Partition dimming method and device based on image region features, equipment and medium

Abstract

The invention relates to a partition dimming method and device based on image region features, equipment and a medium, and the method comprises the steps: obtaining an RGB image, carrying out the brightness grading of the RGB image, and obtaining different brightness grade labels; high-frequency components in the RGB image are separated, image detail classification is carried out based on gray information of the high-frequency components, and different detail level labels are obtained; converting the RGB image into a YUV color space to construct a brightness matrix, and calculating a backlight value of each backlight partition based on the brightness matrix, the brightness level label and the detail level label; and constructing a multi-dimensional state space according to the combined state of the brightness level label and the detail level label, optimizing the S curve parameter based on the multi-dimensional state space, and performing pixel compensation based on the S curve parameter and the backlight value to generate a target compensation image. According to the invention, the accuracy of partition dimming of the display equipment is improved, and the quality of a display image is enhanced.

Description

Technical Field

[0001] The present application relates to the technical field of liquid crystal displays, and in particular to a partition dimming method, device, equipment and medium based on image area features. Background Art

[0002] Liquid crystal displays, as a non-autonomous luminous display device, are highly dependent on the illumination light source provided by the backlight module for their normal display function. However, during normal display, even when displaying a completely black image, the backlight module still maintains its highest brightness state, which not only leads to energy waste, but also often causes light leakage, further reducing the contrast of the displayed image. In order to effectively solve the above problems, reduce power consumption and improve the contrast of the display, relevant scholars have proposed a zone dimming technology. This technology divides the backlight module into multiple areas whose brightness can be independently controlled. During the actual dimming process, the system will flexibly adjust the brightness of each backlight area based on the real-time content of the input image, and accurately adjust the pixel value of the corresponding backlight area through pixel compensation technology.

[0003] In zone-based dynamic dimming technology, precise selection of backlight brightness is crucial for ensuring high performance and efficiency. In zone-based dynamic dimming, various backlight adjustment methods exist, including maximum, average, root mean square, and error correction. Several mainstream backlight brightness selection methods have their own characteristics: the maximum value method selects the maximum RGB brightness value of the sub-pixels in the region as the backlight setting, effectively avoiding overflow distortion during the image dimming process and protecting image details. However, its sensitivity to highlight pixel values ​​may cause a single highlight pixel in an overall dark area to significantly affect the backlight brightness, thereby affecting energy saving. The average value method uses the arithmetic average of the pixel grayscale values ​​in the region as the backlight brightness. It is simple to calculate, but when processing areas with large differences in pixel values, it may not accurately reflect the image brightness characteristics, resulting in distortion. The root mean square method is based on the average value method. Through normalization processing and calculation of the root mean square value, it improves the accuracy of reflecting image brightness characteristics and reduces image distortion. Although the computational complexity is increased, it is still within the reasonable range of technical implementation. The error correction method comprehensively considers the average brightness information of the image and the difference between the maximum brightness and the average brightness, ensuring the simultaneous presentation of image details. However, when processing images with uneven brightness distribution, due to the fixed nature of the correction coefficient, it may perform poorly and is only suitable for image scenes with relatively uniform brightness distribution. Summary of the Invention

[0004] The purpose of the embodiments of the present application is to propose a method, apparatus, device and medium for zone dimming based on image area features, so as to improve the accuracy of zone dimming of a display device and enhance the quality of the displayed image.

[0005] In order to solve the above technical problems, the embodiments of the present application provide a zone dimming method based on image region features, including:

[0006] Acquire an RGB image, and perform brightness grading on the RGB image to obtain different brightness level labels;

[0007] Separating high-frequency components from the RGB image, and performing image detail classification based on grayscale information of the high-frequency components to obtain different detail level labels;

[0008] Convert the RGB image to a YUV color space to construct a brightness matrix, and calculate a backlight value of each backlight partition based on the brightness matrix, the brightness level label, and the detail level label;

[0009] A multidimensional state space is constructed according to the combined state of the brightness level label and the detail level label, S-curve parameters are optimized based on the multidimensional state space, and pixel compensation is performed based on the S-curve parameters and the backlight value to generate a target compensated image.

[0010] In order to solve the above technical problems, the embodiments of the present application provide a zone dimming device based on image area features, including:

[0011] A brightness grading module is used to obtain an RGB image and perform brightness grading on the RGB image to obtain different brightness level labels;

[0012] a detail classification module, configured to separate high-frequency components from the RGB image and classify image details based on the grayscale information of the high-frequency components to obtain different detail level labels;

[0013] a backlight value calculation module, configured to convert the RGB image into a YUV color space to construct a brightness matrix, and calculate a backlight value of each backlight partition based on the brightness matrix, the brightness level label, and the detail level label;

[0014] A compensation image generation module is used to construct a multidimensional state space according to the combination state of the brightness level label and the detail level label, optimize S-curve parameters based on the multidimensional state space, and perform pixel compensation based on the S-curve parameters and the backlight value to generate a target compensated image.

[0015] In order to solve the above technical problems, a technical solution adopted by the present invention is: to provide a display device, including one or more processors; a memory for storing one or more programs, so that the one or more processors can implement any one of the above-mentioned partition dimming methods based on image area features.

[0016] In order to solve the above technical problems, a technical solution adopted by the present invention is: a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements any one of the above-mentioned partition dimming methods based on image area features.

[0017] Embodiments of the present invention provide a method, apparatus, device, and medium for zone dimming based on image region features. The method includes: acquiring an RGB image and performing brightness grading on the RGB image to obtain different brightness level labels; separating high-frequency components from the RGB image and performing image detail classification based on the grayscale information of the high-frequency components to obtain different detail level labels; converting the RGB image to a YUV color space to construct a brightness matrix, and calculating the backlight value of each backlight zone based on the brightness matrix, the brightness level labels, and the detail level labels; constructing a multidimensional state space based on the combined state of the brightness level labels and the detail level labels, optimizing S-curve parameters based on the multidimensional state space, and performing pixel compensation based on the S-curve parameters and the backlight value to generate a target compensated image. The embodiment of the present invention improves the accuracy of zone dimming of a display device and enhances the quality of the displayed image by performing brightness grading and detail classification on the RGB image, calculating the backlight value of each backlight zone based on the brightness level labels and detail level labels, confirming the S-curve parameters, and performing pixel compensation based on the S-curve parameters and the backlight value. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the solutions in this application, a brief introduction will be given below to the drawings required for use in the description of the embodiments of this application. Obviously, the drawings described below are some embodiments of this application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0019] Figure 1 This is a flowchart of the implementation process of the partition dimming method based on image area features provided in an embodiment of the present application;

[0020] Figure 2 This is a flowchart for implementing the first sub-process in the zone dimming method based on image area features provided in an embodiment of the present application;

[0021] Figure 3 This is a flowchart for implementing the second sub-process in the partition dimming method based on image area features provided in an embodiment of the present application;

[0022] Figure 4 This is a flowchart for implementing the third sub-process in the partition dimming method based on image area features provided in an embodiment of the present application;

[0023] Figure 5 This is a flowchart for implementing the fourth sub-process in the partition dimming method based on image area features provided in an embodiment of the present application;

[0024] Figure 6 Schematic diagram of the relationship between the S-curve parameters and the S-curve provided in the embodiment of the present application;

[0025] Figure 7 This is a flowchart for implementing the fifth sub-process in the partition dimming method based on image area features provided in an embodiment of the present application;

[0026] Figure 8 This is a flowchart for implementing the sixth sub-process in the partition dimming method based on image area features provided in an embodiment of the present application;

[0027] Figure 9 Schematic diagram of a zone dimming device based on image area features provided in an embodiment of the present application;

[0028] Figure 10 Schematic diagram of a display device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0029] Unless otherwise defined, all technical and scientific terms used herein have the same meanings as commonly understood by those skilled in the art to which this application belongs. The terms used in the specification of the application are for the purpose of describing specific embodiments only and are not intended to limit this application. The terms "including" and "having" and any variations thereof in the specification and claims of this application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of this application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0030] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0031] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0032] The present invention will be described in detail below with reference to the accompanying drawings and embodiments.

[0033] It should be noted that the zone dimming method based on image area features provided in the embodiments of the present application is generally executed by a server, and accordingly, the zone dimming device based on image area features is generally configured in the server.

[0034] See also Figure 1 , Figure 1 A specific implementation of the regional dimming method based on image area features is shown.

[0035] It should be noted that the method of the present invention is not limited to the method of Figure 1 The process sequence shown is limited to the following steps:

[0036] S1: Acquire an RGB image, and perform brightness grading on the RGB image to obtain different brightness level labels.

[0037] The partition dimming method based on image area features provided in this application is classified according to image brightness and detail features, and the backlight brightness and pixels are compensated and adjusted by the grayscale and detail content of the divided areas, thereby improving the display effect, increasing the contrast, reducing the average power consumption of the display, improving brightness distortion, and enhancing the visual effect.

[0038] Specifically, when viewing a color image in RGB pixel format, the human eye is sensitive to image brightness. To make the classification results more consistent with the characteristics of the human eye, image brightness is divided into three levels. Therefore, in the embodiment of the present application, an RGB image is obtained and brightness is graded to obtain different brightness level labels. In one specific embodiment, the brightness level labels include high brightness, medium brightness, and low brightness.

[0039] See also Figure 2 , Figure 2 A specific implementation of step S1 is shown, which is described in detail as follows:

[0040] S11: Obtain the RGB image and convert it into a grayscale image. S12: Calculate the grayscale value of each pixel in the grayscale image and generate a grayscale histogram of the entire image based on the grayscale value. S13: Calculate the average brightness based on the grayscale histogram of the entire image and perform brightness grading based on the average brightness to obtain different brightness level labels.

[0041] Specifically, the brightness is represented by the grayscale average of the image, and the brightness information of the image is statistically analyzed using a histogram. The image brightness level is [0, 255]. Therefore, in the embodiment of the present application, the RGB image is first converted to a grayscale image, and then the grayscale value of each pixel in the grayscale image is calculated. The grayscale histogram of the entire image is calculated based on the grayscale value. Among them, the brightness statistical histogram of the image I(x, y) is defined as follows:

[0042]

[0043] Among them, s i is the ith level brightness value of I(x,y), h(s i ) is the brightness value s in I(x,y) i where L is the total number of pixels in the image. Finally, the average brightness is calculated based on the grayscale histogram of the entire image, and brightness levels are graded based on the average brightness to obtain different brightness level labels. In one specific embodiment, the human eye perceives images with brightness levels above 200 as brighter; when the brightness level is below 60, the human eye perceives the image as darker. Therefore, using this as the dividing line, images with brightness above 200 are classified as high brightness, those below 60 as low brightness, and those between 60 and 200 as medium brightness.

[0044] S2: Separate the high-frequency components in the RGB image, and perform image detail classification based on the grayscale information of the high-frequency components to obtain different detail level labels.

[0045] Specifically, after feature classification of brightness information, the present application further classifies image details. The details of an image are often evaluated using the frequency of the image. The frequency of an image is an indicator of the severity of the grayscale change in the image. It corresponds to the pixels in the image where the grayscale changes dramatically. Different frequency information has different effects in the image structure. The higher the frequency of the image, the more obvious the change, and the more details the image contains. Conversely, the lower the frequency of the image, the less obvious the change, and the less details the image contains. Therefore, the present application needs to classify details based on the frequency of the image. In the present application, the high-frequency component in the RGB image is separated, and the image details are classified based on the grayscale information of the high-frequency component to obtain different detail level labels.

[0046] See also Figure 3 , Figure 3 A specific implementation of step S2 is shown, which is described in detail as follows:

[0047] S21: Filter the RGB image using a Gaussian filter to generate a low-frequency component, and generate a high-frequency component based on the RGB image and the low-frequency component. S22: Convert the high-frequency component into a grayscale matrix. S23: Generate a target threshold based on the grayscale matrix, and perform image detail classification based on the target threshold using the Otsu algorithm to obtain different detail level labels.

[0048] Specifically, Gaussian filtering is used to separate an image into high-frequency components and low-frequency components. Low-frequency components include the image's outline, color, and brightness, while high-frequency information includes the image's details, texture, and edges. Gaussian filtering is essentially a linear low-pass filter that is widely used in image processing. The conditions for Gaussian filtering are as follows:

[0049]

[0050] Among them, by setting the standard deviation σ and window size radius Determine the Gaussian template to obtain the high and low frequency parts of the image, so the selection of the standard deviation σ and the window size radius r becomes crucial. To ensure that the obtained high frequency information can represent the details extracted from the image, it is necessary to determine the specific data of [r,σ]. This application adopts the Gaussian filter scale selection algorithm and finally determines to set [r,σ] to [30,10]. At this time, it can remove the interference information of the image and highlight the image details. Assume that I is the input image, and the low frequency component I L The calculation formula is: L =I*G(x,y); high frequency component I H The calculation formula is I H =II L , where * represents the convolution operation, I L is the low-frequency image matrix, i.e., the low-frequency component; I H is the high-frequency image matrix, i.e., the high-frequency component; G(x,y) is a two-dimensional Gaussian formula with a standard deviation of σ. Then the high-frequency component is grayed to obtain the grayscale matrix I HG Then, a target threshold is generated based on the grayscale matrix, and the Otsu algorithm is used to perform image detail classification based on the target threshold to obtain different detail level labels.

[0051] See also Figure 4 , Figure 4 A specific implementation of step S23 is shown, which is described in detail as follows:

[0052] S231: Calculate the grayscale histogram of the high-frequency component based on the grayscale matrix to obtain the target grayscale histogram, and determine the frequency and frequency of each grayscale level based on the target grayscale histogram. S232: Calculate the inter-class variance when the grayscale level is the threshold based on the frequency and frequency of each grayscale level, and use the inter-class variance of the maximum value as the target threshold. S233: Segment the RGB image according to the target threshold to generate high grayscale areas and low grayscale areas. S234: Re-adopt the Otsu algorithm to respectively determine the high detail threshold and the low detail threshold of the high grayscale area and the low grayscale area, and classify the image details of the high grayscale area and the low grayscale area based on the high detail threshold and the low detail threshold to obtain different detail level labels.

[0053] In the embodiment of the present application, the Otsu algorithm is used to perform two image detail classifications on the RGB image to obtain different detail level labels. Among them, the Otsu algorithm adaptively determines the optimal segmentation threshold by maximizing the inter-class variance, and is applicable to the bimodal histogram scene of the high-frequency detail image. The Otsu algorithm first separates the high frequency of the image, performs Gaussian filtering on the RGB image, and extracts the high frequency component I from it. HG , and calculate the grayscale histogram of the high-frequency component; then, in order to explore the optimal threshold for high and low frequency classification, traverse the grayscale level t, calculate the inter-class variance when t is the threshold, and select the t that maximizes the inter-class variance as the optimal threshold T opt After the first segmentation, the Otsu algorithm is applied twice to the high / low grayscale areas to obtain the low detail threshold T low and high detail threshold T high , and finally the pixels are classified as low details: g<T low 、Medium details:T low ≤g≤T high , High detail: g>T high Three categories.

[0054] Specifically, the input image is preprocessed by Gaussian filtering to separate the high-frequency components I containing detail information such as texture and edge. HG , calculate I HG Grayscale histogram, count the frequency of each gray level n i and frequency p i =n i / N (N is the total number of pixels). Assume that in an image with rich details, the high-frequency component histogram presents a bimodal distribution, with Peak 1 being the low grayscale area, mainly containing dark details, and Peak 2 being the high grayscale area, mainly containing bright details. The goal of the Otsu algorithm is to find the optimal threshold T opt Separate the two peaks and maximize the difference between the two types of details. The maximum between-class variance is determined as follows:

[0055] First, for each grayscale level, we classify it. Assuming that the total number of pixels in the image is N, we divide the image into two categories C1 and C2, where C1 contains pixels with grayscale values ​​less than or equal to t, and C2 contains pixels with grayscale values ​​greater than t. The value range of t is usually the possible values ​​of the image grayscale, from 0 to 255. For each grayscale level t, we calculate the inter-class variance when t is used as the threshold. The inter-class variance is calculated as follows: count the frequency n of the pixels in C1 and C2 i and frequency p i =n i / N. Let the average grayscale of C1 and C2 be μ1 and μ2 respectively, and the total average grayscale of the image be μ. The calculation formula is:

[0056]

[0057] in, p2=1-p1, (L is the grayscale level of the image). Select t that maximizes the inter-class variance as the target threshold T opt Since the detail area is divided into three parts according to grayscale in this algorithm, multi-level segmentation is required, and the optimal threshold T is used. opt Perform the first segmentation on the image to divide it into high grayscale area and low grayscale area. Apply the Otsu algorithm twice to the high / low grayscale area to obtain the high detail threshold T high and low detail threshold T low Finally, based on the high detail threshold and the low detail threshold, the image details of the high grayscale area and the low grayscale area are classified respectively to obtain different detail level labels.

[0058] The embodiment of the present application realizes the accurate quantitative classification of detail intensity through Otsu's adaptive bimodal separation. It does not need to pre-set thresholds and can automatically adapt to changes in image illumination and contrast, showing strong adaptability. In terms of computational efficiency, the algorithm only needs to traverse the histogram once, and its computational complexity is O(256), which is very efficient. In a physical sense, maximizing the inter-class variance is equivalent to minimizing the intra-class difference. This feature ensures the rationality of detail classification. For example, in image processing, low-detail areas (such as smooth backgrounds) will be filtered out by low thresholds, high-detail areas (such as sharp edges) can be retained by high thresholds, and medium-detail areas (such as gradient textures) can maintain a natural transition, so that the processed image effect is more in line with actual needs.

[0059] S3: Convert the RGB image into a YUV color space to construct a brightness matrix, and calculate the backlight value of each backlight partition based on the brightness matrix, the brightness level label, and the detail level label.

[0060] Specifically, after obtaining the high, medium, and low detail classification ranges, this application uses an error correction algorithm based on image feature classification to determine the optimal backlight value for the image. The error correction algorithm can determine the backlight value based on the average brightness of the image grayscale and the difference between the maximum brightness and the average brightness. Average brightness is the average brightness value of all pixels in the image. In backlight brightness extraction, average brightness is often used as a preliminary estimate of backlight brightness. By calculating the average brightness of the image, a rough backlight brightness value can be obtained. However, this method may be affected by extreme brightness values ​​in the image (such as highlights or shadows), resulting in inaccurate backlight value estimation. Maximum brightness is the brightness of the pixel with the highest brightness value in the image, reflecting the brightest part of the image. However, since maximum brightness typically only represents a very small area in the image, it may not accurately reflect the backlight brightness of the entire image. Average brightness difference refers to the difference between the maximum brightness and average brightness in the image, reflecting the degree of dispersion in the image brightness distribution. By analyzing the average brightness difference, the uniformity of the brightness distribution in the image can be understood. If the average brightness difference is large, it indicates that there are significant brightness differences in the image, and more precise backlight brightness adjustment may be required. The error correction algorithm is a more accurate and robust method for backlight brightness extraction. It combines multiple factors, such as average brightness and the difference between maximum brightness and average brightness, to more accurately assess the brightness distribution characteristics of the image and determine a more reasonable backlight value based on this.

[0061] The error correction algorithm first calculates the image's average brightness and maximum brightness, then calculates the difference between them. By combining these two factors, the algorithm can more accurately determine the image's brightness distribution characteristics. If the average brightness is high and the average brightness difference is small, the image is generally bright and the brightness distribution is relatively uniform. In this case, the backlight brightness may need to be reduced to avoid overexposure. Conversely, if the average brightness is low and the average brightness difference is large, the image is generally dark and has large brightness variations. In this case, the backlight brightness may need to be increased to improve the image's brightness level. Compared to other algorithms, the error correction algorithm offers higher accuracy and robustness. It reduces the interference of extreme brightness values ​​on backlight estimation, improving the stability and accuracy of backlight brightness extraction. Furthermore, the error correction algorithm is adaptable to different types of images and scenes, making it more versatile and adaptable.

[0062] In the embodiment of the present application, the present application converts the RGB image into the YUV color space, constructs the brightness matrix using the brightness equation Y=0.299R+0.587G+0.114B, and implements the backlight control upgrade through triple optimization. The triple optimization is as follows:

[0063] 1. Perception Optimization: The Y component weight distribution strictly matches the sensitivity of the human eye's cone cells (green > red > blue), making the brightness assessment more consistent with subjective feelings; 2. Noise Reduction Optimization: The weighted averaging process can suppress single-channel noise, and combined with Gaussian filtering, the brightness distribution is smoother; 3. Energy Efficiency Optimization: Relying on a hardware acceleration unit to achieve real-time YUV conversion, the calculation efficiency is greatly improved compared to the RGB maximum value method.

[0064] See also Figure 5 , Figure 5 A specific implementation of step S3 is shown, which is described in detail as follows:

[0065] S31: Convert the RGB image to the YUV color space to construct the brightness matrix. S32: Calculate the target average brightness, maximum brightness, and brightness difference of each backlight partition based on the brightness matrix. S33: Calculate the detail content ratio in the image based on the brightness level label and the detail level label, and calculate the dimming factor based on the detail content ratio in the image. S34: Calculate the backlight value of each backlight partition based on the dimming factor, the target average brightness, the maximum brightness, and the brightness difference.

[0066] Specifically, the brightness matrix composed of the image grayscale and the Y component of YUV is used as the basis for backlight selection. In the brightness matrix, the brightness value of each pixel is the value of the Y component. By comparing these values, the maximum brightness matrix can be found. The algorithm is implemented as follows:

[0067]

[0068] Among them, BL max is the maximum brightness value in the area, BL ave is the average brightness value, T bl is the final backlight value output by the area, ω is the dimming factor, BL diff BL is the difference between the maximum brightness value and the average brightness value in the partition. correction Dynamic backlight compensation value. To calculate the dimming factor, the target average brightness, maximum brightness, and brightness difference of each backlight zone are calculated based on the brightness matrix. The detail content ratio in the image is then calculated based on the brightness level label and detail level label. The dimming factor is then calculated based on the detail content ratio in the image. The calculation formula for the detail content ratio in the image is:

[0069]

[0070] in, s i Indicates the value of detail brightness in area I, K detail Indicates I HGThe ratio of details in the area occupied. The calculation formula of dimming factor is: ω=1-K detail .

[0071] Among them, BL correction Will follow BL diff The difference between the two is approximately linear. This means that when the brightness difference between areas increases, a larger backlight correction factor is needed to balance the brightness and ensure display quality. The dimming factor ω is a key parameter used to adjust the backlight brightness of the area in real time. By adjusting ω, power consumption can be minimized while maintaining display quality. ave When BL is larger, correction This means that in brighter areas, the need for backlight correction is lower. At this point, the backlight brightness can be further adjusted according to the content of image details to reduce image distortion. ave At the middle brightness, since the brightness of the input image is relatively uniform, the backlight brightness can be significantly reduced, thereby reducing power consumption. ave When the dimming factor is low and the image has more details, the backlight brightness should be kept high to ensure that these details are preserved. When the image has less details, the brightness dimming factor ω will also be reduced accordingly, allowing the backlight brightness to be further reduced, thereby improving energy saving efficiency.

[0072] S4: constructing a multidimensional state space according to the combined state of the brightness level label and the detail level label, optimizing S-curve parameters based on the multidimensional state space, and performing pixel compensation based on the S-curve parameters and the backlight value to generate a target compensated image.

[0073] Specifically, in order to effectively ensure the image display effect, this application uses the S-curve pixel compensation method for pixel correction. Considering that if all images use the same S-curve for pixel compensation, it will lead to problems such as loss of details, contrast imbalance, color distortion, etc. The equation of the S-curve is as follows:

[0074]

[0075] Among them, L point Represents the pixel value of the inflection point of the S curve, L input represents the input pixel value, a is the S-curve parameter, that is, the curvature parameter. point It is the inflection point of the S-curve and determines the range of pixel adjustment.

[0076]

[0077] Among them, L 20 It is the value corresponding to the grayscale histogram accumulated to 20% of the total number of pixels, L 80It is the value corresponding to 80% of the total number of pixels. The q value of the image is determined according to the brightness and details of the image to adjust the pixels of each partition. Figure 6 , Figure 6 is a schematic diagram of the relationship between the S-curve parameters and the S-curve provided in the embodiment of the present application, Figure 6 As can be seen, the relationship between the S-curve and the S-curve parameter a value is that the more curved the curve, the greater the increase or decrease in pixel value. Therefore, the a value in the S-curve is important for the study of image display. Therefore, in the embodiment of the present application, a multidimensional state space is constructed based on the combination of the brightness level label and the detail level label. The S-curve parameters are optimized based on the multidimensional state space, and pixel compensation is performed based on the S-curve parameters and the backlight value to generate a target compensated image.

[0078] See also Figure 7 , Figure 7 A specific implementation of step S4 is shown, which is described in detail as follows:

[0079] S41: Construct the multidimensional state space according to the combined state of the brightness level label and the detail level label. S42: Initialize the network model and the experience playback buffer, wherein the network model includes an Actor network, a Critic network, and a target network. S43: Use the multidimensional state space as the input state, and perform model training based on the input state through the network model, and determine the S-curve parameters through the trained target network. S44: Compensate the pixel values ​​of the RGB image through the S-curve parameters and the backlight value to generate the target compensated image.

[0080] This embodiment of the present application uses the Deep Deterministic Policy Gradient (DDPG) algorithm to perform S-curve pixel compensation for images of varying detail and brightness. The DDPG algorithm is a deep reinforcement learning algorithm that continuously learns the optimal pixel compensation strategy through interaction between an agent and its environment. During training, the agent selects appropriate pixel compensation factors based on the current image state and continuously optimizes the strategy based on environmental feedback to achieve optimal pixel compensation. The combined states of the brightness level label and detail level label are divided into nine states: high brightness and high detail; high brightness and medium detail; high brightness and low detail; medium brightness and high detail; medium brightness and medium detail; medium brightness and low detail; low brightness and high detail; low brightness and medium detail; and low brightness and low detail. Specifically, the brightness level label of each physical partition is encoded as a discrete value: low brightness = 0, medium brightness = 1, high brightness = 2; and the detail level label of each physical partition is encoded as a discrete value: low detail = 0, medium detail = 1, high detail = 2. A 9-dimensional state space is generated using the formula state_id = 3 × brightness code + detail code, and this 9-dimensional state space is used as the input state. The input state is fed into the Actor network to guide the selection of strategies. The action space A is defined as the S-curvature value a, which varies between [0, 0.03]. This parameter is used to adjust the response characteristics of the LCD pixels, thereby affecting the image display effect.

[0081] Specifically, the DDPG algorithm leverages the powerful nonlinear fitting capabilities of deep neural networks to simulate high-dimensional, continuous real-world spatial environments and learn the complex relationships within them. The algorithmic flow is shown in Algorithm 2. In the DDPG algorithm, the actor network and the critic network are two key components: the actor network outputs a deterministic action (i.e., a policy for adjusting parameter a) based on the current input state, while the critic network evaluates the expected long-term cumulative reward of that action. To improve the algorithm's stability and convergence, DDPG introduces an experience replay mechanism and a target network. The experience replay mechanism stores experience samples (state, action, reward, next state) of the agent's interaction with the environment in a replay buffer. During training, random sampling is performed to break sample correlation and improve utilization. The target network uses a soft update mechanism to periodically and slowly copy the parameters of the online network to the target network, making parameter changes smoother, thereby reducing training oscillations and ensuring stable convergence of the algorithm. Therefore, before model training, the Actor network, Critic network, and target network must be initialized. The multidimensional state space is used as the input state, and the network model is trained based on the input state. The S-curve parameters are determined using the trained target network. The pixel values ​​of the RGB image are compensated using the S-curve parameters and the backlight value to generate the target compensated image.

[0082] In the DDPG algorithm, each state transition is accompanied by a corresponding reward signal. The goal of the agent is to learn a strategy that maximizes the cumulative reward by selecting its optimal action given the current state. k+1 |s k )=P(s k+1 |s1,s2,...,s k ), a typical MDP problem consists of {S,A,R,P}. (1) State space S: S={s1,s2,...,s k} is the set of all states. At any given point in time, the agent is in a state in the state space. k ∈S represents the state at time step k. (2) Action space A: A={a1,a2,...,a k} are all possible actions that the agent can take. Each state corresponds to a set of optional actions. k∈A represents the action taken at time step k. (3) Reward function R: The reward function defines the immediate reward provided by the environment to the agent after the agent performs an action. (4) State transition probability P: The state transition probability describes the probability that the environment transitions from one state to another after the agent performs an action. This probability is expressed as P(s k+1 |s k ,a k ), that is, in state s k Next, perform action a k Then transfer to state s k+1 The Markov property requires that the probability of state transitions depends only on the current state and the action taken, and not on previous states. Through these components, the MDP provides a formal framework for describing the environment, agents, and interactions in reinforcement learning problems. In the MDP, the agent learns and optimizes its policy to solve optimization problems using methods such as value functions or policy gradients.

[0083] See also Figure 8 , Figure 8 A specific implementation of step S43 is shown, which is described in detail as follows:

[0084] S431: Use the multidimensional state space as the input state, and input the input state into the Actor network so that the Actor network outputs the current optimal S-curve parameters according to the current state, and uses the current optimal S-curve parameters as the action. S432: Adjust the liquid crystal pixels based on the current optimal S-curve parameters to generate an adjusted image, and calculate the peak signal-to-noise ratio based on the adjusted image, and use the peak signal-to-noise ratio as a reward. S433: Store the input state, the action, and the reward in the experience replay buffer. S434: Randomly sample experience data from the experience replay buffer, and update the parameters of the Actor network and the Critic network based on the experience data, and use the soft update method to update the parameters of the objective function to generate the trained target network. S435: Determine the S-curve parameters based on the trained target network.

[0085] Specifically, the Peak Signal to Noise Ratio (PSNR) is used as a reward in this application. The Peak Signal to Noise Ratio provides a quantitative measure of image quality by calculating the difference in pixel values ​​between the original image and the processed image. It can stably and repeatedly reflect the changes in image quality after various processing, so that the image quality under different algorithms, systems or parameter settings can be directly compared. In addition, the value of PSNR is closely related to the degree of image distortion. In the display screen optimization problem, when adjusting parameters such as image brightness, contrast, and details, different degrees of distortion may be introduced. PSNR can accurately capture these distortions. The higher its value, the smaller the difference between the processed image and the original image, that is, the lower the distortion, the better the image quality; conversely, the lower the PSNR value, the more serious the image distortion. Among them, PSNR is defined as the formula:

[0086]

[0087] Among them, T MSE The mean square error between the original image and the processed image, T PSNR is the peak signal-to-noise ratio, and L is the grayscale level of the image. The calculation formula of the soft update method is:

[0088]

[0089] Among them, τ is the soft update coefficient, θ μ is the Actor network parameter, θ μ' is the target Actor network parameter, θ Q is the Critic network parameter, θ Q' is the target Critic network parameter.

[0090] In an embodiment of the present application, an RGB image is acquired and brightness graded to obtain different brightness grade labels; high-frequency components in the RGB image are separated, and image details are classified based on the grayscale information of the high-frequency components to obtain different detail grade labels; the RGB image is converted to a YUV color space to construct a brightness matrix, and the backlight value of each backlight partition is calculated based on the brightness matrix, the brightness grade label, and the detail grade label; a multidimensional state space is constructed based on the combined state of the brightness grade label and the detail grade label, S-curve parameters are optimized based on the multidimensional state space, and pixel compensation is performed based on the S-curve parameters and the backlight value to generate a target compensated image. The embodiment of the present invention performs brightness grading and detail classification on the RGB image, calculates the backlight value of each backlight partition based on the brightness grade label and the detail grade label, confirms the S-curve parameters, performs pixel compensation based on the S-curve parameters and the backlight value, and generates a target compensated image, which is conducive to improving the accuracy of the display device's partition dimming and enhancing the quality of the displayed image.

[0091] The embodiment of the present application can significantly reduce the average power consumption of the display through a hierarchical reinforcement learning algorithm based on image classification, and has made great improvements to the texture details and brightness distortion of the image, and has greatly improved the PSNR and average information entropy. The algorithm retains more image texture details, reduces power consumption while improving image display quality, and basically meets the requirement of selecting the best between power consumption and display quality. The DDPG algorithm is used to compensate pixels through different brightness and detail features, so that the average power consumption of the display is significantly reduced. At the same time, the PSNR and average information entropy are greatly improved, which means that the degree of image distortion is significantly reduced, and the increase in average information entropy indicates that the image contains more information and the texture details are better preserved.

[0092] Please refer to Figure 9 , as a response to the above Figure 1 The present application provides an embodiment of a zone dimming device based on image area features. Figure 1 Corresponding to the method embodiment shown, the apparatus can be specifically applied to various display devices.

[0093] like Figure 9 As shown, the regional dimming device based on image region features of this embodiment includes: a brightness grading module 51, a detail classification module 52, a backlight value calculation module 53 and a compensation image generation module 54, wherein:

[0094] A brightness grading module 51 is used to obtain an RGB image and perform brightness grading on the RGB image to obtain different brightness level labels;

[0095] a detail classification module 52 for separating high-frequency components from the RGB image and performing image detail classification based on grayscale information of the high-frequency components to obtain different detail level labels;

[0096] a backlight value calculation module 53 for converting the RGB image into a YUV color space to construct a brightness matrix, and calculating the backlight value of each backlight partition based on the brightness matrix, the brightness level label, and the detail level label;

[0097] The compensation image generation module 54 is used to construct a multidimensional state space according to the combination state of the brightness level label and the detail level label, optimize the S-curve parameters based on the multidimensional state space, and perform pixel compensation based on the S-curve parameters and the backlight value to generate a target compensated image.

[0098] Furthermore, the brightness grading module 51 includes:

[0099] An image acquisition unit, configured to acquire the RGB image and convert the RGB image into a grayscale image;

[0100] A grayscale value calculation unit, configured to calculate the grayscale value of each pixel in the grayscale image and generate a grayscale histogram of the entire image based on the grayscale value;

[0101] The average brightness calculation unit is used to calculate the average brightness according to the grayscale histogram of the whole image, and perform brightness grading according to the average brightness to obtain different brightness level labels.

[0102] Furthermore, the detail classification module 52 includes:

[0103] a high-frequency component generating unit, configured to filter the RGB image using a Gaussian filter to generate a low-frequency component, and generate a high-frequency component based on the RGB image and the low-frequency component;

[0104] A grayscale matrix conversion unit, configured to convert the high-frequency component into a grayscale matrix;

[0105] A target threshold generating unit is configured to generate a target threshold based on the grayscale matrix, and to perform image detail classification based on the target threshold using an Otsu algorithm to obtain different detail level labels.

[0106] Furthermore, the target threshold generating unit includes:

[0107] a grayscale histogram generating unit, configured to calculate a grayscale histogram of the high-frequency component based on the grayscale matrix to obtain a target grayscale histogram, and determine the frequency and number of each grayscale level based on the target grayscale histogram;

[0108] an inter-class variance calculation unit, configured to calculate the inter-class variance when the gray level is a threshold value based on the gray level frequency and the frequency, and use the maximum inter-class variance as the target threshold value;

[0109] An image segmentation unit, configured to segment the RGB image using the target threshold to generate a high grayscale area and a low grayscale area;

[0110] A secondary classification unit is used to re-adopt the Otsu algorithm to respectively determine the high detail threshold and the low detail threshold of the high grayscale area and the low grayscale area, and classify the image details of the high grayscale area and the low grayscale area based on the high detail threshold and the low detail threshold to obtain different detail level labels.

[0111] Furthermore, the backlight value calculation module 53 includes:

[0112] a brightness matrix construction unit, configured to convert the RGB image into the YUV color space to construct the brightness matrix;

[0113] a brightness difference calculation unit, configured to calculate a target average brightness, a maximum brightness, and a brightness difference of each of the backlight subareas based on the brightness matrix;

[0114] a dimming factor calculation unit, configured to calculate a detail content ratio in an image based on the brightness level label and the detail level label, and calculate a dimming factor based on the detail content ratio in the image;

[0115] A backlight value generating unit is configured to calculate the backlight value of each of the backlight subareas based on the dimming factor, the target average brightness, the maximum brightness, and the brightness difference.

[0116] Furthermore, the compensation image generation module 54 includes:

[0117] a multidimensional state space generating unit, configured to construct the multidimensional state space according to a combination state of the brightness level label and the detail level label;

[0118] Initialization unit, used to initialize the network model and experience replay buffer, where the network model includes the Actor network, the Critic network, and the target network;

[0119] a model training unit, configured to use the multidimensional state space as an input state, perform model training based on the input state through the network model, and determine the S-curve parameters through the trained target network;

[0120] A pixel compensation unit is used to compensate the pixel values ​​of the RGB image using the S-curve parameters and the backlight value to generate the target compensated image.

[0121] Furthermore, the model training unit includes:

[0122] an action confirmation unit, configured to use the multidimensional state space as the input state, input the input state into an Actor network, so that the Actor network outputs current optimal S-curve parameters according to the current state, and uses the current optimal S-curve parameters as an action;

[0123] a reward confirmation unit, configured to adjust liquid crystal pixels based on the current optimal S-curve parameters, generate an adjusted image, calculate a peak signal-to-noise ratio based on the adjusted image, and use the peak signal-to-noise ratio as a reward;

[0124] a data storage unit, configured to store the input state, the action, and the reward in the experience replay buffer;

[0125] A training completion unit, configured to randomly sample experience data from the experience replay buffer, update parameters of the actor network and the critic network based on the experience data, and update parameters of the objective function using a soft update method to generate the trained target network;

[0126] A parameter determination unit is used to determine the S-curve parameters according to the trained target network.

[0127] In order to solve the above technical problems, the present application also provides a display device. Figure 10 , Figure 10 This is a basic structural block diagram of the display device in this embodiment.

[0128] The display device 6 includes a memory 61, a processor 62, and a network interface 63 that are interconnected through a system bus. It should be noted that Figure 10 Only a display device 6 having three components, namely a memory 61, a processor 62, and a network interface 63, is shown. However, it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead. It should be understood by those skilled in the art that the display device herein is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to a microprocessor, an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA), a digital signal processor (DSP), an embedded device, etc.

[0129] Display devices can be computing devices such as desktop computers, laptops, PDAs, and cloud servers. Display devices can interact with users through keyboards, mice, remote controls, touchpads, or voice-activated devices.

[0130] The memory 61 includes at least one type of readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 61 can be an internal storage unit of the display device 6, such as the hard disk or memory of the display device 6. In other embodiments, the memory 61 can also be an external storage device of the display device 6, such as a plug-in hard disk equipped on the display device 6, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. Of course, the memory 61 can also include both the internal storage unit of the display device 6 and its external storage device. In this embodiment, the memory 61 is generally used to store the operating system and various application software installed on the display device 6, such as the program code of the zone dimming method based on image area features. In addition, the memory 61 can also be used to temporarily store various types of data that have been output or are to be output.

[0131] In some embodiments, the processor 62 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip. The processor 62 is generally used to control the overall operation of the display device 6. In this embodiment, the processor 62 is used to execute program code stored in the memory 61 or process data, such as executing the program code of the aforementioned zone dimming method based on image region features to implement various embodiments of the zone dimming method based on image region features.

[0132] The network interface 63 may include a wireless network interface or a wired network interface. The network interface 63 is generally used to establish a communication connection between the display device 6 and other electronic devices.

[0133] The present application also provides another embodiment, namely, providing a computer-readable storage medium, which stores a computer program. The computer program can be executed by at least one processor to enable the at least one processor to perform the steps of the above-mentioned partition dimming method based on image area features.

[0134] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus the necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods of each embodiment of the present application.

[0135] Obviously, the embodiments described above are only some of the embodiments of the present application, rather than all of the embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application has been described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions described in the aforementioned specific embodiments, or to make equivalent replacements for some of the technical features therein. Any equivalent structure made using the contents of the present application specification and the accompanying drawings, directly or indirectly used in other related technical fields, is also within the scope of protection of the present application.

Claims

1. A zone dimming method based on image region features, characterized in that: include: Acquire an RGB image, and perform brightness grading on the RGB image to obtain different brightness level labels; Separating high-frequency components from the RGB image, and performing image detail classification based on grayscale information of the high-frequency components to obtain different detail level labels; Convert the RGB image to a YUV color space to construct a brightness matrix, and calculate a backlight value of each backlight partition based on the brightness matrix, the brightness level label, and the detail level label; A multidimensional state space is constructed according to the combined state of the brightness level label and the detail level label, S-curve parameters are optimized based on the multidimensional state space, and pixel compensation is performed based on the S-curve parameters and the backlight value to generate a target compensated image.

2. The method for zone dimming based on image region features according to claim 1, characterized in that: The step of acquiring an RGB image and performing brightness grading on the RGB image to obtain different brightness level labels includes: Acquire the RGB image, and convert the RGB image into a grayscale image; Calculating the grayscale value of each pixel in the grayscale image, and generating a grayscale histogram of the entire image based on the grayscale values; The average brightness is calculated according to the grayscale histogram of the entire image, and brightness is graded according to the average brightness to obtain different brightness level labels.

3. The method for zone dimming based on image region features according to claim 1, wherein: The separating of the high-frequency components in the RGB image and performing image detail classification based on the grayscale information of the high-frequency components to obtain different detail level labels include: Filtering the RGB image using a Gaussian filter to generate a low-frequency component, and generating a high-frequency component based on the RGB image and the low-frequency component; Converting the high frequency component into a grayscale matrix; A target threshold is generated based on the grayscale matrix, and the Otsu algorithm is used to perform image detail classification based on the target threshold to obtain different detail level labels.

4. The method for zone dimming based on image region features according to claim 3, wherein: Generating a target threshold based on the grayscale matrix, and using the Otsu algorithm to perform image detail classification based on the target threshold to obtain different detail level labels, includes: Calculating a grayscale histogram of the high-frequency component based on the grayscale matrix to obtain a target grayscale histogram, and determining the frequency and number of each grayscale level based on the target grayscale histogram; Calculate the inter-class variance when the gray level is a threshold based on the gray level frequency and the frequency, and use the maximum inter-class variance as the target threshold; Segmenting the RGB image using the target threshold to generate high grayscale areas and low grayscale areas; The Otsu algorithm is re-adopted to respectively determine the high detail threshold and the low detail threshold of the high grayscale area and the low grayscale area, and image details of the high grayscale area and the low grayscale area are respectively classified based on the high detail threshold and the low detail threshold to obtain different detail level labels.

5. The method for zone dimming based on image region features according to claim 1, characterized in that: The converting the RGB image into a YUV color space to construct a brightness matrix, and calculating the backlight value of each backlight partition based on the brightness matrix, the brightness level label, and the detail level label, includes: Converting the RGB image to the YUV color space to construct the brightness matrix; Calculate the target average brightness, maximum brightness and brightness difference of each backlight partition based on the brightness matrix; Calculating a detail content ratio in the image based on the brightness level label and the detail level label, and calculating a dimming factor based on the detail content ratio in the image; The backlight value of each of the backlight partitions is calculated based on the dimming factor, the target average brightness, the maximum brightness, and the brightness difference.

6. The method for zone dimming based on image region features according to any one of claims 1 to 5, characterized in that: The step of constructing a multidimensional state space according to the combined state of the brightness level label and the detail level label, optimizing S-curve parameters based on the multidimensional state space, and performing pixel compensation based on the S-curve parameters and the backlight value to generate a target compensated image includes: constructing the multidimensional state space according to the combined state of the brightness level label and the detail level label; Initialize the network model and experience replay buffer, where the network model includes the Actor network, Critic network, and target network; Taking the multidimensional state space as an input state, performing model training based on the input state through the network model, and determining the S-curve parameters through the trained target network; The pixel values ​​of the RGB image are compensated by using the S-curve parameters and the backlight value to generate the target compensated image.

7. The method for zone dimming based on image region features according to claim 6, wherein: The method of using the multidimensional state space as an input state, performing model training based on the input state by the network model, and determining the S-curve parameters by the trained target network includes: Using the multidimensional state space as the input state, inputting the input state into an Actor network, so that the Actor network outputs current optimal S-curve parameters according to the current state, and using the current optimal S-curve parameters as actions; Adjusting liquid crystal pixels based on the current optimal S-curve parameters to generate an adjusted image, and calculating a peak signal-to-noise ratio based on the adjusted image, using the peak signal-to-noise ratio as a reward; Storing the input state, the action, and the reward in the experience replay buffer; Randomly sampling experience data from the experience replay buffer, and updating the parameters of the Actor network and the Critic network based on the experience data, and updating the parameters of the objective function using a soft update method to generate the trained target network; The S-curve parameters are determined according to the trained target network.

8. A zone dimming device based on image area features, characterized in that: include: A brightness grading module is used to obtain an RGB image and perform brightness grading on the RGB image to obtain different brightness level labels; a detail classification module, configured to separate high-frequency components from the RGB image and classify image details based on the grayscale information of the high-frequency components to obtain different detail level labels; a backlight value calculation module, configured to convert the RGB image into a YUV color space to construct a brightness matrix, and calculate a backlight value of each backlight partition based on the brightness matrix, the brightness level label, and the detail level label; A compensation image generation module is used to construct a multidimensional state space according to the combination state of the brightness level label and the detail level label, optimize S-curve parameters based on the multidimensional state space, and perform pixel compensation based on the S-curve parameters and the backlight value to generate a target compensated image.

9. A display device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the zone dimming method based on image area features as claimed in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the partitioned dimming method based on image area features according to any one of claims 1 to 7 is implemented.

Images