Image data conversion method and device, display device and storage medium

Abstract

The invention discloses an image data conversion method and device, a display device and a storage medium. The method comprises the following steps: establishing a neural network model for image data conversion, and training the model by using a preset input and output image data set to obtain a trained model; the input of the neural network model is RGB three-primary-color component data of an input image and input backlight voltage, and the output of the neural network model is RGBW four-primary-color component data of an output image and output backlight voltage; optimizing the trained neural network model based on the display mode to obtain an optimized neural network model corresponding to at least one display mode; and receiving an input image, selecting the matched optimized neural network model according to the current display mode, and inputting the three-primary-color component data of the input image and the input backlight voltage into the optimized neural network model for image data conversion. The display effect and the backlight power consumption of the output image are optimized according to the characteristics of different display modes.

Description

Technical Field

[0001] The present invention relates to but is not limited to the field of display technology, and in particular to a method, device, display device and storage medium for converting image data. Background Art

[0002] With the development of display technology, display technologies with diversified structures that are conducive to high resolution have been developed one after another. Figure 1 As shown in the figure, the RGBW technology adds a white W sub-pixel to the original three primary colors of red, green, blue, and RGB, making it a four-color pixel design. When displaying images of the same brightness, the power consumption of the RGBW pixel architecture is lower than that of the RGB pixel architecture; and under the same power consumption, the brightness of the RGBW pixel architecture is greatly improved compared to the RGB pixel architecture.

[0003] Therefore, the RGBW pixel architecture has greatly improved the brightness of the picture compared to the traditional RGB pixel architecture, but at the same time, the addition of W sub-pixels will also cause problems such as the picture color becoming lighter. For projector products with RGBW pixel architecture, the input image is RGB data. After the projector motherboard converts RGB data into RGBW data, the addition of W sub-pixels often causes problems such as loss of picture saturation, which affects the user's viewing experience. Summary of the invention

[0004] In a first aspect, an embodiment of the present disclosure provides a method for converting image data, comprising:

[0005] Establishing a neural network model for image data conversion, and training the neural network model using a preset input image data set and an output image data set to obtain a trained neural network model; wherein the input of the neural network model is the red, green, blue (RGB) three-primary color component data of the input image and an input backlight voltage, and the output of the neural network model is the red, green, blue, white (RGBW) four-primary color component data of the output image and an output backlight voltage;

[0006] Optimizing the trained neural network model based on the display mode to obtain an optimized neural network model corresponding to at least one display mode;

[0007] An input image is received, a matching optimized neural network model is selected according to a current display mode, and RGB three primary color component data of the input image and an input backlight voltage are input into the optimized neural network model to convert image data.

[0008] In a second aspect, an embodiment of the present disclosure provides an apparatus for converting image data, comprising: a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the steps of the above-mentioned method for converting image data are implemented.

[0009] In a third aspect, an embodiment of the present disclosure provides a computer-readable storage medium storing a computer program, which implements the steps of the above-mentioned method for converting image data when executed by a processor.

[0010] In a fourth aspect, an embodiment of the present disclosure provides a display device, comprising: the above-mentioned device for converting image data.

[0011] The method, device, display device and storage medium for converting image data provided by the embodiments of the present disclosure can select a matching neural network optimization model according to the current display mode after receiving the input image. Since the neural network model compensates for the influence of the W sub-pixel on the display according to the characteristics of the display mode during the optimization training process, the display quality of the output image can be significantly improved after the image data is converted using the optimized neural network model. In addition, the backlight power consumption accounts for a large proportion of the overall power consumption of the display device. By using the method for converting image data of the embodiments of the present disclosure, the backlight voltage is used as the input of the neural network model. The model can make full use of the favorable influence of the W primary color component on the display brightness to realize the dynamic adjustment of the backlight voltage, thereby reducing the backlight power consumption. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The accompanying drawings are used to provide an understanding of the technical solution of the present disclosure and constitute a part of the specification. Together with the embodiments of the present disclosure, they are used to explain the technical solution of the present disclosure and do not constitute a limitation on the technical solution of the present disclosure.

[0013] Figure 1 A schematic diagram of converting RGB image data into RGBW image data in the related art;

[0014] Figure 2 A flowchart of a method for converting image data provided by an embodiment of the present disclosure;

[0015] Figure 3 A schematic diagram of the structure of a convolutional neural network model provided in an embodiment of the present disclosure;

[0016] Figure 4 A schematic diagram of the structure of a device for converting image data provided by an embodiment of the present disclosure. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solutions and advantages of the present disclosure clearer, the embodiments of the present disclosure will be described in detail in conjunction with the accompanying drawings below. Note that the embodiments can be implemented in multiple different forms. A person of ordinary skill in the art can easily understand the fact that the methods and contents can be transformed into various forms without departing from the purpose and scope of the present disclosure. Therefore, the present disclosure should not be interpreted as being limited to the contents described in the following embodiments. In the absence of conflict, the embodiments in the present disclosure and the features in the embodiments can be arbitrarily combined with each other.

[0018] In the present specification, ordinal numbers such as “first”, “second” and “third” are provided to avoid confusion among constituent elements, and are not intended to limit the number.

[0019] like Figure 2 As shown, the embodiment of the present disclosure provides a method for converting image data, including:

[0020] Step S10, establishing a neural network model for image data conversion, and training the neural network model using a preset input image data set and an output image data set to obtain a trained neural network model; wherein the input of the neural network model is the red, green, blue (RGB) three-primary color component data of the input image and the input backlight voltage, and the output of the neural network model is the red, green, blue, white (RGBW) four-primary color component data of the output image and the output backlight voltage;

[0021] Step S20, optimizing the trained neural network model based on the display mode to obtain an optimized neural network model corresponding to at least one display mode;

[0022] Step S30, receiving an input image, selecting a matching optimized neural network model according to the current display mode, and inputting the RGB three primary color component data of the input image and the input backlight voltage into the optimized neural network model to convert the image data.

[0023] The method for converting image data provided in the above embodiment establishes a neural network model for image data conversion, trains the neural network model using a preset input image data set and an output image data set to obtain a trained neural network model; wherein the input of the neural network model is the red, green, blue (RGB) three primary color component data of the input image and the input backlight voltage, and the output of the neural network model is the red, green, blue, white (RGBW) four primary color component data of the output image and the output backlight voltage; optimizes the trained neural network model based on the display mode to obtain an optimized neural network model corresponding to at least one display mode; receives an input image, selects a matching optimized neural network model according to the current display mode, and inputs the RGB three primary color component data of the input image and the input backlight voltage into the optimized neural network model to convert the image data. The method provided in the above embodiment can optimize the display effect of the output image and the backlight power consumption according to the characteristics of different display modes.

[0024] In an exemplary embodiment, the red, green, blue, RGB three-primary color component data of the input image includes: the RGB three-primary color component data of all pixels of the input image; the input backlight voltage of the input image includes: the input backlight voltage corresponding to a frame of the input image (the input backlight voltage value corresponding to all pixels on the input image is the same); the red, green, blue, white, RGBW four-primary color component data of the output image includes: the RGBW four-primary color component data of all pixels of the output image; the output backlight voltage includes: the output backlight voltage corresponding to a frame of the output image (the output backlight voltage value corresponding to all pixels on the output image is the same). The input backlight voltage and the output backlight voltage are the same voltage value for the entire picture.

[0025] In an exemplary embodiment, the neural network model includes: a convolutional neural network (CNN) model. In other embodiments, the neural network model may also be other types of neural network models, such as a deep neural network (DNN) model.

[0026] In an exemplary embodiment, Figure 3 As shown, the convolutional neural network model includes the following structural layers connected in series in sequence: an input layer, a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer, a third convolutional layer, a third pooling layer, a fully connected layer and an output layer.

[0027] Among them, the input layer is used for data input. The convolution layer (the first convolution layer, the second convolution layer and the third convolution layer) is used to use the convolution kernel for feature extraction and feature mapping. The pooling layer (the first pooling layer, the second pooling layer and the third pooling layer) is used for downsampling, sparse processing of feature maps, and reducing the amount of data calculation. The fully connected layer is used for refitting to reduce the loss of feature information. The output layer is used to output the results.

[0028] In an exemplary embodiment, the first convolution layer, the second convolution layer and the third convolution layer include: a convolution kernel layer, a batch normalization layer and an excitation layer.

[0029] In an exemplary embodiment, the number and size of the convolution kernels of the first convolution layer, the second convolution layer, and the third convolution layer can be set according to the display resolution requirement. For example, for a resolution of 1920*1080, the first convolution layer can use 32 5*5 convolution kernels; the second convolution layer can use 64 3*3 convolution kernels; and the third convolution layer can use 128 3*3 convolution kernels.

[0030] In an exemplary embodiment, the excitation layer uses a ReLU activation function.

[0031] In an exemplary embodiment, the first pooling layer and the second pooling layer are maximum pooling layers; and the third pooling layer is a global average pooling layer.

[0032] In an exemplary embodiment, a plurality of residual blocks are further included between the first convolutional layer and the first pooling layer, and between the second convolutional layer and the second pooling layer, each residual block including two convolutional layers and a jump connection. The jump connection is to connect the input directly to the output bypassing the two convolutional layers so that information can be directly propagated across layers to avoid gradient vanishing and gradient exploding problems. The use of the residual block structure can help the convolutional neural network model better capture the features of the input image. And batch normalization is used to accelerate training and improve accuracy. In addition, a global average pooling layer is used to reduce the amount of calculation, while improving the stability and reliability of the network.

[0033] The convolutional neural network model algorithm can obtain accurate output values ​​through the training of key parameters, making the display performance of the picture better. Larger convolution kernels can be used in the input layer and the first convolution layer to process high-resolution input images. Using batch normalization can accelerate training and improve accuracy. The use of the first pooling layer and the second pooling layer can further reduce the dimension of the convolution layer output and improve the computational efficiency and speed of the convolutional neural network model. The use of the global average pooling layer can reduce the amount of calculation and improve the stability and reliability of the convolutional neural network model.

[0034] In an exemplary embodiment, the preset input image data set and output image data set include: a set of pure color pictures in the DCI-P3 color gamut. DCI-P3 is a color gamut used in digital cinemas. It is a color gamut standard dominated by human visual experience. By traversing the pure color pictures in the DCI-P3 color gamut to train the convolutional neural network model, since the RGB data of each pixel on the input image is the same, the effect of dimensionality reduction is achieved, which greatly reduces the amount of calculation.

[0035] In an exemplary embodiment, the input image includes: a static image or a dynamic image; the dynamic image includes: a video or an animation.

[0036] In an exemplary embodiment, the number of nodes in each layer of the convolutional neural network model can be determined according to the requirements of the input image. For example, when the input image (such as a video) needs to be processed quickly, the number of nodes in each layer of the convolutional neural network model can be reduced. When the picture quality of the input image (such as a picture) needs to be improved, the number of nodes in each layer of the convolutional neural network model can be increased.

[0037] In an exemplary embodiment, the input backlight voltage of any frame of input image in the preset input image data set is the same as the output backlight voltage of the corresponding output image. When the convolutional neural network model is trained using the preset input image data set and the output image data set, the backlight voltage is set to be the same as the input value and the output value, which can reduce the complexity of the convolutional neural network model and save calculation time.

[0038] In an exemplary embodiment, the optimizing the trained neural network model based on the display mode includes:

[0039] According to different display modes, a policy gradient-based reinforcement learning method is used to optimize the trained neural network model; wherein the policy gradient-based reinforcement learning method includes determining a reward function according to the display mode.

[0040] In an exemplary embodiment, the optimizing the trained neural network model by using a policy gradient-based reinforcement learning method further includes:

[0041] Define a state space, an action space and a policy network; wherein the state of the state space is the input data of the trained neural network model; the action of the action space is the output data of the trained neural network model; and the policy network is the trained neural network model;

[0042] The policy network is trained using a RF (REINFORCE) algorithm based on Monte Carlo sampling, and a gradient-based optimization method is used during the training process to accelerate the training of the policy network, wherein the gradient-based optimization method is, for example, an Adam algorithm.

[0043] In an exemplary embodiment, the display mode may be set by a user.

[0044] In an exemplary embodiment, the display mode includes at least one of the following: a default mode, an eye protection mode, a cinema mode, and a custom mode. The display mode is a mode setting related to the display effect of the picture, and different display modes differ in display indicators (for example, the saturation, brightness, contrast, etc. of the picture). Among them, the default mode is the default setting of the system. The eye protection mode mainly reduces the output of blue light. The cinema mode is a mode that simulates the viewing scene setting of a cinema. The custom mode is a mode in which the user can set the display indicators independently.

[0045] In an exemplary embodiment, the custom mode may provide at least one of the following display indicators for the user to set: image saturation, image brightness, and image contrast.

[0046] In an exemplary embodiment, the reward function may include a penalty term of at least one of the following control indicators: saturation of the output image, brightness of the output image, color gamut of the output image, and output backlight voltage of the output image.

[0047] In an exemplary embodiment, the reward and punishment rules of the reward function include: when the actual value of the control index of the output image is greater than the target value of the control index, the parameters of the trained neural network model are updated for punishment; when the actual value of the control index of the output image is less than the target value of the control index, the parameters of the trained neural network model are updated for reward.

[0048] In an exemplary embodiment, when the display mode is a default mode, determining the reward function according to the display mode includes performing at least one of the following processes:

[0049] A saturation penalty term is set in the reward function, the control index of the saturation penalty term is the saturation of the output image, and the reward and punishment rule of the saturation penalty term is: if the saturation of the output image is greater than a first saturation threshold, the parameters of the trained neural network model are updated for punishment; if the saturation of the output image is less than the first saturation threshold, the parameters of the trained neural network model are updated for reward;

[0050] A brightness penalty term is set in the reward function, the control index of the brightness penalty term is the brightness of the output image, and the reward and punishment rule of the brightness penalty term is: if the brightness of the output image is greater than a first brightness threshold, the parameters of the trained neural network model are updated for punishment; if the brightness of the output image is less than the first brightness threshold, the parameters of the trained neural network model are updated for reward;

[0051] A color gamut penalty term is set in the reward function, the control index of the color gamut penalty term is the color gamut of the output image, and the reward and punishment rule of the color gamut penalty term is: if the proportion of pixel points in the color gamut of the output image that fall outside the target color gamut range is greater than a first color gamut threshold, the parameters of the trained neural network model are updated for punishment; if the proportion of pixel points in the color gamut of the output image that fall outside the target color gamut range is less than the first color gamut threshold, the parameters of the trained neural network model are updated for reward;

[0052] A backlight power consumption penalty item is set in the reward function, and the control indicator of the backlight power consumption penalty item is the output backlight voltage of the output image. The reward and punishment rules of the backlight power consumption penalty item are as follows: if the output backlight voltage of the output image is greater than a first voltage threshold, the parameters of the trained neural network model are updated for penalty; if the output backlight voltage of the output image is less than the first voltage threshold, the parameters of the trained neural network model are updated for reward.

[0053] In an exemplary embodiment, when the display mode is a cinema mode, determining the reward function according to the display mode includes performing at least one of the following processes:

[0054] A saturation penalty term is set in the reward function, the control index of the saturation penalty term is the saturation of the output image, and the reward and punishment rule of the saturation penalty term is: if the saturation of the output image is greater than a second saturation threshold, the parameters of the trained neural network model are updated for punishment; if the saturation of the output image is less than the second saturation threshold, the parameters of the trained neural network model are updated for reward;

[0055] A brightness penalty term is set in the reward function, the control index of the brightness penalty term is the brightness of the output image, and the reward and punishment rule of the brightness penalty term is: if the brightness of the output image is greater than the second brightness threshold, the parameters of the trained neural network model are updated for punishment; if the brightness of the output image is less than the second brightness threshold, the parameters of the trained neural network model are updated for reward;

[0056] A color gamut penalty term is set in the reward function, the control index of the color gamut penalty term is the color gamut of the output image, and the reward and punishment rule of the color gamut penalty term is: if the proportion of pixel points in the color gamut of the output image that fall outside the target color gamut range is greater than a second color gamut threshold, the parameters of the trained neural network model are updated for punishment; if the proportion of pixel points in the color gamut of the output image that fall outside the target color gamut range is less than the second color gamut threshold, the parameters of the trained neural network model are updated for reward;

[0057] A backlight power consumption penalty item is set in the reward function, and the control indicator of the backlight power consumption penalty item is the output backlight voltage of the output image. The reward and punishment rules of the backlight power consumption penalty item are as follows: if the output backlight voltage of the output image is greater than the second voltage threshold, the parameters of the trained neural network model are updated for penalty; if the output backlight voltage of the output image is less than the second voltage threshold, the parameters of the trained neural network model are updated for reward.

[0058] In an exemplary embodiment, when the display mode is an eye protection mode, the reward function determined according to the display mode includes: setting a blue light penalty term in the reward function, the control index of the blue light penalty term is the intensity of the blue light component in the output image, and the reward and punishment rule of the blue light penalty term is: if the intensity of the blue light component in the output image is greater than the intensity threshold, then the parameters of the trained neural network model are updated for punishment; if the intensity of the blue light component in the output image is less than the intensity threshold, then the parameters of the trained neural network model are updated for reward. Introducing a blue light penalty term in the eye protection mode can reduce the output of blue light and protect the viewer's eyesight.

[0059] In an exemplary embodiment, the method further includes: detecting the intensity of a blue light component of the output image by a sensor.

[0060] Introducing saturation, brightness, and color gamut penalty terms into the reward function can optimize the proportion of the white primary color component W in the output image, thus avoiding or reducing the adverse effects of the white primary color component on the display effect (for example, fading the color W).

[0061] Introducing the backlight power consumption penalty term in the reward function can make full use of the beneficial effect of the white W primary color component on the display brightness, further reduce the backlight voltage, and reduce the backlight power consumption.

[0062] In an exemplary embodiment, the first saturation threshold is less than the second saturation threshold; the first brightness threshold is greater than the second brightness threshold, and the first color gamut threshold is greater than the second color gamut threshold. Compared with the default mode, the target saturation of the cinema mode is set to be greater than the target saturation of the default mode, which can improve the picture contrast and clarity of the cinema mode. Setting the target brightness of the cinema mode to be less than the target brightness of the default mode can match the viewing environment of the cinema mode (dim light) and reduce power consumption. Setting the target color gamut width requirement of the cinema mode to be greater than the target color gamut width requirement of the default mode can enrich the picture color of the cinema mode.

[0063] On the RGBW screen, the R, G, and B sub-pixels are responsible for displaying colors, while the W sub-pixel is used to increase brightness and white performance. Assume that in the cinema mode, the reward function sets the target value of saturation (the second saturation threshold) to 50 and the target value of brightness (the second brightness threshold) to 110nit. Without changing the hardware, because the voltage and grayscale of the pixel are related, the average saturation (saturation of the image) and average brightness (brightness of the image) of all pixels in the entire picture can be obtained by projecting RGBW into the HSV (hue (H), saturation (S), brightness (V)) color space. In cinema mode, if the saturation of the output image is greater than 50, the parameters of the trained neural network model are updated for punishment, and if the saturation of the output image is less than 50, the parameters of the trained neural network model are updated for reward. If the brightness of the output image is greater than 110nit, the parameters of the trained neural network model are updated for punishment, and if the brightness of the output image is less than 110nit, the parameters of the trained neural network model are updated for reward.

[0064] like Figure 4 As shown, an embodiment of the present disclosure provides a device for converting image data, including: a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the steps of the above-mentioned method for converting image data are implemented.

[0065] An embodiment of the present disclosure provides a computer-readable storage medium storing a computer program, which implements the steps of the above-mentioned method for converting image data when executed by a processor.

[0066] The embodiment of the present disclosure also provides a display device, including the above-mentioned device for converting image data.

[0067] In an exemplary embodiment, the display device includes: a projector.

[0068] For projector products with RGBW pixel architecture, the input image is RGB data. After the projector converts the RGB data into RGBW data, the related technology will cause problems such as loss of saturation of the projected image due to the addition of W sub-pixels, affecting the user's viewing experience.

[0069] By using the method for converting image data of the embodiment of the present disclosure, after receiving the input image, the projector can select a matching neural network model optimization model according to the current display mode. Since the neural network model compensates for the influence of the W sub-pixel on the display according to the characteristics of the display mode during the optimization training process, the display quality of the projected image can be significantly improved after the image data is converted using the optimized neural network model. In addition, the backlight power consumption accounts for a large proportion of the overall power consumption of the projector. By using the method for converting image data of the embodiment of the present disclosure, the backlight voltage is used as the input of the neural network model. The model can make full use of the favorable influence of the W primary color component on the display brightness to realize the dynamic adjustment of the backlight voltage, thereby reducing the backlight power consumption.

[0070] The display device may also include a liquid crystal display device, an OLED (Organic Light-Emitting Diode) display device. The display device may also be any product or component with a display function, such as a mobile phone, a tablet computer, a television, a monitor, a notebook computer, a digital photo frame, a navigator, etc. Other essential components of the display device should be understood by those skilled in the art, and will not be described in detail here, nor should they be used as limitations on the present invention.

[0071] It will be appreciated by those skilled in the art that all or some of the steps, systems, and functional modules / units in the methods disclosed above may be implemented as software, firmware, hardware, and appropriate combinations thereof. In hardware implementations, the division between the functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed by several physical components in cooperation. Some or all components may be implemented as software executed by a processor, such as a digital signal processor or a microprocessor, or implemented as hardware, or implemented as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or temporary medium). As known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, it is well known to those of ordinary skill in the art that communication media typically contain computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0072] It should be noted that the above embodiments or implementations are merely illustrative and not restrictive. Therefore, the present disclosure is not limited to the contents specifically shown and described herein. Various modifications, substitutions or omissions may be made to the forms and details of implementation without departing from the scope of the present disclosure.

Claims

1. A method for converting image data, comprising: Establishing a neural network model for image data conversion, and training the neural network model using a preset input image data set and an output image data set to obtain a trained neural network model; wherein the input of the neural network model is the red, green, blue (RGB) three-primary color component data of the input image and an input backlight voltage, and the output of the neural network model is the red, green, blue, white (RGBW) four-primary color component data of the output image and an output backlight voltage; Optimizing the trained neural network model based on the display mode to obtain an optimized neural network model corresponding to at least one display mode; An input image is received, a matching optimized neural network model is selected according to a current display mode, and RGB three primary color component data of the input image and an input backlight voltage are input into the optimized neural network model to convert image data.

2. The method according to claim 1, characterized in that: The optimizing the trained neural network model based on the display mode includes: According to different display modes, a policy gradient-based reinforcement learning method is used to optimize the trained neural network model; wherein the policy gradient-based reinforcement learning method includes determining a reward function according to the display mode.

3. The method according to claim 2, characterized in that: The method of optimizing the trained neural network model by using a policy gradient-based reinforcement learning method also includes: Define a state space, an action space and a policy network; wherein the state of the state space is the input data of the trained neural network model; the action of the action space is the output data of the trained neural network model; and the policy network is the trained neural network model; The policy network is trained using a RF algorithm based on Monte Carlo sampling, and a gradient-based optimization method is used during the training process to accelerate the training of the policy network.

4. The method according to claim 2, characterized in that: The reward function includes a penalty term of at least one of the following control indicators: saturation of the output image, brightness of the output image, color gamut of the output image, and output backlight voltage of the output image.

5. The method according to claim 4, characterized in that: The reward and punishment rules of the reward function include: when the actual value of the control index of the output image is greater than the target value of the control index, the parameters of the trained neural network model are updated for punishment; when the actual value of the control index of the output image is less than the target value of the control index, the parameters of the trained neural network model are updated for reward.

6. The method according to claim 1, characterized in that: The display mode includes at least one of the following: a default mode, an eye protection mode, a theater mode, and a custom mode.

7. The method according to claim 6, characterized in that: The custom mode provides at least one of the following display indicators for the user to set: image saturation, image brightness, and image contrast.

8. The method according to claim 2, characterized in that: When the display mode is the default mode, determining the reward function according to the display mode includes performing at least one of the following processes: A saturation penalty term is set in the reward function, the control index of the saturation penalty term is the saturation of the output image, and the reward and punishment rule of the saturation penalty term is: if the saturation of the output image is greater than a first saturation threshold, the parameters of the trained neural network model are updated for punishment; if the saturation of the output image is less than the first saturation threshold, the parameters of the trained neural network model are updated for reward; A brightness penalty term is set in the reward function, the control index of the brightness penalty term is the brightness of the output image, and the reward and punishment rule of the brightness penalty term is: if the brightness of the output image is greater than a first brightness threshold, the parameters of the trained neural network model are updated for punishment; if the brightness of the output image is less than the first brightness threshold, the parameters of the trained neural network model are updated for reward; A color gamut penalty term is set in the reward function, the control index of the color gamut penalty term is the color gamut of the output image, and the reward and punishment rule of the color gamut penalty term is: if the proportion of pixel points in the color gamut of the output image that fall outside the target color gamut range is greater than a first color gamut threshold, the parameters of the trained neural network model are updated for punishment; if the proportion of pixel points in the color gamut of the output image that fall outside the target color gamut range is less than the first color gamut threshold, the parameters of the trained neural network model are updated for reward; A backlight power consumption penalty item is set in the reward function, and the control indicator of the backlight power consumption penalty item is the output backlight voltage of the output image. The reward and punishment rules of the backlight power consumption penalty item are as follows: if the output backlight voltage of the output image is greater than a first voltage threshold, the parameters of the trained neural network model are updated for penalty; if the output backlight voltage of the output image is less than the first voltage threshold, the parameters of the trained neural network model are updated for reward.

9. The method according to claim 2, characterized in that: When the display mode is the cinema mode, determining the reward function according to the display mode includes performing at least one of the following processes: A saturation penalty term is set in the reward function, the control index of the saturation penalty term is the saturation of the output image, and the reward and punishment rule of the saturation penalty term is: if the saturation of the output image is greater than a second saturation threshold, the parameters of the trained neural network model are updated for punishment; if the saturation of the output image is less than the second saturation threshold, the parameters of the trained neural network model are updated for reward; A brightness penalty term is set in the reward function, the control index of the brightness penalty term is the brightness of the output image, and the reward and punishment rule of the brightness penalty term is: if the brightness of the output image is greater than the second brightness threshold, the parameters of the trained neural network model are updated for punishment; if the brightness of the output image is less than the second brightness threshold, the parameters of the trained neural network model are updated for reward; A color gamut penalty term is set in the reward function, the control index of the color gamut penalty term is the color gamut of the output image, and the reward and punishment rule of the color gamut penalty term is: if the proportion of pixel points in the color gamut of the output image that fall outside the target color gamut range is greater than a second color gamut threshold, the parameters of the trained neural network model are updated for punishment; if the proportion of pixel points in the color gamut of the output image that fall outside the target color gamut range is less than the second color gamut threshold, the parameters of the trained neural network model are updated for reward; A backlight power consumption penalty item is set in the reward function, and the control indicator of the backlight power consumption penalty item is the output backlight voltage of the output image. The reward and punishment rules of the backlight power consumption penalty item are as follows: if the output backlight voltage of the output image is greater than the second voltage threshold, the parameters of the trained neural network model are updated for penalty; if the output backlight voltage of the output image is less than the second voltage threshold, the parameters of the trained neural network model are updated for reward.

10. The method according to claim 2, characterized in that: When the display mode is the eye protection mode, the reward function is determined according to the display mode, including: setting a blue light penalty item in the reward function, the control indicator of the blue light penalty item is the intensity of the blue light component in the output image, and the reward and punishment rules of the blue light penalty item are: if the intensity of the blue light component in the output image is greater than the intensity threshold, then the parameters of the trained neural network model are updated for punishment; if the intensity of the blue light component in the output image is less than the intensity threshold, then the parameters of the trained neural network model are updated for reward.

11. The method according to claim 1, characterized in that: The input image includes: a static image or a dynamic image; the dynamic image includes: a video or an animation.

12. The method according to claim 1, characterized in that: The neural network model includes: a convolutional neural network model.

13. A device for converting image data, comprising: A memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the steps of the method for converting image data according to any one of claims 1 to 12 are implemented.

14. A computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the computer program implements the steps of the method for converting image data according to any one of claims 1 to 12.

15. A display device, comprising: The device for converting image data as claimed in claim 13.

Images