An image segmentation method and system for double-layer liquid crystal screen display

Abstract

The application discloses an image segmentation method and system for double-layer liquid crystal screen display, the double-layer liquid crystal screen is a front liquid crystal screen close to an audience and a rear liquid crystal screen close to a backlight module, and the method comprises the following steps: constructing a convolutional neural network model, the convolutional neural network model is used for processing an input image to obtain a first image and a second image, and the second image is sent to the rear liquid crystal screen, wherein the first image is used for being sent to the front liquid crystal screen; through the first image, the second image is subjected to image reconstruction, the image reconstruction is used for improving the artifact problem of double-screen liquid crystal display and improving the image display quality; and the second image after reconstruction is sent to the rear liquid crystal screen; the application has great optimization effects on the artifact phenomenon and the image quality of the double-layer screen, and the calculation speed is greatly improved compared with an image segmentation algorithm based on viewpoint compensation.

Description

Technical Field

[0001] This invention relates to the field of liquid crystal display technology, and more specifically, to an image segmentation method and system for dual-layer liquid crystal display. Background Technology

[0002] High dynamic range (HDR) allows viewers to see more accurate and precise details. Traditional LCDs, due to inherent light leakage, cannot achieve HDR, a significant challenge posed by OLED displays. Dual-layer LCDs use a highly transparent optical adhesive to bond two LCD panels together, effectively reducing light leakage and improving contrast. Because of the two panels, the displayed image must be segmented into two sub-images, transmitted to the back panel (near the backlight unit) and the front panel (near the viewer), respectively. Current image segmentation algorithms essentially blur the image sent to the back panel (near the backlight), creating small partitions with uniform brightness. While blur-based image segmentation algorithms are simple, they produce artifacts when viewed at an angle (at a certain viewing angle), affecting display quality. Another image segmentation algorithm based on viewpoint compensation eliminates parallax errors while considering display quality, but its computational load is enormous, hindering fast processing. Summary of the Invention

[0003] To address the aforementioned problems, the present invention aims to provide an image segmentation method and system for dual-layer liquid crystal displays, which avoids artifacts when viewed at an angle in existing dual-layer liquid crystal displays and improves display quality.

[0004] To achieve the above technical objectives, this application provides an image segmentation method for dual-layer LCD screen displays, wherein the dual-layer LCD screens are a front LCD screen closer to the viewer and a rear LCD screen closer to the backlight module, comprising the following steps:

[0005] A convolutional neural network model is constructed to process the input image to obtain a first image and a second image, and the second image is sent to the rear LCD screen, wherein the first image is used to be sent to the front LCD screen.

[0006] The second image is reconstructed from the first image. The image reconstruction is used to improve the artifact problem of dual-screen LCD display and improve the image display quality.

[0007] The reconstructed second image will be sent to the LCD screen.

[0008] Preferably, in the process of constructing the convolutional neural network model, a training dataset for training the convolutional neural network model is constructed based on the input image, wherein the images in the training dataset are all images of the same size with vertical lines.

[0009] Preferably, during the construction of the training dataset, each image in the training dataset is divided into 3 channels, and each image contains 10 blocks. Before testing, the order of the images in the training dataset is shuffled 6 times.

[0010] Preferably, in the process of constructing the convolutional neural network model, the convolutional neural network model includes an input layer, a convolutional layer and an output layer, wherein the convolutional neural network model consists of 8 sequentially stacked two-dimensional convolutions;

[0011] The input layer consists of five 3x3 convolutional kernels;

[0012] The convolutional layer consists of 64 3x3 convolutional kernels, and the convolutional layer is a Conv convolutional layer;

[0013] The output layer consists of two 64-fold 3x3 convolutional kernels.

[0014] Preferably, in the process of constructing the convolutional neural network model, the layers of the convolutional neural network model are fully connected. The fully connected layer is used to represent the feature map of the previous layer of the convolutional neural network model, which serves as the input of the current layer, and the feature map of the current layer serves as the input of the next layer.

[0015] Preferably, during the image reconstruction of the second image, the pixel values ​​of the first and second images are normalized and then multiplied by a shift to obtain the reconstructed image. The shift multiplication is performed five times, with the first image remaining stationary. In the first two of the first to fifth multiplications, the second image is positioned to the left of the first image, in the third multiplication it is directly opposite the first image, and in the fourth and fifth multiplications it is positioned to the right of the first image. The displacement distance is equal each time.

[0016] Preferably, during the process of acquiring the reconstructed image, the loss function between the reconstructed image and the original image is obtained and used as the loss function of the convolutional neural network model;

[0017] The loss function is expressed as:

[0018]

[0019] Among them, I input Represents the input image, I ref This represents the reconstructed image after the input image has been processed by the network. n represents the number of pixels in the width and height of the input image. Loss represents the loss function. The input image is rectangular.

[0020] Preferably, during the acquisition of the reconstructed image, the reconstructed image is evaluated using the peak signal-to-noise ratio (PSNR), wherein the formula for calculating the PSNR is:

[0021] Where n is the number of bits per pixel, and MES is the mean square error;

[0022] The calculation formula for MES is:

[0023] Where MES represents the mean square error of I(i,j) and K(i,j), I(i,j) and K(i,j) represent the gray values ​​of the processed image and the original image at pixel (i,j) respectively, i and j represent the pixel coordinate positions, M is the height of the image and N is the width of the image.

[0024] This invention also discloses an image segmentation system for dual-layer LCD screen display, wherein the two LCD screens are a front LCD screen closer to the viewer and a rear LCD screen closer to the backlight module, comprising:

[0025] The image acquisition module is used to acquire the input image;

[0026] The image processing module is used to process the input image by constructing a convolutional neural network model to obtain a first image and a second image, and to send the first image to the front LCD screen;

[0027] The image reconstruction module is used to reconstruct the second image from the first image and send the reconstructed second image to the rear LCD screen. The image reconstruction is used to improve the artifact problem of dual-screen LCD display and improve the image display quality.

[0028] Preferably, the image segmentation system further includes:

[0029] Data storage module, used to store the first image and the second image;

[0030] The communication module is used to send the first image and the second image to the control system of the dual-layer LCD screen, wherein the control system is used to represent the system or program that controls the dual-layer LCD screen to display images.

[0031] The communication module is also used to receive the first image;

[0032] Image segmentation systems are also used as control systems for dual-layer LCD screens.

[0033] The present invention discloses the following technical effects:

[0034] This invention effectively improves the artifact problem in dual-screen LCD displays and enhances image display quality, while significantly increasing image processing speed. The proposed image segmentation algorithm utilizes a convolutional neural network (CNN). Through training on a large dataset, it simulates the real image of the display as seen by the human eye from different viewpoints and compares it with a test image. By forward propagation and updating the CNN parameters, it considers the display effects from multiple viewpoints. Compared to existing blurring algorithms, this invention significantly optimizes both artifact phenomena and image quality in dual-screen displays, and also offers a substantial improvement in computational speed compared to viewpoint-compensation-based image segmentation algorithms. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments 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 drawings can be obtained based on these drawings without creative effort.

[0036] Figure 1 This is a schematic diagram of the convolutional neural network structure described in this invention;

[0037] Figure 2 This is a schematic diagram of the method steps described in this invention;

[0038] Figure 3 This is a schematic diagram of the system structure described in this invention. Detailed Implementation

[0039] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0040] like Figure 1-3 As shown, this invention provides an image segmentation method for dual-layer LCD screen displays, wherein the dual-layer LCD screens are a front LCD screen closer to the viewer and a rear LCD screen closer to the backlight module, comprising the following steps:

[0041] A convolutional neural network model is constructed to process the input image to obtain a first image and a second image, and the second image is sent to the rear LCD screen, wherein the first image is used to be sent to the front LCD screen.

[0042] The second image is reconstructed from the first image. The image reconstruction is used to improve the artifact problem of dual-screen LCD display and improve the image display quality.

[0043] The reconstructed second image will be sent to the LCD screen.

[0044] More preferably, in the process of constructing the convolutional neural network model, a training dataset for training the convolutional neural network model is constructed based on the input image, wherein the images in the training dataset are all images of the same size with vertical lines.

[0045] More preferably, during the construction of the training dataset, each image in the training dataset is divided into 3 channels, and each image contains 10 blocks. Before testing, the order of the images in the training dataset is shuffled 6 times.

[0046] More preferably, in the process of constructing the convolutional neural network model, the convolutional neural network model includes an input layer, a convolutional layer and an output layer, wherein the convolutional neural network model consists of 8 sequentially stacked two-dimensional convolutions;

[0047] The input layer consists of five 3x3 convolutional kernels;

[0048] The convolutional layer consists of 64 3x3 convolutional kernels, and the convolutional layer is a Conv convolutional layer;

[0049] The output layer consists of two 64-fold 3x3 convolutional kernels.

[0050] More preferably, in the process of constructing the convolutional neural network model, the layers of the convolutional neural network model are fully connected. The fully connected layer is used to represent the feature map of the previous layer of the convolutional neural network model, which serves as the input of the current layer, and the feature map of the current layer serves as the input of the next layer.

[0051] More preferably, during the image reconstruction of the second image, the pixel values ​​of the first and second images are normalized and then multiplied by a shift to obtain the reconstructed image. The shift multiplication is performed five times, with the first image remaining stationary. In the first two of the first to fifth multiplications, the second image is positioned to the left of the first image, in the third multiplication it is directly opposite the first image, and in the fourth and fifth multiplications it is positioned to the right of the first image. The displacement distance is equal each time.

[0052] More preferably, during the process of acquiring the reconstructed image, the loss function between the reconstructed image and the original image is obtained and used as the loss function of the convolutional neural network model;

[0053] The loss function is expressed as:

[0054]

[0055] Among them, I input Represents the input image, I ref This represents the reconstructed image after the input image has been processed by the network. n represents the number of pixels in the width and height of the input image. Loss represents the loss function. The input image is rectangular.

[0056] More preferably, during the acquisition of the reconstructed image, the reconstructed image is evaluated using the peak signal-to-noise ratio (PSNR), wherein the formula for calculating the PSNR is:

[0057]

[0058] Where n is the number of bits per pixel, and MES is the mean square error;

[0059] The calculation formula for MES is:

[0060] Where MES represents the mean square error of I(i,j) and K(i,j), I(i,j) and K(i,j) represent the gray values ​​of the processed image and the original image at pixel (i,j) respectively, i and j represent the pixel coordinate positions, M is the height of the image and N is the width of the image.

[0061] This invention also discloses an image segmentation system for dual-layer LCD screen display, wherein the two LCD screens are a front LCD screen closer to the viewer and a rear LCD screen closer to the backlight module, comprising:

[0062] The image acquisition module is used to acquire the input image;

[0063] The image processing module is used to process the input image by constructing a convolutional neural network model to obtain a first image and a second image, and to send the first image to the front LCD screen;

[0064] The image reconstruction module is used to reconstruct the second image from the first image and send the reconstructed second image to the rear LCD screen. The image reconstruction is used to improve the artifact problem of dual-screen LCD display and improve the image display quality.

[0065] More preferably, the image segmentation system further includes:

[0066] Data storage module, used to store the first image and the second image;

[0067] The communication module is used to send the first image and the second image to the control system of the dual-layer LCD screen, wherein the control system is used to represent the system or program that controls the dual-layer LCD screen to display images.

[0068] The communication module is also used to receive the first image;

[0069] Image segmentation systems are also used as control systems for dual-layer LCD screens.

[0070] Example 1: This invention provides an image segmentation algorithm based on convolutional neural networks for dual-layer LCD screen displays, wherein the two LCD screens are a front LCD screen closer to the viewer and a rear LCD screen closer to the backlight module. The method specifically includes the following steps:

[0071] Step (1): Construct a training dataset and perform preprocessing operations on the training dataset;

[0072] Step (2): Use 8 sequentially stacked 2D convolutions to form a convolutional neural network model;

[0073] Step (3): After steps (1) and (2), two test images are generated. The first image represents the front LCD screen close to the audience, and the second image represents the rear LCD screen close to the backlight module. The pixel values ​​of the two images are normalized and then multiplied by offset to obtain the reconstructed image. The offset multiplication is performed 5 times in total. The first image remains unchanged. The second image is located to the left of the first image in the first two times from the first to the fifth time, directly facing the first image in the third time, and to the right of the first image in the fourth and fifth times. The displacement distance is equal each time.

[0074] Step (4): Set the loss function of the network and optimize the loss function;

[0075] Step (5): Perform image processing based on the trained network model and evaluate the reconstructed image using peak signal-to-noise ratio;

[0076] Step (1) involves preprocessing the training dataset, including the following steps:

[0077] 1.1: Select training samples from the training dataset as the original training set. The images in the training samples are all of the same size and contain vertical lines.

[0078] 1.2: Divide each image in the training dataset into 3 channels, and each image contains 10 blocks. Shuffle the order of all images in the dataset 6 times before testing.

[0079] The network model consists of an input layer, a convolutional layer, and an output layer. In the network model, each layer is fully connected. The feature map of the previous layer is used as the input of the current layer, and the feature map of the current layer is used as the input of the next layer.

[0080] The input layer consists of five 3x3 convolutional kernels;

[0081] The intermediate convolutional layer consists of 64 3x3 convolutional kernels;

[0082] The output layer consists of two convolutional kernels, each containing 64 3x3 kernels.

[0083] The convolutional layer is a Conv convolutional layer;

[0084] The loss function of the 5 reconstructed images obtained in step (3) and the original test image is: Where I input Represents the input image, I ref This represents the image reconstructed after the input image is processed by the network, where n represents the number of pixels in the width and height of the input image (the input image is square), and Loss represents the loss function.

[0085] The formula for calculating peak signal-to-noise ratio is:

[0086] Where n is the number of bits per pixel, and MES is the mean squared error.

[0087] The calculation formula for MES is:

[0088] Where MES represents the mean square error of I(i,j) and K(i,j), I(i,j) and K(i,j) represent the grayscale values ​​of the processed image and the original image at pixel (i,j), respectively, where i and j represent the pixel coordinates, M is the height of the image, and N is the width of the image.

[0089] This invention effectively improves the artifact problem in dual-screen LCD displays and enhances image display quality, while significantly increasing image processing speed. The proposed image segmentation algorithm utilizes a convolutional neural network (CNN). Through training on a large dataset, it simulates the real image of the display as seen by the human eye from different viewpoints and compares it with a test image. By forward propagation and updating the CNN parameters, it considers the display effects from multiple viewpoints. Compared to existing blurring algorithms, this invention significantly optimizes both artifact phenomena and image quality in dual-screen displays, and also offers a substantial improvement in computational speed compared to viewpoint-compensation-based image segmentation algorithms.

[0090] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. In addition, the terms "first", "second", "third", etc. are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0091] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention. All should be covered 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. An image segmentation method for a dual-layer LCD screen display, wherein the dual-layer LCD screen is a front LCD screen closer to the viewer and a rear LCD screen closer to the backlight module, characterized in that, Includes the following steps: A convolutional neural network model is constructed, which is used to process the input image to obtain a first image and a second image, and send the second image to the rear LCD screen, wherein the first image is used to be sent to the front LCD screen; Using the first image, the second image is reconstructed. The image reconstruction is used to improve the artifact problem of dual-screen liquid crystal display and improve the image display quality. The reconstructed second image is sent to the rear LCD screen; In the process of building a convolutional neural network model, a training dataset for training the convolutional neural network model is constructed based on the input image, wherein the images in the training dataset are all images of the same size with vertical lines; In the process of constructing the training dataset, each image in the training dataset is divided into 3 channels, and each image contains 10 blocks. Before testing, the order of the images in the training dataset is shuffled 6 times. In the process of constructing a convolutional neural network model, the convolutional neural network model includes an input layer, a convolutional layer and an output layer, wherein the convolutional neural network model consists of 8 sequentially stacked two-dimensional convolutions; The input layer consists of five 3x3 convolutional kernels; The convolutional layer comprises 64 3x3 convolutional kernels, wherein the convolutional layer is a Conv convolutional layer; The output layer includes two 64-fold 3x3 convolutional kernels; In the process of constructing a convolutional neural network model, the layers of the convolutional neural network model are fully connected. The fully connected layer is used to represent the feature map of the previous layer of the convolutional neural network model, which serves as the input of the current layer. The feature map of the current layer serves as the input of the next layer. In the process of reconstructing the second image, the pixel values ​​of the first image and the second image are normalized and then multiplied by a staggered method to obtain the reconstructed image. The staggered multiplication is performed five times. The first image remains stationary. The second image is located to the left of the first image in the first two of the first to fifth multiplications, directly opposite the first image in the third multiplication, and to the right of the first image in the fourth and fifth multiplications. The displacement distance is equal each time.

2. The image segmentation method for dual-layer liquid crystal display according to claim 1, characterized in that, During the process of acquiring the reconstructed image, the loss function between the reconstructed image and the original image is obtained and used as the loss function of the convolutional neural network model; The loss function is expressed as: Among them, I input Represents the input image, I ref The input image is reconstructed after processing by the network, where n represents the length and width of the input image in pixels, and Loss represents the loss function. The input image is rectangular.

3. The image segmentation method for dual-layer liquid crystal display according to claim 2, characterized in that, During the acquisition of the reconstructed image, the reconstructed image is evaluated using the Peak Signal-to-Noise Ratio (PSNR), wherein the formula for calculating the PSNR is: Where n is the number of bits per pixel, and MES is the mean square error; The calculation formula for MES is: Where MES represents the mean square error of I(i,j) and K(i,j), I(i,j) and K(i,j) represent the gray values ​​of the processed image and the original image at pixel (i,j) respectively, i and j represent the pixel coordinate positions, M is the height of the image and N is the width of the image.

4. An image segmentation system for dual-layer LCD screen display, wherein the two LCD screens are a front LCD screen closer to the viewer and a rear LCD screen closer to the backlight module, characterized in that, include: The image acquisition module is used to acquire the input image; The image processing module is used to process the input image by constructing a convolutional neural network model to obtain a first image and a second image, and to send the first image to the front LCD screen; In the process of building a convolutional neural network model, a training dataset for training the convolutional neural network model is constructed based on the input image, wherein the images in the training dataset are all images of the same size with vertical lines; In the process of constructing the training dataset, each image in the training dataset is divided into 3 channels, and each image contains 10 blocks. Before testing, the order of the images in the training dataset is shuffled 6 times. In the process of constructing a convolutional neural network model, the convolutional neural network model includes an input layer, a convolutional layer and an output layer, wherein the convolutional neural network model consists of 8 sequentially stacked two-dimensional convolutions; The input layer consists of five 3x3 convolutional kernels; The convolutional layer comprises 64 3x3 convolutional kernels, wherein the convolutional layer is a Conv convolutional layer; The output layer includes two 64-fold 3x3 convolutional kernels; In the process of constructing a convolutional neural network model, the layers of the convolutional neural network model are fully connected. The fully connected layer is used to represent the feature map of the previous layer of the convolutional neural network model, which serves as the input of the current layer. The feature map of the current layer serves as the input of the next layer. An image reconstruction module is used to reconstruct the second image from the first image and send the reconstructed second image to the rear LCD screen. The image reconstruction is used to improve the artifact problem of the dual-screen LCD display and improve the image display quality. In the process of reconstructing the second image, the pixel values ​​of the first image and the second image are normalized and then multiplied by a staggered method to obtain the reconstructed image. The staggered multiplication is performed five times. The first image remains stationary. The second image is located to the left of the first image in the first two of the first to fifth multiplications, directly opposite the first image in the third multiplication, and to the right of the first image in the fourth and fifth multiplications. The displacement distance is equal each time.

5. The image segmentation system for dual-layer liquid crystal display according to claim 4, characterized in that, The image segmentation system further includes: Data storage module, used to store the first image and the second image; A communication module is used to send the first image and the second image to the control system of the dual-layer LCD screen, wherein the control system is used to represent a system or program that controls the dual-layer LCD screen to display images; The communication module is also used to receive the first image; The image segmentation system is also used as the control system for the dual-layer liquid crystal screen.

Images