Image display method and apparatus, electronic device, and storage medium

Abstract

The present application provides an image display method and apparatus, an electronic device, and a storage medium. The method comprises: acquiring a color lookup table, the color lookup table being a mapping relationship, which is learned by a neural network model, between an original image and a target image, the brightness of the target image being lower than the brightness of the original image, and the difference value between the contrast of the target image and the contrast of the original image being smaller than a target threshold; and on the basis of the color lookup table, performing mapping processing on an image to be displayed, so as to obtain a displayed image. According to the method, the brightness of the displayed image is reduced, and the difference value between the contrast of the displayed image and the contrast of the image to be displayed is smaller than the target threshold; moreover, the brightness reduction of the displayed image reduces the power consumption, and the difference value between the contrast of the displayed image and the contrast of the image to be displayed is smaller than the target threshold, so that the quality of the displayed image is ensured, thereby realizing low-power-consumption and high-quality image display.

Description

Image display method, device, electronic device and storage medium Technical Field

[0001] The present application relates to the field of computer vision technology, and in particular to an image display method, device, electronic device, and storage medium. Background Art

[0002] Organic Light Emitting Diode (OLED) is widely used in smart terminal devices for image display due to its high contrast, fast response speed, saturation, and flexible bending. The power consumption of OLED display screen (i.e., smart terminal device screen) accounts for the main part of the energy consumption of the entire smart terminal device.

[0003] In related technologies, the primary approach to reducing energy consumption in OLED displays is to lower the pixel brightness of the displayed image. However, this reduction in pixel brightness can affect the quality of the resulting image. Therefore, while this approach reduces power consumption, it cannot guarantee the quality of the resulting image, resulting in poor image quality after power reduction. Therefore, achieving low-power and high-quality image display is a pressing technical issue.

[0004] Summary of the Invention

[0005] In view of the above problems, embodiments of the present application provide an image display method, device, electronic device, and storage medium to overcome the above problems or at least partially solve the above problems.

[0006] In a first aspect of the embodiments of the present application, a method for displaying an image is disclosed, the method comprising:

[0007] Obtaining a color lookup table, where the color lookup table is a mapping relationship between an original image and a target image learned by a neural network model, wherein the brightness of the target image is lower than the brightness of the original image, and the difference between the contrast of the target image and the contrast of the original image is less than a target threshold;

[0008] According to the color lookup table, mapping processing is performed on the image to be displayed to obtain a display image.

[0009] According to a second aspect of the embodiments of the present application, an image display device is disclosed, comprising:

[0010] A first acquisition module is configured to acquire a color lookup table, wherein the color lookup table is a mapping relationship between an original image and a target image learned through a neural network model, wherein the brightness of the target image is lower than the brightness of the original image, and the difference between the contrast of the target image and the contrast of the original image is less than a target threshold;

[0011] The first mapping module is configured to perform mapping processing on the image to be displayed according to the color lookup table to obtain a display image.

[0012] A third aspect of an embodiment of the present application discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and runnable on the processor. When the processor executes the computer program, the steps of the image display method described in the first aspect of the embodiment of the present application are implemented.

[0013] A fourth aspect of the embodiments of the present application discloses a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the image display method described in the first aspect of the embodiments of the present application are implemented.

[0014] The embodiments of the present application include the following advantages:

[0015] In an embodiment of the present application, a color lookup table is obtained, and a mapping process is performed on the image to be displayed according to the color lookup table to obtain a display image. Since the color lookup table is a mapping relationship between the original image and the target image learned by a neural network model, and the brightness of the target image is lower than the brightness of the original image, and the difference between the contrast of the target image and the contrast of the original image is less than the target threshold. Therefore, the brightness of the display image obtained according to the color lookup table is also lower than the brightness of the image to be displayed, and the difference between the contrast and the contrast of the image to be displayed is less than the target threshold. The reduction in the brightness of the display image achieves a reduction in power consumption, and the difference between the contrast of the display image and the contrast of the image to be displayed is less than the target threshold, which ensures the quality of the display image. In this way, low-power and high-quality image display is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0017] FIG1 is a flowchart of the steps of an image display method provided in an embodiment of the present application;

[0018] FIG2 is a schematic diagram of a process for generating a true value image according to an embodiment of the present application;

[0019] FIG3 is a schematic diagram of the structure of a ground truth image generation model provided in an embodiment of the present application;

[0020] FIG4 is a flowchart of another image display method provided by an embodiment of the present application;

[0021] FIG5 is a schematic diagram of a mapping curve provided in an embodiment of the present application;

[0022] FIG6 is a flowchart of another image display method provided in an embodiment of the present application;

[0023] FIG. 7 is a flow chart of an image display method of a multi-color lookup table provided in an embodiment of the present application.

[0024] FIG8 is a schematic diagram of the structure of a neural network model provided in an embodiment of the present application;

[0025] FIG9 is a schematic diagram of the structure of another neural network model provided in an embodiment of the present application;

[0026] FIG10 is a schematic diagram of analysis results of a sculpture scene image provided by an embodiment of the present application;

[0027] FIG11 is a schematic diagram of an analysis result of a plant scene image provided in an embodiment of the present application;

[0028] FIG12 is a schematic diagram of an analysis result of an animal scene image provided in an embodiment of the present application;

[0029] FIG13 is a schematic structural diagram of an image display device provided in an embodiment of the present application;

[0030] FIG14 is a schematic structural diagram of an electronic device provided in an embodiment of the present application. Specific embodiments

[0031] To make the above-mentioned purposes, features, and advantages of this application more clearly understood, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of this application.

[0032] To facilitate understanding of the technical solutions provided by this application, the following is a brief description of the OLED display screen. In OLED display technology, low power consumption and high contrast are two key core advantages. Compared with traditional liquid crystal displays, OLED displays do not require a backlight source but use organic light-emitting materials, so they consume less energy. At the same time, they do not require a liquid crystal layer to control the degree of light transmittance. When displaying deep black, there will be no backlight leakage, and a deeper black can be presented, thereby achieving a higher contrast. In addition, the OLED display screen also has pixel-level lighting control capabilities, that is, only the pixels that need to be displayed will be activated, and other pixels will remain off, which also helps to reduce screen-side power consumption, and by improving and optimizing the molecular structure of organic materials, higher quantum efficiency and lower display power consumption can be achieved.

[0033] In smart terminal devices, the power consumption of OLED displays still accounts for a major portion of the overall energy consumption of the smart terminal device. Currently, the main approach to reducing the energy consumption of OLED displays is to reduce the pixel brightness of the displayed image, but this reduction in pixel brightness can affect the quality of the resulting displayed image. Therefore, to overcome the limitations of existing technologies, the present invention provides an image display method that effectively reduces power consumption while ensuring high-quality image display. The image display method of the present invention is described in detail below.

[0034] An image display method provided by an embodiment of the present application is shown in FIG1 , which is a flowchart of the steps of an image display method provided by an embodiment of the present application. As shown in FIG1 , an image display method provided by an embodiment of the present application may include steps S110 to S120:

[0035] Step S110: Obtain a color lookup table, where the color lookup table is a mapping relationship between an original image and a target image learned through a neural network model, the brightness of the target image is lower than the brightness of the original image, and the difference between the contrast of the target image and the contrast of the original image is less than a target threshold.

[0036] Step S120: performing mapping processing on the image to be displayed according to the color lookup table to obtain a display image.

[0037] In the embodiments of the present application, the original image is an image with high brightness and contrast, and the original image does not meet the requirements of low-power image display. In practical applications, the original image can be an image directly captured by an image acquisition device (e.g., a camera) or an image obtained from an image database (e.g., the DIV2K database).

[0038] The target image refers to an image that meets the requirements of low power consumption and high-quality display. The brightness of the target image is lower than the brightness of the original image, and the difference between the contrast of the target image and the contrast of the original image is less than the target threshold. Among them, the target threshold is a smaller difference set according to the high-quality image display requirements. If the image quality requirement is higher, the target threshold is set smaller. The difference between the contrast of the target image and the contrast of the original image is less than the target threshold, which means that the contrast of the target image has not changed significantly relative to the contrast of the original image. In some embodiments, the contrast of the target image is lower than the contrast of the original image, and the contrast value of the target image below the original image does not exceed the target threshold. In other embodiments, the contrast of the target image is higher than the contrast of the original image, and the contrast value of the target image above the original image does not exceed the target threshold. In other words, the target image is an image with low brightness and high contrast relative to the original image, and the target image can meet the requirements of low power consumption and high-quality display.

[0039] The color lookup table is a mapping relationship between the original image and the target image learned by the neural network model. Therefore, the color lookup table can process the image to be displayed into a low-brightness and high-contrast display image, that is, the brightness of the display image is lower than the brightness of the image to be displayed, and the contrast of the display image is based on the contrast consistency of the image to be displayed.

[0040] A color lookup table is a mapping table of pixel grayscale values. The color lookup table uses pixel grayscale values ​​as indexes and the values ​​after the grayscale values ​​are mapped as the table contents. Mapping the image to be displayed based on the color lookup table to obtain the displayed image specifically includes: using the grayscale value of a pixel in the image to be displayed as an index, finding the value after the grayscale value is mapped from the color mapping table as the mapped grayscale value of the pixel, traversing all pixels in the image to be displayed, obtaining the mapped grayscale value of each pixel, and then obtaining the displayed image based on the grayscale values ​​of all pixels.

[0041] For example, the image to be displayed has 64 pixels. The grayscale values ​​of the 64 pixels are used as indexes respectively. The grayscale values ​​mapped are found in the color lookup table as the grayscale values ​​of the pixels after mapping. The displayed image is obtained according to the grayscale values ​​of the 64 pixels after mapping.

[0042] In a specific embodiment, mapping processing is performed on the image to be displayed according to the color lookup table to obtain a display image, including: extracting a brightness channel image and a color channel image from the image to be displayed, and processing the brightness channel image into a low-brightness brightness channel image; mapping processing is performed on the low-brightness brightness channel image according to the color lookup table to obtain a mapped brightness channel image; and fusing the color channel image of the image to be displayed with the mapped brightness channel image to obtain a display image.

[0043] In an embodiment of the present application, the image to be displayed is an RGB image, and extracting a luminance channel image and a color channel image from the image to be displayed specifically includes: converting the display image from the RGB color model to the YUV color model, and then extracting the luminance channel image (i.e., the Y channel image) and the color channel image (i.e., the U and V channel images) from the converted image. Processing the luminance channel image into a low-brightness luminance channel image specifically includes: multiplying the luminance channel image by a luminance coefficient to obtain a low-brightness luminance channel image, wherein the luminance coefficient is a number between 0.5 and 0.7. Since the mapping process is based on the low-brightness luminance channel image, the brightness of the mapped luminance channel image is reduced, while the color channel image remains unchanged, so that the final generated display image is a low-brightness and high-contrast image.

[0044] In summary, in the embodiment of the present application, the image to be displayed is mapped according to the color lookup table to obtain a display image. Since the color lookup table is a mapping relationship between the original image and the target image learned by the neural network model, and the brightness of the target image is lower than the brightness of the original image, and the difference between the contrast of the target image and the contrast of the original image is less than the target threshold. Therefore, the brightness of the display image obtained according to the color lookup table is also lower than the brightness of the image to be displayed, and the difference between the contrast and the contrast of the image to be displayed is less than the target threshold. The reduction in the brightness of the display image achieves a reduction in power consumption, and the difference between the contrast of the display image and the contrast of the image to be displayed is less than the target threshold, which ensures the quality of the display image. In this way, low-power and high-quality image display is achieved.

[0045] The color lookup table of the embodiment of the present application can be obtained by performing unsupervised training on a neural network model, or by performing supervised training. The two methods are described below.

[0046] Method 1: Unsupervised training to obtain a color lookup table:

[0047] In which, the neural network model includes: a first image processing model, the first image processing model is trained to obtain a generated image with low brightness and high contrast, the low brightness means that the brightness of the generated image is lower than the brightness of the original image, and the high contrast means that the difference between the contrast of the generated image and the contrast of the original image is less than the target threshold.

[0048] Specifically, the color lookup table and the first image processing model are trained by the following steps A1 to A4:

[0049] Step A1: Obtain a color lookup table to be trained, and obtain the original image.

[0050] Step A2: inputting the original image into a first image processing model to be trained for unsupervised training to obtain a first generated image.

[0051] Step A3: Obtain a first loss value based on the first generated image and the original image.

[0052] Step A4: Update the model parameters of the first image processing model to be trained according to the first loss value to obtain the first image processing model, and update the color lookup table to be trained according to the first loss value to obtain the color lookup table.

[0053] In the embodiments of the present application, the first image processing model is trained to generate a low-brightness, high-contrast generated image. In some embodiments, the contrast of the generated image is lower than that of the original image, and the contrast value of the generated image below the original image does not exceed the target threshold. In other embodiments, the contrast of the generated image is higher than that of the original image, and the contrast value of the generated image above the original image does not exceed the target threshold.

[0054] The original images are obtained from an image dataset (e.g., DIV2K, Flickr2K). In a specific embodiment, to ensure the diversity of the original image data, the images obtained from the image dataset are randomly cropped to obtain a large number of partial images. The partial images are used as original images for model training to achieve data enhancement.

[0055] The color lookup table to be trained is updated with each training, and a color lookup table is obtained after the training is completed. The first image processing model processes the input original image according to the current color lookup table to be trained to obtain a first generated image. Specifically, the original image is converted from the RGB color model to the YUV color model, and the brightness channel image (i.e., the Y channel image) and the color channel images (i.e., the U and V channel images) are extracted from the converted image, and the brightness channel image is processed into a low-brightness brightness channel image; then, the first image processing model maps the low-brightness brightness channel image according to the color lookup table to be trained to obtain a mapped brightness channel image; the color channel image of the original image is fused with the mapped brightness channel image to obtain the first generated image.

[0056] In order to achieve the training goal of the first image processing model to obtain a generated image with low brightness and high contrast, the loss functions used in the unsupervised training include: a power consumption loss function, a structural information loss function, and a contrast loss function; wherein the power consumption loss function aims to reduce the brightness of the generated image, the structural information loss function aims to minimize the structural information loss of the generated image, and the contrast loss function aims to minimize the contrast loss of the generated image.

[0057] Specifically, the power consumption loss function is used to ensure that the power consumption saving rate of the generated image compared to the original image can be adjusted arbitrarily, while the overall grayscale value of the brightness channel image of the generated image is maintained at a certain ratio (for example, a ratio of 0.6) less than the overall grayscale value of the brightness channel of the original image, so as to achieve the purpose of reducing the image brightness.

[0058] For example, the power loss function L R Expressed as:

[0059] in, is the pixel grayscale mean of the brightness channel image, R is the pixel grayscale mean of the brightness channel image of the generated image, N represents the number of pixels, Y i Represents the brightness channel pixel value of the i-th pixel.

[0060] The structural information loss function is used to ensure that the structural information between the generated image and the original image is not lost in large quantities. For example, the structural information loss function L s Expressed as: L s =1-ssim|Xf(X D )|

[0061] Among them, X is the brightness channel image of the original image, f(X D ) is used to generate images, and ssim() is used to calculate X and f(X D )’s structural similarity.

[0062] The contrast loss function is used to ensure that the contrast of the generated image does not drop significantly compared to the original image. For example, the contrast loss function L c Expressed as: L c =|std(X)-std(f(X D ))|

[0063] Among them, std() is the standard deviation of the image information. The standard deviation can reflect the discrete degree of a data set. When used in the image field, it can represent the contrast of an image.

[0064] In the embodiment of the present application, based on the aforementioned power loss function, structural information loss function, and contrast loss function, power loss values, structural information loss values, and contrast loss values ​​are calculated for the first generated image and the original image, respectively. The power loss values, structural information loss values, and contrast loss values ​​are then added together to obtain a first loss value. The model parameters of the first image processing model to be trained are then updated based on the first loss value, enabling the first image processing model to generate generated images with low brightness and high contrast. Furthermore, the color lookup table to be trained is updated based on the first loss value, enabling the color lookup table to represent the mapping relationship between the original image and the target image.

[0065] Method 2: Supervised training method to obtain the color lookup table:

[0066] In which, the neural network model includes: a second image processing model, the second image processing model has the training goal of minimizing the loss between the generated image and the true value image, the brightness of the true value image is lower than the brightness of the original image, and the difference between the contrast of the true value image and the contrast of the original image is less than the target threshold.

[0067] Specifically, the color lookup table and the second image processing model are trained by the following steps B1 to B4:

[0068] Step B1: Obtain a color lookup table to be trained, and obtain an original image and a true value image corresponding to the original image.

[0069] Step B2: Input the original image into a second image processing model to be trained for supervised training to obtain a second generated image.

[0070] Step B3: Obtain a second loss value based on the second generated image and the true value image.

[0071] Step B4: Update the model parameters of the second image processing model to be trained according to the second loss value to obtain the second image processing model, and update the color lookup table to be trained according to the second loss value to obtain the color lookup table.

[0072] In an embodiment of the present application, the original image and the ground-truth image corresponding to the original image are used as an image pair to perform supervised training on the second image processing model. In some embodiments, the contrast of the ground-truth image is lower than the contrast of the original image, and the contrast value of the ground-truth image below the original image does not exceed the target threshold. In other embodiments, the contrast of the ground-truth image is higher than the contrast of the original image, and the contrast value of the ground-truth image above the original image does not exceed the target threshold.

[0073] During training, the color lookup table to be trained is updated with each training session. After training, a color lookup table is obtained. The second image processing model processes the input original image according to the current color lookup table to be trained to obtain a second generated image. Specifically, the second image processing model processes the luminance channel image of the original image into a low-luminance luminance channel image, and maps the low-luminance luminance channel image according to the color lookup table to be trained to obtain a mapped luminance channel image. The color channel image of the original image is fused with the mapped luminance channel image to obtain the second generated image.

[0074] A second loss value is obtained based on the second generated image and the true value image, including: calculating the loss value based on the pixel grayscale values ​​of the second generated image and the pixel grayscale values ​​of the true value image to obtain the second loss value, and then taking minimizing the loss between the generated image and the true value image as the training goal, updating the model parameters of the second image processing model to be trained based on the second loss value, so that the second image processing model can generate a generated image with low brightness and high contrast; updating the color lookup table to be trained based on the second loss value, so that the color lookup table can represent the mapping relationship between the original image and the target image.

[0075] Optionally, considering that there is no general dataset for low-brightness and high-contrast display, the present embodiment proposes a method for generating a true value image based on unsupervised training. Specifically, obtaining the true value image corresponding to the original image includes steps C1 and C2:

[0076] Step C1: Construct a true value image generation model, wherein the true value image generation model is obtained by performing unsupervised training with the generation of true value images as the training goal.

[0077] Step C2: Input the original image into the true value image generation model for processing to obtain the true value image.

[0078] In an embodiment of the present application, the loss functions used in the unsupervised training of the true image generation model include: power consumption loss function, structural information loss function and contrast loss function; wherein, the power consumption loss function aims to reduce the brightness of the generated image, the structural information loss function aims to minimize the structural information loss of the generated image, and the contrast loss function aims to minimize the contrast loss of the generated image.

[0079] Specifically, the specific expressions of the power consumption loss function, the structural information loss function and the contrast loss function are as described above and will not be described in detail here.

[0080] The power loss function ensures that the power saving rate of the generated image output by the true image generation model compared to the original image can be arbitrarily adjusted, while the overall grayscale value of the brightness channel image of the generated image is kept at a certain ratio lower than the overall grayscale value of the brightness channel of the original image to achieve the purpose of reducing the image brightness. The structural information loss function ensures that the structural information between the generated image output by the true image generation model and the original image is not lost in large quantities. The contrast loss function ensures that the contrast of the generated image output by the true image generation model is not significantly reduced relative to the original image. Therefore, the true image generation model trained according to the above power loss function, structural information loss function, and contrast loss function can obtain a low-brightness and high-contrast true image based on the original image.

[0081] As shown in Figure 2, the original image is input into the true value image generation model for processing to obtain the true value image, specifically including: extracting the brightness channel image and color channel image from the original image, and processing the brightness channel image into a low-brightness brightness channel image; inputting the low-brightness brightness channel image into the true value image generation model for processing to obtain a processed brightness channel image; and fusing the processed brightness channel image with the color channel image to obtain the true value image.

[0082] In a specific implementation method, the true value image generation model is a Unet model with a residual structure. Specifically, the true value image generation model includes an encoder and a decoder, and the encoder and the decoder are connected via a residual structure.

[0083] Furthermore, the original image is input into the true value image generation model for processing to obtain the true value image, including: using the encoder to extract features of the original image to obtain feature information; using the decoder to restore the feature information to obtain a restored image; and obtaining the true value image based on the restored image and the original image.

[0084] The encoder is responsible for extracting feature information, while the decoder is responsible for restoring it. As shown in Figure 3, the encoder consists of four feature extraction modules, each consisting of two convolutional layers and one maximum pooling layer. The decoder also consists of four feature restoration modules, each consisting of one upsampling layer and two convolutional layers. The encoder and decoder are connected via residual connections. By adding residual connections to the ground truth image generation model, image texture details can be reconstructed, thereby reducing the training difficulty and improving the training accuracy of the image generation model.

[0085] In the embodiments of the present application, the original image refers to the image to be displayed by the display device. Specifically, in order to enable the image display method in the embodiments of the present application to process images of any dimension and any scene type, the original image includes: images that meet the target resolution in different brightness scenes and images that meet the target resolution in different target object scenes. Different brightness scenes include: dark light scenes, bright light scenes, backlight scenes, etc.; different target object scenes include: human scenes, building scenes, plant scenes, etc. The target resolution refers to the resolution of the image, which is determined according to the display requirements of the display device. For example, if the display resolution requirement of the display device is 2K, the original image is an image with a 2K resolution.

[0086] Among them, images that meet the target resolution in different brightness scenes refer to: images of different brightness scenes corresponding to the same target object at the target resolution. Images that meet the target resolution in different target object scenes refer to: images of different target objects corresponding to the same brightness scene at the target resolution. In practical applications, in order to make the training results more accurate, images of different brightness scenes corresponding to a variety of different target objects (including images of different target objects corresponding to the same brightness scene) are used as original images for model training. For example, the target object scenes include: object A, object B and object C, and the brightness scenes include: brightness 1 and brightness 2; then the original images include: image of object A in brightness 1 scene, image of object A in brightness 2 scene, image of object B in brightness 1 scene, image of object B in brightness 2 scene, image of object C in brightness 1 scene, and image of object C in brightness 2 scene.

[0087] For example, to meet the low-power, high-contrast display requirements of a dual-FPGA, 95-inch OLED display device, the original image's luminance channel image (Y channel image) is first uniformly resized to 480×270 to meet the resolution requirement of the 95-inch OLED display device's 8K resolution after four downsampling. At the same time, the color channel images (U and V channel images) remain unchanged. In addition, to generate a low-power, high-contrast image, the luminance channel image is multiplied by a coefficient in the range of 0.5 to 0.7, processing it into a low-brightness luminance channel image and inputting it into the neural network model for training.

[0088] 4 , which is a flowchart of another method for displaying an image provided by an embodiment of the present application, wherein the method displays an image based on multiple color lookup tables. As shown in FIG4 , when there are multiple color lookup tables, the method may include steps S410 to S440:

[0089] Step S410: Obtain multiple color lookup tables, where the color lookup tables are mapping relationships between the original image and the target image learned through a neural network model, the brightness of the target image is lower than the brightness of the original image, and the difference between the contrast of the target image and the contrast of the original image is less than a target threshold.

[0090] Step S420: inputting the image to be displayed into the neural network model to predict the fusion coefficient to obtain a plurality of fusion coefficients, wherein the plurality of fusion coefficients correspond one-to-one to the plurality of color lookup tables.

[0091] Step S430: performing weighted summation on the multiple color lookup tables according to the multiple fusion coefficients to obtain a fused color lookup table.

[0092] Step S440: performing mapping processing on the image to be displayed according to the fusion color lookup table to obtain a display image.

[0093] In the embodiments of the present application, the number of color lookup tables is determined based on the required display image quality and the image processing speed. The greater the number of color lookup tables, the more compatible the mapping relationship between the fused color lookup table and the image to be displayed. The resulting image, mapped according to the fused color lookup table, is more capable of achieving low power consumption and high-quality display. However, excessive color lookup tables can reduce image processing speed. Therefore, the number of color lookup tables is determined by considering both the required display image quality and the image processing speed.

[0094] The image to be displayed is fed into the neural network model, which extracts features from the image and generates multiple fusion coefficients. These fusion coefficients correspond to multiple color lookup tables. For example, if there are eight color lookup tables, the image to be displayed is fed into the neural network model for feature extraction, and the fusion coefficients for each of the eight color lookup tables are generated.

[0095] After obtaining multiple fusion coefficients, each fusion coefficient can be used to weight each color lookup table, and then the multiple weighted color lookup tables are averaged to obtain a fused color lookup table. In this way, a fused color lookup table is obtained by averaging multiple color lookup tables. The fused color lookup table is the mapping relationship that best matches the image to be displayed, thus avoiding errors in a single color lookup table that result in excessive brightness or low contrast in the displayed image.

[0096] In one example, the fused color lookup table can be represented by a mapping curve. For example, FIG5 is a schematic diagram of a mapping curve provided in an embodiment of the present application. As shown in FIG5 , the horizontal axis represents the pixel grayscale value of the original image, and the vertical axis represents the pixel grayscale value of the target image. Mapping the image to be displayed according to the fused color lookup table specifically includes: according to the pixel grayscale value of the image to be displayed (the horizontal axis), finding the mapped pixel grayscale value (the vertical axis) from the mapping curve, and then obtaining the display image according to the grayscale values ​​of all pixels.

[0097] 6 , FIG6 is a flowchart of another image display method provided in an embodiment of the present application. The method may include steps S610 to S660:

[0098] Step S610: Obtain multiple color lookup tables, where the color lookup tables are mapping relationships between the original image and the target image learned through a neural network model, the brightness of the target image is lower than the brightness of the original image, and the difference between the contrast of the target image and the contrast of the original image is less than a target threshold.

[0099] Step S620: extracting a brightness channel image from the image to be displayed, and processing the brightness channel image into a low-brightness brightness channel image.

[0100] Step S630: Input the low-brightness brightness channel image into the neural network model to predict the fusion coefficient to obtain multiple fusion coefficients.

[0101] Step S640: performing weighted summation on the multiple color lookup tables according to the multiple fusion coefficients to obtain a fused color lookup table.

[0102] Step S650: performing mapping processing on the low-brightness luminance channel image according to the fused color lookup table to obtain a generated luminance channel image.

[0103] Step S660: Fusing the color channel image of the image to be displayed with the generated brightness channel image to obtain a display image.

[0104] In an embodiment of the present application, in order to generate a display image with low brightness and high contrast, the brightness channel image of the image to be displayed is processed into a low-brightness brightness channel image, and then the neural network model performs feature extraction based on the low-brightness brightness channel image to obtain multiple fusion coefficients. The fusion color lookup table obtained based on the multiple fusion coefficients satisfies the mapping relationship of low brightness and high contrast, and the fusion color lookup table is the mapping relationship that is most suitable for the image to be displayed, so that the display image finally generated is an image with low brightness and high contrast.

[0105] For example, FIG7 is a flowchart of an image display method for a multi-color lookup table provided by an embodiment of the present application. Specifically, a luminance channel image and a color channel image are extracted from the image to be displayed, and the luminance channel image is processed into a low-luminance luminance channel image; the low-luminance luminance channel image is input into a neural network model for fusion coefficient prediction, and a plurality of fusion coefficients (i.e., w1, w2, ..., w n-1 ,w n ); performing weighted summation on multiple color lookup tables according to multiple fusion coefficients to obtain a fused color lookup table, and performing mapping processing on the low-brightness brightness channel image according to the fused color lookup table to obtain a generated brightness channel image; finally, fusing the color channel image of the image to be displayed with the generated brightness channel image to obtain a display image.

[0106] The multiple color lookup tables of the embodiment of the present application can be obtained by performing unsupervised training on a neural network model, or by performing supervised training. The two methods are described below.

[0107] Method 1: Unsupervised training method to obtain multiple color lookup tables:

[0108] Among them, the neural network model includes: a first fusion coefficient determination model, which takes the generation of a low-brightness and high-contrast generated image as a training goal, the low brightness means that the brightness of the generated image is lower than the brightness of the original image, and the high contrast means that the difference between the contrast of the generated image and the contrast of the original image is less than the target threshold.

[0109] Specifically, the plurality of color lookup tables and the first fusion coefficient determination model are trained by the following steps D1 to D5:

[0110] Step D1: Acquire multiple color lookup tables to be trained, and acquire the original image.

[0111] Step D2: inputting the original image into the first fusion coefficient determination model to be trained for unsupervised training to obtain a plurality of first fusion coefficients for training, wherein the plurality of first fusion coefficients for training correspond one-to-one to the plurality of color lookup tables to be trained.

[0112] Step D3: Obtain a third generated image according to the multiple first fusion coefficients for training, the multiple color lookup tables to be trained, and the original image.

[0113] Step D4: Obtain a third loss value based on the third generated image and the original image.

[0114] Step D5: Update the model parameters of the first fusion coefficient determination model to be trained according to the third loss value to obtain the first fusion coefficient determination model, and update the multiple color lookup tables to be trained according to the third loss value to obtain multiple color lookup tables.

[0115] In the embodiment of the present application, the first fusion coefficient determination model is trained to obtain a generated image with low brightness and high contrast. In some embodiments, the contrast of the generated image is lower than the contrast of the original image, and the contrast value of the generated image below the original image does not exceed the target threshold. In other embodiments, the contrast of the generated image is higher than the contrast of the original image, and the contrast value of the generated image above the original image does not exceed the target threshold.

[0116] The multiple color lookup tables to be trained are updated with each training session, and after the training session, multiple color lookup tables are obtained. The first fusion coefficient determination module extracts features from the original image to obtain multiple first fusion coefficients for training. Specifically, the original image is converted from an RGB color model to a YUV color model, a luminance channel image and a color channel image are extracted from the converted image, and the luminance channel image is processed into a low-luminance luminance channel image. The first fusion coefficient determination module then extracts features from the low-luminance luminance channel image to obtain multiple first fusion coefficients for training.

[0117] Obtaining a third generated image based on multiple first fusion coefficients for training, the multiple color lookup tables to be trained, and the original image, specifically includes: processing the brightness channel image of the original image into a low-brightness brightness channel image, performing weighted summation on the multiple color lookup tables to be trained according to the first fusion coefficient for training to obtain a fused color lookup table for training, mapping the low-brightness brightness channel image according to the fused color lookup table for training to obtain a mapped brightness channel image, and fusing the mapped brightness channel image with the color channel image of the original image to obtain the third generated image.

[0118] In order to achieve the training goal of determining the model of the first fusion coefficient to obtain a generated image with low brightness and high contrast, the loss functions used in the unsupervised training include: power consumption loss function, structural information loss function and contrast loss function; wherein, the power consumption loss function aims to reduce the brightness of the generated image, the structural information loss function aims to minimize the structural information loss of the generated image, and the contrast loss function aims to minimize the contrast loss of the generated image.

[0119] The aforementioned power loss function, structural information loss function, and contrast loss function ensure that the overall brightness of the displayed image is reduced to any desired ratio, while ensuring that the displayed image's structural information, such as texture, and overall image contrast approximate the original image. Specifically, the specific representations of the power loss function, structural information loss function, and contrast loss function are as described above and will not be detailed here.

[0120] In an embodiment of the present application, the loss value between the third generated image and the original image is calculated based on the power loss function, structural information loss function, and contrast loss function described above to obtain a third loss value. The third loss value includes the power loss value, structural information loss value, and contrast loss value between the third generated image and the original image. Therefore, the model parameters of the first fusion coefficient determination model to be trained are updated based on the third loss value, so that the first fusion coefficient determination model can generate a generated image with low brightness and high contrast. Furthermore, the multiple color lookup tables to be trained are updated based on the third loss value, so that the color lookup tables can represent the mapping relationship between the original image and the target image.

[0121] In this way, the first fusion coefficient determination model to be trained and the multiple color lookup tables to be trained can be trained simultaneously, resulting in a paired first fusion coefficient determination model and multiple color lookup tables. For each paired first fusion coefficient determination model and multiple color lookup tables, the most suitable fusion color lookup table for each image to be displayed can be quickly and accurately obtained, thereby achieving a display image that meets low power consumption and high-quality display requirements.

[0122] Method 2: Supervised training method to obtain multiple color lookup tables:

[0123] In which, the neural network model includes: a second fusion coefficient determination model, which takes minimizing the loss between the generated image and the true value image as the training goal, the brightness of the true value image is lower than the brightness of the original image, and the difference between the contrast of the true value image and the contrast of the original image is less than the target threshold.

[0124] Specifically, the plurality of color lookup tables and the second fusion coefficient determination model are trained by the following steps E1 to E5:

[0125] Step E1: Acquire multiple color lookup tables to be trained, and acquire the original image and a true value image corresponding to the original image.

[0126] Step E2: Inputting the original image into the second fusion coefficient determination model to be trained for supervised training to obtain a plurality of second fusion coefficients for training, wherein the plurality of second fusion coefficients for training correspond one-to-one to the plurality of color lookup tables to be trained.

[0127] Step E3: Obtain a fourth generated image according to the plurality of second fusion coefficients for training, the plurality of color lookup tables to be trained, and the original image.

[0128] Step E4: Obtain a fourth loss value based on the fourth generated image and the true value image.

[0129] Step E5: Update the model parameters of the second fusion coefficient determination model to be trained according to the fourth loss value to obtain the second fusion coefficient determination model, and update the multiple color lookup tables to be trained according to the fourth loss value to obtain multiple color lookup tables.

[0130] In an embodiment of the present application, the original image and the true value image corresponding to the original image are used as an image pair to perform supervised training on the second image processing model. In some embodiments, the contrast of the true value image is lower than the contrast of the original image, and the contrast value of the true value image lower than the original image does not exceed the target threshold. In other embodiments, the contrast of the true value image is higher than the contrast of the original image, and the contrast value of the true value image higher than the original image does not exceed the target threshold. The multiple color lookup tables to be trained are updated through each training, and after the training is completed, multiple color lookup tables are obtained. The second fusion coefficient determination module extracts features from the original image to obtain multiple second fusion coefficients for training. Specifically, the original image is converted from the RGB color model to the YUV color model, and the brightness channel image and the color channel image are extracted from the converted image, and the brightness channel image is processed into a low-brightness brightness channel image; then the second fusion coefficient determination model extracts features from the low-brightness brightness channel image to obtain multiple second fusion coefficients for training.

[0131] A fourth generated image is obtained based on multiple second fusion coefficients for training, the multiple color lookup tables to be trained, and the original image, specifically including: processing the brightness channel image of the original image into a low-brightness brightness channel image, performing weighted summation on the multiple color lookup tables to be trained according to the second fusion coefficients for training to obtain a fused color lookup table for training, mapping the low-brightness brightness channel image according to the fused color lookup table for training to obtain a mapped brightness channel image, and fusing the mapped brightness channel image with the color channel image of the original image to obtain the fourth generated image.

[0132] A fourth loss value is obtained based on the fourth generated image and the true value image, including: calculating the loss value based on the pixel grayscale values ​​of the fourth generated image and the pixel grayscale values ​​of the true value image to obtain the fourth loss value, and then taking minimizing the loss between the generated image and the true value image as the training goal, updating the model parameters of the fourth fusion coefficient determination model to be trained based on the fourth loss value, so that the second fusion coefficient determination model can generate a generated image with low brightness and high contrast; updating the multiple color lookup tables to be trained based on the fourth loss value, so that the color lookup tables can represent the mapping relationship between the original image and the target image.

[0133] In this way, the second fusion coefficient determination model to be trained and the multiple color lookup tables to be trained can be trained simultaneously, resulting in a second fusion coefficient determination model and multiple color lookup tables for use in pairs. For each paired second fusion coefficient determination model and multiple color lookup tables, the fusion color lookup table that best matches each image to be displayed can be quickly and accurately obtained, thereby achieving a display image that meets low power consumption and high-quality display requirements.

[0134] Optionally, the first fusion coefficient determination model and the second fusion coefficient determination model can be a neural network model as shown in Figure 8. The entire network structure adopts a total of six convolution layers. The first layer inputs 1 channel and outputs 16 channels, with a 3×3 convolution kernel and a step size of 2. The second to fifth layers input 16 channels and output 16 channels, with a 3×3 convolution kernel and a step size of 2. The sixth layer inputs 16 channels and outputs 8 channels (assuming the number of color lookup tables is 8), with a 3×3 convolution kernel and a step size of 1. In this neural network model, the ReLU function is used as the activation function after all convolution layers. Finally, the softmax function is used to output the fusion coefficient corresponding to each color lookup table.

[0135] In an optional embodiment, the neural network model includes: N feature extraction modules and 1 fusion module, and the N feature extraction modules are respectively deployed on N computing resources; the image to be displayed is input into the neural network model for fusion coefficient prediction to obtain multiple fusion coefficients, including: dividing the image to be displayed into N sub-images of the same size; inputting the N sub-images into the N feature extraction modules respectively to obtain N image extraction features; inputting the N image extraction features into the fusion module to obtain fusion features, and obtaining multiple fusion coefficients based on the fusion features.

[0136] In an embodiment of the present application, N computing resources (for example, FPGA computing resources) are used to process the image to be displayed, so as to quickly obtain multiple fusion coefficients corresponding to the image to be displayed. Specifically, in order to meet the deployment requirements of N computing resources, the neural network model includes N feature extraction modules, wherein each feature extraction module uses six convolution layers to implement feature extraction. The first convolution layer inputs 1 channel and outputs 16 channels, a 3×3 convolution kernel, and a step size of 2. The second to fifth convolution layers all input 16 channels, output 16 channels, a 3×3 convolution kernel, and a step size of 2. The sixth convolution layer inputs 16 channels, outputs 8 channels (assuming the number of color lookup tables is 8), a 3×3 convolution kernel, and a step size of 1. The Relu function is used as the activation function after all convolution layers. After the N feature extraction modules complete the feature extraction of the N sub-images, the fusion module fuses the features extracted from the N images and outputs the fusion coefficients corresponding to each color lookup table through the Softmax function.

[0137] For example, FIG9 is a structural diagram of another neural network model provided by an embodiment of the present application. For the deployment of dual computing resources (i.e., N is equal to 2), a low-brightness luminance channel image of size 480×270 is divided into two 240×270 sub-images, and the two sub-images are respectively input into two feature extraction modules for feature extraction to obtain two image extraction features; the two image extraction features are input into the fusion module to obtain fusion features, and the fusion coefficient corresponding to each color lookup table is output through the Softmax function. The prediction of multiple fusion coefficients is completed by a neural network model that meets the dual computing resources, so as to match and integrate the best fusion color lookup table for the image to be displayed based on multiple fusion coefficients to achieve low power consumption and high contrast image display.

[0138] In an optional embodiment, after obtaining the display image, the method further includes: displaying the display image using an organic light emitting diode display device.

[0139] In the embodiment of the present application, the brightness of the displayed image is lower than the brightness of the image to be displayed, and the difference between the contrast and the contrast of the image to be displayed is less than a target threshold. Therefore, the displayed image satisfies low power consumption and high-quality image display requirements. The organic light-emitting diode display device of the present application has the characteristics of low power consumption and high contrast. Therefore, the organic light-emitting diode display device is used to display the display image, achieving adaptive contrast improvement when the overall brightness of the displayed image of the organic light-emitting diode display device is low.

[0140] Furthermore, in order to better illustrate the image display method provided by the embodiment of the present application, images in different scenes are processed according to the provided image display method, and the processed images are subjected to index analysis. Figure 10 is a schematic diagram of the analysis results of a sculpture scene image provided by the embodiment of the present application, Figure 11 is a schematic diagram of the analysis results of a plant scene image provided by the embodiment of the present application, and Figure 12 is a schematic diagram of the analysis results of an animal scene image provided by the embodiment of the present application. Among them, the low-brightness image is a low-brightness brightness channel image after the brightness channel image of the image to be displayed is multiplied by 0.5, the display image is the display image obtained by the embodiment of the present application, and the same-brightness contrast image is an image obtained by multiplying the brightness of the image to be displayed by the same brightness channel image reduction ratio as the display image. From the perspective of subjective visual effects, it can be seen that the peak brightness, contrast and structural details of the displayed image are better than those of the low-brightness image and the same-brightness contrast image. In terms of objective indicators, the mean value of the brightness channel of the displayed image decreased by between 30% and 50%, achieving the goal of reducing power consumption. At the same time, the standard deviation decreased, and the PSNR (Peak Signal to Noise Ratio) and SSIM (Structure Similarity Index Measure) were all better than those of low-brightness images and images with the same brightness, achieving the goal of low-brightness and high contrast of image information.

[0141] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application.

[0142] The present application also provides an image display device, as shown in FIG13 , which is a schematic structural diagram of an image display device provided in the present application. The device includes:

[0143] A first acquisition module 1310 is configured to acquire a color lookup table, wherein the color lookup table is a mapping relationship between an original image and a target image learned through a neural network model, wherein the brightness of the target image is lower than the brightness of the original image, and the difference between the contrast of the target image and the contrast of the original image is less than a target threshold;

[0144] The first mapping module 1320 is configured to perform mapping processing on the image to be displayed according to the color lookup table to obtain a display image.

[0145] In an optional embodiment, the neural network model includes: a first image processing model, wherein the first image processing model is trained to obtain a generated image with low brightness and high contrast, wherein the low brightness means that the brightness of the generated image is lower than the brightness of the original image, and the high contrast means that the difference between the contrast of the generated image and the contrast of the original image is less than the target threshold;

[0146] The color lookup table and the first image processing model are obtained by training a first training module, and the first training module includes:

[0147] A second acquisition module is used to acquire a color lookup table to be trained, and to acquire the original image;

[0148] A first processing module is configured to input the original image into a first image processing model to be trained for unsupervised training to obtain a first generated image;

[0149] A first loss module, configured to obtain a first loss value according to the first generated image and the original image;

[0150] A first updating module is used to update the model parameters of the first image processing model to be trained according to the first loss value to obtain the first image processing model, and to update the color lookup table to be trained according to the first loss value to obtain the color lookup table.

[0151] In an optional embodiment, the neural network model includes: a second image processing model, wherein the second image processing model is trained to minimize the loss between a generated image and a true image, the brightness of the true image is lower than the brightness of the original image, and the difference between the contrast of the true image and the contrast of the original image is less than the target threshold;

[0152] The color lookup table and the second image processing model are obtained by training a second training module, and the second training module includes:

[0153] A third acquisition module is used to obtain a color lookup table to be trained, and to obtain an original image and a true value image corresponding to the original image;

[0154] A second processing module is configured to input the original image into a second image processing model to be trained for supervised training to obtain a second generated image;

[0155] A second loss module, configured to obtain a second loss value according to the second generated image and the true value image;

[0156] The second updating module is used to update the model parameters of the second image processing model to be trained according to the second loss value to obtain the second image processing model, and to update the color lookup table to be trained according to the second loss value to obtain the color lookup table.

[0157] In an optional embodiment, when there are multiple color lookup tables, the device further includes:

[0158] A first prediction module is configured to input the image to be displayed into the neural network model to predict a fusion coefficient, thereby obtaining a plurality of fusion coefficients, wherein the plurality of fusion coefficients correspond one-to-one to the plurality of color lookup tables;

[0159] A first weighting module, configured to perform weighted summation on the plurality of color lookup tables according to the plurality of fusion coefficients to obtain a fused color lookup table;

[0160] The second mapping module is configured to perform mapping processing on the image to be displayed according to the fused color lookup table to obtain a display image.

[0161] In an optional embodiment, the neural network model includes: a first fusion coefficient determination model, wherein the first fusion coefficient determination model is trained to obtain a generated image with low brightness and high contrast, wherein the low brightness means that the brightness of the generated image is lower than the brightness of the original image, and the high contrast means that the difference between the contrast of the generated image and the contrast of the original image is less than the target threshold;

[0162] The plurality of color lookup tables and the first fusion coefficient determination model are obtained by training through a third training module, and the third training module includes:

[0163] a fourth acquisition module, configured to acquire a plurality of color lookup tables to be trained, and to acquire the original image;

[0164] a third processing module, configured to input the original image into a first fusion coefficient determination model to be trained for unsupervised training, to obtain a plurality of first fusion coefficients for training, wherein the plurality of first fusion coefficients for training correspond one-to-one to the plurality of color lookup tables to be trained;

[0165] a first generating module, configured to obtain a third generated image according to the plurality of first fusion coefficients for training, the plurality of color lookup tables to be trained, and the original image;

[0166] a third loss module, configured to obtain a third loss value according to the third generated image and the original image;

[0167] The third updating module is used to update the model parameters of the first fusion coefficient determination model to be trained according to the third loss value to obtain the first fusion coefficient determination model, and to update the multiple color lookup tables to be trained according to the third loss value to obtain multiple color lookup tables.

[0168] In an optional embodiment, the neural network model includes: a second fusion coefficient determination model, which takes minimizing the loss between the generated image and the true value image as a training goal, the brightness of the true value image is lower than the brightness of the original image, and the difference between the contrast of the true value image and the contrast of the original image is less than the target threshold;

[0169] The plurality of color lookup tables and the second fusion coefficient determination model are obtained by training through a fourth training module, and the fourth training module includes:

[0170] a fifth acquisition module, configured to acquire a plurality of color lookup tables to be trained, and to acquire the original image and a true value image corresponding to the original image;

[0171] a fourth processing module, configured to input the original image into a second fusion coefficient determination model to be trained for supervised training, to obtain a plurality of second fusion coefficients for training, wherein the plurality of second fusion coefficients for training correspond one-to-one to the plurality of color lookup tables to be trained;

[0172] a second generating module, configured to obtain a fourth generated image according to the plurality of second fusion coefficients for training, the plurality of color lookup tables to be trained, and the original image;

[0173] a fourth loss module, configured to obtain a fourth loss value based on the fourth generated image and the true image;

[0174] A fourth updating module is used to update the model parameters of the second fusion coefficient determination model to be trained according to the fourth loss value to obtain the second fusion coefficient determination model, and to update the multiple color lookup tables to be trained according to the fourth loss value to obtain multiple color lookup tables.

[0175] In an optional embodiment, the device further includes:

[0176] A construction module is used to construct a true value image generation model, wherein the true value image generation model is obtained by unsupervised training with the generation of true value images as the training goal;

[0177] The fifth processing module is used to input the original image into the true value image generation model for processing to obtain the true value image.

[0178] In an optional embodiment, the loss functions used in the unsupervised training include: power loss function, structural information loss function and contrast loss function;

[0179] The power consumption loss function aims to reduce the brightness of the generated image, the structure information loss function aims to minimize the structure information loss of the generated image, and the contrast loss function aims to minimize the contrast loss of the generated image.

[0180] In an optional embodiment, the device further includes:

[0181] A first extraction module is used to extract a brightness channel image from the image to be displayed, and process the brightness channel image into a low-brightness brightness channel image;

[0182] The second prediction module is used to input the low-brightness brightness channel image into the neural network model to predict the fusion coefficient and obtain multiple fusion coefficients.

[0183] In an optional embodiment, the second mapping module includes:

[0184] A first mapping submodule is configured to perform mapping processing on the low-brightness luminance channel image according to the fused color lookup table to obtain a generated luminance channel image;

[0185] The first fusion module is configured to fuse the color channel image of the image to be displayed with the generated brightness channel image to obtain a display image.

[0186] In an optional embodiment, the neural network model includes: N feature extraction modules and 1 fusion module, and the N feature extraction modules are respectively deployed on N computing resources; the second prediction module includes:

[0187] A segmentation module, configured to segment the image to be displayed into N sub-images of the same size;

[0188] A second extraction module is used to input the N sub-images into the N feature extraction modules respectively to obtain N image extraction features;

[0189] The fusion module is used to input the N image extraction features into the fusion module to obtain fusion features, and obtain multiple fusion coefficients based on the fusion features.

[0190] In an optional embodiment, the true value image generation model includes: an encoder and a decoder, wherein the encoder and the decoder are connected via a residual; and the fifth processing module includes:

[0191] A third extraction module is used to extract features from the original image using the encoder to obtain feature information;

[0192] A restoration module, configured to perform restoration processing on the feature information using the decoder to obtain a restored image;

[0193] A true value image module is used to obtain the true value image according to the restored image and the original image.

[0194] In an optional embodiment, the original image includes an image to be displayed by a display device.

[0195] In an optional embodiment, the device further includes:

[0196] The display module is used to display the display image using an organic light emitting diode display device.

[0197] The present application also provides an electronic device, as shown in Figure 14, which is a schematic diagram of the structure of an electronic device provided in the present application. As shown in Figure 14, the electronic device 1400 includes a memory 1410 and a processor 1420. The memory 1410 and the processor 1420 are connected via a bus. The memory 1410 stores a computer program that can be executed on the processor 1420 to implement the steps of the image display method described in the present application.

[0198] An embodiment of the present application further provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the image display method described in the embodiment of the present application are implemented.

[0199] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0200] The embodiments of the present application are described with reference to the flowcharts and / or block diagrams of the methods, devices, and equipment according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate a device for implementing the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0201] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device that implements the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0202] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce computer-implemented processing, so that the instructions executed on the computer or other programmable terminal device provide steps for implementing the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0203] Although preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they become aware of the basic inventive concepts. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present invention.

[0204] Finally, it should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or terminal device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or terminal device that includes the element.

[0205] The above is a detailed introduction to the image display method, device, electronic device and storage medium provided by the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea; at the same time, for general technical personnel in this field, based on the ideas of the present application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. An image display method, characterized in that: The method comprises: Obtaining a color lookup table, where the color lookup table is a mapping relationship between an original image and a target image learned by a neural network model, wherein the brightness of the target image is lower than the brightness of the original image, and the difference between the contrast of the target image and the contrast of the original image is less than a target threshold; According to the color lookup table, mapping processing is performed on the image to be displayed to obtain a display image.

2. The method according to claim 1, characterized in that The neural network model includes: a first image processing model, wherein the first image processing model is trained to obtain a generated image with low brightness and high contrast, wherein the low brightness means that the brightness of the generated image is lower than the brightness of the original image, and the high contrast means that the difference between the contrast of the generated image and the contrast of the original image is less than the target threshold; The color lookup table and the first image processing model are trained by the following steps: Obtaining a color lookup table to be trained, and obtaining the original image; Inputting the original image into a first image processing model to be trained for unsupervised training to obtain a first generated image; Obtaining a first loss value according to the first generated image and the original image; The model parameters of the first image processing model to be trained are updated according to the first loss value to obtain the first image processing model, and the color lookup table to be trained is updated according to the first loss value to obtain the color lookup table.

3. The method according to claim 1, characterized in that The neural network model includes: a second image processing model, wherein the second image processing model is trained to minimize the loss between a generated image and a true image, the brightness of the true image is lower than the brightness of the original image, and the difference between the contrast of the true image and the contrast of the original image is less than the target threshold; The color lookup table and the second image processing model are trained by the following steps: Obtaining a color lookup table to be trained, and obtaining an original image and a true value image corresponding to the original image; Inputting the original image into a second image processing model to be trained for supervised training to obtain a second generated image; Obtaining a second loss value according to the second generated image and the true value image; The model parameters of the second image processing model to be trained are updated according to the second loss value to obtain the second image processing model, and the color lookup table to be trained is updated according to the second loss value to obtain the color lookup table.

4. The method according to claim 1, wherein In the case where there are multiple color lookup tables, the method further includes: Inputting the image to be displayed into the neural network model to predict a fusion coefficient to obtain a plurality of fusion coefficients, wherein the plurality of fusion coefficients correspond one to one to the plurality of color lookup tables; Performing weighted summation on the multiple color lookup tables according to the multiple fusion coefficients to obtain a fused color lookup table; Mapping the image to be displayed according to the color lookup table to obtain a display image includes: According to the fusion color lookup table, mapping processing is performed on the image to be displayed to obtain a display image.

5. The method according to claim 4, characterized in that The neural network model includes: a first fusion coefficient determination model, wherein the first fusion coefficient determination model is trained to obtain a generated image with low brightness and high contrast, wherein the low brightness means that the brightness of the generated image is lower than the brightness of the original image, and the high contrast means that the difference between the contrast of the generated image and the contrast of the original image is less than the target threshold; The plurality of color lookup tables and the first fusion coefficient determination model are trained by the following steps: Obtaining a plurality of color lookup tables to be trained, and obtaining the original image; Inputting the original image into a first fusion coefficient determination model to be trained for unsupervised training to obtain a plurality of first fusion coefficients for training, wherein the plurality of first fusion coefficients for training correspond one-to-one to the plurality of color lookup tables to be trained; Obtaining a third generated image according to the plurality of first fusion coefficients for training, the plurality of color lookup tables to be trained, and the original image; Obtaining a third loss value according to the third generated image and the original image; The model parameters of the first fusion coefficient determination model to be trained are updated according to the third loss value to obtain the first fusion coefficient determination model, and the multiple color lookup tables to be trained are updated according to the third loss value to obtain the multiple color lookup tables.

6. The method according to claim 4, characterized in that The neural network model includes: a second fusion coefficient determination model, with minimizing the loss between a generated image and a true image as a training goal, the brightness of the true image is lower than the brightness of the original image, and the difference between the contrast of the true image and the contrast of the original image is less than the target threshold; The plurality of color lookup tables and the second fusion coefficient determination model are trained by the following steps: Acquire a plurality of color lookup tables to be trained, and acquire the original image and a true value image corresponding to the original image; Inputting the original image into a second fusion coefficient determination model to be trained for supervised training to obtain a plurality of second fusion coefficients for training, wherein the plurality of second fusion coefficients for training correspond one-to-one to the plurality of color lookup tables to be trained; Obtaining a fourth generated image according to the plurality of second fusion coefficients for training, the plurality of color lookup tables to be trained, and the original image; Obtaining a fourth loss value according to the fourth generated image and the true value image; The model parameters of the second fusion coefficient determination model to be trained are updated according to the fourth loss value to obtain the second fusion coefficient determination model, and the multiple color lookup tables to be trained are updated according to the fourth loss value to obtain the multiple color lookup tables.

7. The method according to claim 3 or 6, characterized in that Obtaining a true value image corresponding to the original image, including: Constructing a true value image generation model, wherein the true value image generation model is obtained by performing unsupervised training with generating true value images as a training goal; The original image is input into the true value image generation model for processing to obtain the true value image.

8. The method according to any one of claims 2, 5 and 7, characterized in that: The loss functions used in the unsupervised training include: power loss function, structural information loss function and contrast loss function; The power consumption loss function aims to reduce the brightness of the generated image, the structure information loss function aims to minimize the structure information loss of the generated image, and the contrast loss function aims to minimize the contrast loss of the generated image.

9. The method according to claim 4, characterized in that The method further comprises: Extracting a brightness channel image from the image to be displayed, and processing the brightness channel image into a low-brightness brightness channel image; The image to be displayed is input into the neural network model to predict the fusion coefficient to obtain multiple fusion coefficients, including: The low-brightness brightness channel image is input into the neural network model to predict the fusion coefficient to obtain multiple fusion coefficients.

10. The method according to claim 9, characterized in that Mapping the image to be displayed according to the fused color lookup table to obtain a display image, including: Performing mapping processing on the low-brightness luminance channel image according to the fused color lookup table to obtain a generated luminance channel image; The color channel image of the image to be displayed is fused with the generated brightness channel image to obtain a display image.

11. The method according to claim 4, characterized in that The neural network model includes: N feature extraction modules and 1 fusion module, and the N feature extraction modules are respectively deployed on N computing resources; The image to be displayed is input into the neural network model to predict the fusion coefficient to obtain multiple fusion coefficients, including: Dividing the image to be displayed into N sub-images of the same size; Inputting N sub-images into the N feature extraction modules respectively to obtain N image extraction features; The N image extraction features are input into the fusion module to obtain fusion features, and a plurality of fusion coefficients are obtained based on the fusion features.

12. The method according to claim 7, characterized in that The true value image generation model includes: an encoder and a decoder, wherein the encoder and the decoder are connected via a residual; Inputting the original image into the true value image generation model for processing to obtain the true value image, including: Using the encoder to extract features from the original image to obtain feature information; Using the decoder to restore the feature information to obtain a restored image; The true value image is obtained according to the restored image and the original image.

13. The method according to any one of claims 1 to 12, characterized in that: The original image refers to the image to be displayed by the display device.

14. The method according to any one of claims 1 to 12, characterized in that: The method further includes: displaying the display image using an organic light emitting diode display device.

15. An image display device, characterized in that: The device comprises: A first acquisition module is configured to acquire a color lookup table, wherein the color lookup table is a mapping relationship between an original image and a target image learned through a neural network model, wherein the brightness of the target image is lower than the brightness of the original image, and the difference between the contrast of the target image and the contrast of the original image is less than a target threshold; The first mapping module is configured to perform mapping processing on the image to be displayed according to the color lookup table to obtain a display image.

16. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the image display method according to any one of claims 1 to 14 are implemented.

17. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the image display method according to any one of claims 1 to 14 are implemented.

Images