominance component reduces computational cost. The conversion process uses fixed-point integer arithmetic to avoid floating-point overhead.

[0024] Calculate the gradient magnitude variance of each candidate block. If the gradient magnitude variance is greater than the preset gradient magnitude threshold, then divide the corresponding candidate block into four sub-blocks. Recursively proceed until the minimum block size is reached. If the gradient magnitude variance is not greater than the preset gradient magnitude threshold, the partitioning is stopped; using the motion vector field calculated in the previous frame, a non-uniform block mesh is obtained according to the partitioning result, and motion vector spatial domain prediction and temporal prediction are performed on each block in the current frame.

[0025] It should be noted that this partitioning result is a non-uniform block mesh: small-sized blocks are used for motion boundaries and detailed regions to accurately capture motion; large-sized blocks are used for smooth regions to reduce computational cost. Compared with fixed-size blocks, this method can reduce the number of blocks while maintaining the accuracy of motion estimation.

[0026] Furthermore, the LCD screen power consumption optimization method based on image content analysis also includes: In spatial prediction, the median of the motion vectors of the adjacent left, top, and right top blocks of the current block is taken as the prediction value. In temporal prediction, linear extrapolation is performed based on the motion vectors of the blocks with the same position as the current block in the previous frame, combined with the global motion trend. Calculate the predicted residual energy. If the residual energy is less than the residual energy threshold, the predicted motion vector is directly used as the motion vector of the current block, and the motion search of the current block is skipped. It should be noted that the global motion trend is the camera motion parameters. If the residual energy is less than the residual energy threshold, the predicted motion vector is directly used as the motion vector of the current block, and the motion search of the current block is skipped. This mechanism utilizes time correlation to skip the motion search of some blocks, thereby reducing computational power consumption.

[0027] For blocks that cannot be reused by time prediction, a range search is initiated. During the range search, the motion vector magnitude of the block corresponding to the previous frame and the region where the block is located are used as the target range for the search. Specifically, if the motion vector magnitude of the block in the previous frame is large, the super-resolution SR is increased; if the motion vector magnitude of the block in the previous frame is small, the super-resolution SR is smaller. If the block belongs to a texture-rich region with high gradient variance, the super-resolution SR is appropriately increased to capture fine motion; if it belongs to a smooth region, the super-resolution SR is decreased to avoid mismatches.

[0028] After the motion search is completed, if the matching cost is lower than a preset multiple of the optimal cost per integer pixel, subpixel interpolation and thinning are performed to obtain the subpixel precision motion vector. Median filtering and consistency verification are then performed on the motion vector field to obtain the motion vector field data in the image data.

[0029] It should be noted that after completing the integer-pixel motion search, if the matching cost is less than 1.1 times the optimal integer-pixel cost, sub-pixel interpolation and thinning are performed. A reference block with 1 / 2 pixel precision is obtained using 2×2 bilinear interpolation, sub-pixel SAD is calculated, and the optimal sub-pixel displacement is selected. The sub-pixel search is performed only on the eight and a half neighboring pixels of the optimal integer-pixel point to avoid computational explosion during full sub-pixel search. Spatial median filtering (3×3 filtering window) is applied to the extracted motion vector field to remove isolated outlier vectors. The median filtering hardware implementation is simple, requiring only a comparator and a multiplexer.

[0030] Perform forward and backward consistency checks: For each block, calculate the forward motion vector from frame N-1 to frame N and the backward motion vector from frame N to frame N-1. If the difference between the two is greater than a threshold (e.g., 2 pixels), the block is marked as an "unreliable block," and its motion vector is replaced by a weighted interpolation of the motion vectors of surrounding reliable blocks. This step effectively eliminates erroneous motion vectors in occluded areas.

[0031] Furthermore, in the LCD screen power consumption optimization method based on image content analysis, spatial clustering analysis is performed on the motion vector field data in the image data. The specific results of the motion vector field analysis are as follows: Calculate the amplitude features, orientation angle features, amplitude temporal stability features, orientation rate of change features, and local motion consistency features of each block in the motion vector field data of the image data, and construct a multidimensional motion feature vector for each block; Based on the multidimensional motion feature vector of each block, the motion vector amplitude of all blocks in the whole picture is counted, a histogram is generated, and the amplitude corresponding to the first local minimum point of the histogram is used as the low speed threshold. The second local minimum point of the histogram is found as the high-speed threshold. At the same time, motion noise basis estimation is introduced, and the amplitude corresponding to the histogram that is lower than the preset quantile threshold is used as the noise threshold. For example, the amplitude corresponding to the top 5 percentile of the histogram is taken as the noise threshold, and blocks with amplitudes less than the noise threshold and amplitude time stability characteristics less than the noise fluctuation threshold are marked as static points, thus forming static regions.

[0032] Based on the multidimensional motion feature vector of each block and the noise threshold, spatial clustering analysis is performed on each block to generate static regions, low-speed motion regions and high-speed motion regions. Spatial clustering analysis methods include K-means clustering, density-based clustering methods, etc.

[0033] The motion vector field analysis results are generated based on the static region, the low-speed motion region, and the high-speed motion region, and then output.

[0034] It should be noted that, among them, blocks that satisfy the following conditions are classified as low-speed motion regions: low-speed threshold < current block amplitude characteristic < high-speed threshold, amplitude time stability characteristic less than preset threshold, and local motion consistency characteristic lower than preset threshold; blocks that satisfy the following conditions are classified as high-speed motion regions: current block amplitude characteristic greater than high-speed threshold, amplitude time stability characteristic greater than preset threshold, and local motion consistency characteristic higher than preset threshold.

[0035] Furthermore, in the LCD screen power consumption optimization method based on image content analysis, a human eye gaze prediction model is constructed based on deep learning, specifically as follows: Historical gaze center point data of static areas, low-speed motion areas, and high-speed motion areas are collected. A human eye gaze prediction model is constructed based on deep learning, and the historical gaze center point data of static areas, low-speed motion areas, and high-speed motion areas are used as model input. The probability of each pixel belonging to the human eye's gaze point is used as the model output, and the human eye gaze prediction model is used to train the training data. The human eye gaze prediction model is considered complete when its prediction accuracy exceeds a preset prediction accuracy threshold.

[0036] It should be noted that the parameters in a deep learning network include: The encoder employs a 3-layer depthwise separable convolution, followed by batch normalization and ReLU6 activation at each layer. The input is a downsampled grayscale image, with the resolution reduced to 1 / 4 of the original. For example, for a 1080p input, the encoder processes a 270p image. The number of channels is 16, 24, and 32, respectively. The stride is 2, 1, and 1, respectively. The encoder output feature map size is the same as the original. Figure 1 / 4, number of channels 32.

[0037] Bottleneck layer: A lightweight attention module—Channel-Spatial Joint Attention (CSSE)—with only 1.2k parameters. This module first performs global average pooling to obtain channel weights, then performs 3×3 depthwise convolution to obtain spatial weights, and finally multiplies the two to obtain a weighted feature map.

[0038] Decoder: Two transposed convolutional layers, upsampled by 2x and 2x respectively, each followed by a depthwise separable convolution, with the output channel count initially at 16 and then at 1. The final output is a saliency probability map with the same resolution as the input, representing the probability that each pixel belongs to the human eye's gaze point.

[0039] Output layer: Sigmoid activation, output range [0,1].

[0040] Furthermore, in the LCD screen power consumption optimization method based on image content analysis, a human eye gaze prediction model is used to predict areas with high visual salience, specifically: Collect image data of the LCD screen content within a preset time period, and input the image data of the LCD screen content within the preset time period into the human eye gaze prediction model for prediction. By prediction, the probability that each pixel in the image data of the LCD screen content belongs to the human eye's gaze point within a preset time is obtained, and the probability that each pixel in the image data of the LCD screen content belongs to the human eye's gaze point within the preset time is calculated to generate a probability map. The system identifies regions in the probability map whose probabilities are greater than a preset probability threshold, and generates regions with high visual saliency based on these regions.

[0041] Furthermore, in the LCD screen power consumption optimization method based on image content analysis, the refresh rate of each area within the LCD screen is initialized according to the motion vector field analysis results, specifically as follows: Based on the motion vector field analysis results, static region, low-speed motion region and high-speed motion region are obtained, and refresh rate is configured for static region, low-speed motion region and high-speed motion region; Specifically, static areas perform intra-frame partial refresh, low-speed motion areas perform frequency reduction refresh, and high-speed motion areas maintain full frame rate refresh.

[0042] Furthermore, in the LCD screen power consumption optimization method based on image content analysis, the refresh rate of each area within the LCD screen is optimized based on areas with high visual salience. Specifically: Determine whether there are visually salient regions in static, low-speed motion, and high-speed motion regions; When there are areas with high visual saliency in static areas, low-speed motion areas, and high-speed motion areas, increase the refresh rate of the corresponding areas. When there are no visually significant areas in the static, low-speed motion, and high-speed motion regions, the current refresh rate remains unchanged.

[0043] It should be noted that this method can optimize the refresh rate of each area within the LCD screen by combining areas with high visual salience. This can further optimize power consumption while improving the visual effect and avoiding the optimization of the user's focus area.

[0044] Furthermore, the LCD screen power consumption optimization method based on image content analysis also includes: The hash fingerprint comparison algorithm is used to calculate two consecutive frames of image data to obtain a perceptual hash fingerprint. Based on the perceptual hash fingerprint, it is determined whether the similarity between the two frames is greater than a preset similarity threshold. When the similarity between two frames determined by the perceptual hash fingerprint is greater than the preset similarity threshold, the reading operation of the frame buffer is stopped, and the driving signal of the previous frame is directly reused. When the change in the two frames of the perceptual hash fingerprint is detected to be significantly lower than the preset range threshold, the driver IC only performs data reading and data line driving on the pixel blocks that have changed, and reuses the driving state of the previous frame for the unchanged areas. If there are discrepancies in the hash fingerprint determination, the frame is further divided into several pixel blocks, the difference metric of each block is calculated, and the location and range of the changed pixel blocks are identified.

[0045] It should be noted that when multiple consecutive frames are detected to be identical (i.e., hash fingerprints match perfectly), the display driver IC enters an ultra-low power mode, stops reading from the frame buffer, and directly reuses the drive signal from the previous frame. Only the basic refresh timing of the panel needs to be maintained, such as a global refresh every 200ms to prevent liquid crystal polarization. When the detected change range is small, i.e., the proportion of changed pixels is less than a threshold, such as 15%, the driver IC only performs data reading and data line driving on the changed pixels, reusing the drive state from the previous frame for the unchanged areas. In this mode, the power consumption of the driver IC is proportional to the proportion of the changed area, saving a significant amount of dynamic power consumption compared to the full-frame drive mode. Furthermore, when a regular change in the image is detected, such as a scrolling list or a gradual transition animation, some drive signals can be generated locally, reducing the amount of data read from the frame buffer and further reducing data bandwidth requirements.

[0046] In addition, this method also includes: By configuring a camera device in a preset range area of ​​the LCD screen, the camera device is used to collect image data information of the LCD screen within a preset time period, and feature extraction is performed on the image data information of the LCD screen within the preset time period. By feature extraction, the visual reflective area of ​​the LCD screen during display is obtained, and the position information of each block or pixel is obtained based on the visual reflective area of ​​the LCD screen during display. The refresh rate of the corresponding pixel is reduced based on the position information of each block or the pixel it belongs to; Simultaneously, based on the light reflection characteristics, the visual reflection area of ​​the LCD screen during display is estimated to obtain the light source area, and relevant lighting control prompts are generated according to the light source area.

[0047] It should be noted that, in reality, when affected by external lighting, the LCD screen will be affected by light, resulting in reflective areas. These reflective areas are often not within the area of ​​human visual focus. Therefore, based on the position information of each block or pixel, the refresh rate of the corresponding pixel is reduced. This can be adjusted according to external conditions, and further reduce power consumption based on actual circumstances.

[0048] In addition, this method also includes: The system acquires video frame data information displayed on the LCD screen, classifies scene categories based on the video frame data information displayed on the LCD screen, obtains the scene category classification results, and constructs a scene-power consumption strategy mapping table. For each scenario category, a multi-dimensional power consumption optimization strategy is predefined. The power consumption of the scenario category classification results is optimized using the multi-dimensional power consumption optimization strategy. At the same time, a strategy switching buffer and a hysteresis threshold are set. When the scene confidence vector changes beyond the switching threshold within the forward time window, a policy update is triggered; otherwise, the current policy is maintained. The actual power consumption data after collection and execution is compared with the expected power consumption pre-calculated by the strategy. If the actual power consumption deviation exceeds the set threshold, the strategy parameters are adaptively adjusted and the strategy parameters in the mapping table are updated.

[0049] It's important to note that a scene-power consumption strategy mapping table is constructed, pre-defining multi-dimensional power optimization strategies for each scene category. Strategy definitions include: in text reading scenarios, prioritizing a reduction in refresh rate to the lowest sustainable level (e.g., 1Hz), moderately reducing backlight brightness while maintaining high contrast to ensure text readability, and disabling or significantly reducing animation rendering effects. Another example is in video playback scenarios, maintaining a higher refresh rate (e.g., 60Hz), employing dynamic dimming of the backlight based on motion vector analysis to maintain HDR dynamic range. Yet another example is in image browsing scenarios, using a color-perception-guided pixel-driven voltage adjustment scheme to prioritize color accuracy. This solution achieves differentiated power consumption management "tailored to the scene" through accurate semantic scene identification, significantly reducing power consumption compared to a uniform strategy, and maximizing energy savings without the user noticing.

[0050] like Figure 2 As shown, the second aspect of the present invention provides a liquid crystal display power consumption optimization system based on screen content analysis, including a memory and a processor. The memory includes a liquid crystal display power consumption optimization method program based on screen content analysis. When the liquid crystal display power consumption optimization method program based on screen content analysis is executed by the processor, it implements any of the steps of the liquid crystal display power consumption optimization method based on screen content analysis.

[0051] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0052] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0053] In addition, in the various embodiments of the present invention, each functional unit can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0054] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0055] Alternatively, if the integrated units of this invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this invention, or the parts that contribute to the prior art, 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, server, or network device, etc.) to execute all or part of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

[0056] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for optimizing LCD screen power consumption based on image content analysis, characterized in that, include: Image data is continuously extracted from the LCD screen, and motion vector field data in the image data is obtained by extracting the motion vector field from the image data. By performing spatial clustering analysis on the motion vector field data in the image data, and using the motion vector field analysis results, a human eye gaze prediction model is constructed based on deep learning. The human eye gaze prediction model is used to predict areas with high visual saliency, and the refresh rate of each area within the LCD screen is initialized based on the motion vector field analysis results. The refresh rate of each area within the LCD screen is optimized using areas with high visual salience.

2. The LCD screen power consumption optimization method based on image content analysis according to claim 1, characterized in that, Image data is continuously extracted from the LCD screen, and motion vector field data is obtained from the image data by performing motion vector field extraction on the image data. Specifically: Image data of the current frame and the previous frame are simultaneously obtained from the frame buffer of the display driver IC in a dual-port parallel reading mode. The motion complexity index of the previous frame image is obtained, and the read image data is adaptively downsampled according to the motion complexity index of the previous frame image. The downsampled image is converted from the RGB color space to the YCbCr color space, and the luminance component is extracted for motion estimation. The luminance image of the current frame is recursively divided into quadtrees based on the gradient complexity within the block. Calculate the gradient magnitude variance of each candidate block. If the gradient magnitude variance is greater than a preset gradient magnitude threshold, then the corresponding candidate block is further divided into four sub-blocks. This process is repeated recursively until the minimum block size is reached. If the gradient magnitude variance is not greater than the preset gradient magnitude threshold, the partitioning is stopped; using the motion vector field calculated in the previous frame, a non-uniform block grid is obtained according to the partitioning result, and motion vector spatial domain prediction and temporal prediction are performed on each block in the current frame.

3. The LCD screen power consumption optimization method based on image content analysis according to claim 2, characterized in that, Also includes: In the spatial domain prediction, the median of the motion vectors of the adjacent left, top, and right top blocks of the current block is taken as the prediction value. In the temporal domain prediction, linear extrapolation is performed based on the motion vectors of the blocks with the same position as the current block in the previous frame, combined with the global motion trend. Calculate the predicted residual energy. If the residual energy is less than the residual energy threshold, the predicted motion vector is directly used as the motion vector of the current block, and the motion search of the current block is skipped. For blocks that cannot be reused by time prediction, a range search is initiated. During the range search, the motion vector magnitude of the block corresponding to the previous frame and the region where the block is located are used as the target range for the search. After the motion search is completed, if the matching cost is lower than a preset multiple of the optimal cost per integer pixel, subpixel interpolation and thinning are performed to obtain the subpixel precision motion vector. Median filtering and consistency verification are then performed on the motion vector field to obtain the motion vector field data in the image data.

4. The LCD screen power consumption optimization method based on image content analysis according to claim 1, characterized in that, By performing spatial clustering analysis on the motion vector field data in the image data, the specific results of the motion vector field analysis are as follows: Calculate the amplitude features, orientation angle features, amplitude time stability features, orientation change rate features, and local motion consistency features of each block in the motion vector field data of the image data, and construct a multidimensional motion feature vector for each block; Based on the multidimensional motion feature vector of each block, the motion vector amplitude of all blocks in the whole picture is statistically analyzed, a histogram is generated, and the amplitude corresponding to the first local minimum point of the histogram is used as the low speed threshold. The second local minimum point of the histogram is found as the high-speed threshold. At the same time, motion noise basis estimation is introduced, and the amplitude corresponding to the histogram that is lower than the preset quantile threshold is used as the noise threshold. Based on the multidimensional motion feature vector of each block and the noise threshold, spatial clustering analysis is performed on each block to generate static regions, low-speed motion regions and high-speed motion regions. Based on the static region, low-speed motion region, and high-speed motion region, a motion vector field analysis result is generated, and the motion vector field analysis result is output.

5. The LCD screen power consumption optimization method based on image content analysis according to claim 1, characterized in that, A human eye gaze prediction model is constructed based on deep learning, specifically as follows: Historical gaze center point data of static areas, low-speed motion areas, and high-speed motion areas are collected. A human eye gaze prediction model is constructed based on deep learning, and the historical gaze center point data of static areas, low-speed motion areas, and high-speed motion areas are used as model input. The probability of each pixel belonging to the human eye's gaze point is used as the model output, and the human eye gaze prediction model is used to train the training data. The human eye gaze prediction model is trained when its prediction accuracy is greater than a preset prediction accuracy threshold.

6. The LCD screen power consumption optimization method based on image content analysis according to claim 5, characterized in that, The human eye gaze prediction model is used to predict regions with high visual saliency, specifically as follows: Collect image data of the screen content within a preset time period, and input the image data of the screen content within the preset time period into the human eye gaze prediction model for prediction. By prediction, the probability that each pixel in the image data of the LCD screen's content within a preset time period belongs to the human eye's gaze point is obtained, and the probability that each pixel in the image data of the LCD screen's content within the preset time period belongs to the human eye's gaze point is calculated to generate a probability map. The locations of regions in the probability map with a probability greater than a preset probability threshold are obtained, and regions with high visual saliency are generated based on the regions in the probability map with a probability greater than the preset probability threshold.

7. The LCD screen power consumption optimization method based on image content analysis according to claim 1, characterized in that, The refresh rate of each area within the LCD screen is initialized based on the motion vector field analysis results, specifically as follows: Based on the motion vector field analysis results, static region, low-speed motion region, and high-speed motion region are obtained, and refresh rates are configured for static region, low-speed motion region, and high-speed motion region. Specifically, static areas perform intra-frame partial refresh, low-speed motion areas perform frequency reduction refresh, and high-speed motion areas maintain full frame rate refresh.

8. The LCD screen power consumption optimization method based on image content analysis according to claim 1, characterized in that, The refresh rate of each area within the LCD screen is optimized using areas with high visual saliency, specifically as follows: Determine whether there are visually salient regions in static, low-speed motion, and high-speed motion regions; When there are areas with high visual saliency in the static area, low-speed motion area, and high-speed motion area, increase the refresh rate of the corresponding area. When there are no visually significant areas in the static area, low-speed motion area, and high-speed motion area, the current refresh rate remains unchanged.

9. The LCD screen power consumption optimization method based on image content analysis according to claim 1, characterized in that, Also includes: The hash fingerprint comparison algorithm is used to calculate two consecutive frames of image data to obtain a perceptual hash fingerprint. Based on the perceptual hash fingerprint, it is determined whether the similarity between the two frames is greater than a preset similarity threshold. When the similarity between two frames determined by the perceptual hash fingerprint is greater than the preset similarity threshold, the reading operation of the frame buffer is stopped, and the driving signal of the previous frame is directly reused. When the change in the two frames of the perceptual hash fingerprint is detected to be significantly lower than the preset range threshold, the driver IC only performs data reading and data line driving on the pixel blocks that have changed, and reuses the driving state of the previous frame for the unchanged areas. If there are discrepancies in the hash fingerprint determination, the frame is further divided into several pixel blocks, the difference metric of each block is calculated, and the location and range of the changed pixel blocks are identified.

10. A liquid crystal display power consumption optimization system based on image content analysis, characterized in that, The device includes a memory and a processor. The memory includes a program for optimizing the power consumption of a liquid crystal display (LCD) screen based on screen content analysis. When the processor executes the program for optimizing the power consumption of an LCD screen based on screen content analysis, it implements the steps of the LCD screen power consumption optimization method as described in any one of claims 1-9.

Images