Display method, display system and application of electronic device

Abstract

The present invention discloses a display method, a display system and an application of an electronic device, which specifically include: inputting an initial image; obtaining a first image correction parameter data set, the first image correction parameter data set including image correction parameter data for correcting unevenness within a display screen and / or inconsistency between the display screen and a reference screen; correcting the initial image using a first convolution kernel neural network and the first image correction parameter data set to obtain a corrected image of the initial image; outputting the corrected image data to a display screen for display, so that the display screen displays an image after correcting unevenness within the display screen and / or inconsistency between the display screen and the reference screen, so as to solve the technical problems of low precision and complex operation of existing display screen correction based on hardware parameters.

Description

Technical Field

[0001] The present invention belongs to the technical field of display screen correction, and in particular relates to a display method, a display system and applications of an electronic device. Background Art

[0002] Generally speaking, the display parameter correction of the display screen is mainly based on the compensation algorithm and parameter setting based on the hardware parameter correction. Taking gamma correction as an example, due to the inconsistency of the electro-optical characteristics of the red, green and blue colors of the LCD screen, the color difference of each grayscale is large, and the color of each grayscale needs to be corrected. In particular, the grayscale error of the dark field is very obvious, and the color error of each grayscale cannot be eliminated by white balance adjustment. Only when the color of each grayscale is consistent can the color temperature be adjusted to the required color temperature through the white balance adjustment of the bright and dark fields. By measuring the curve between the image brightness and the input voltage, such as by giving a fixed register value (input voltage or other hardware parameters) to a certain binding point of the module to be tested, a set of xyLv data is measured using a colorimeter, and this set of data is used to match the data of the existing adjusted screen, and the register value of the closest screen is obtained as the initial value of the module to be tested, and the register value (input voltage or other hardware parameters) of the module to be tested is obtained by the approximation algorithm as its corresponding hardware adjustment parameter. After correcting the GAMMA curve, the following goals can be achieved: the color of the dark field grayscale is significantly improved, the color error of each grayscale is significantly reduced, the dark field color details are clear, the image brightness and color are consistent, the brightness is good, and the contrast is obvious.

[0003] In addition, since each screen has its own specificity, such as the display characteristics of each display unit in each screen are not exactly the same, and the brightness and color are not uniform among the blocks of the segmented display screen (LCD w / 2D BLU or wall screen), measurement and verification are required. Figure 1 FIG. 1 is a flow chart of display screen correction in the prior art. Figure 1 As shown, the traditional verification method is divided into two steps: first, measure the non-uniformity of each block, and then compensate for the non-uniformity. The compensation method is determined by the characteristics of the specific non-uniformity. The usual practice is to record the non-uniformity in the storage unit (flash) of the display driver chip, and set the compensation function and its related parameters in the driver chip. The compensation parameters are determined according to the non-uniformity and these parameters are called to generate a compensated image output to the display. In some systems, the non-uniformity storage unit, compensation algorithm and related parameters are completed by the main chip of the system (TV or mobile phone).

[0004] In addition, the existing compensation methods for image compensation are usually based on the physical model of mura generation, and the measured image is fitted to estimate the compensation value of each pixel at each grayscale, which does not involve decoding and re-encoding the input image based on the convolution kernel neural network. Moreover, the parameter adjustment process involves many manual operations based on experience.

[0005] In the display correction process of the above display screen, the corresponding compensation algorithm and parameter setting require the participation of the display driver chip manufacturer and the data provided by the display screen manufacturer, which in turn needs to be completed by the machine of the measurement equipment manufacturer. At the same time, the current compensation is mostly based on the brightness value of a single channel or RGB three channels. The compensation algorithm for chromaticity deviation requires more calculations and measurements, and the effect inspection after compensation is mostly based on visual evaluation. Therefore, the display correction based on hardware parameters involves multiple suppliers and the process is lengthy, and in many cases it is difficult to meet user needs. Summary of the invention

[0006] In view of the above defects or improvement needs of the prior art, the present invention provides a display method, a display system and an application of an electronic device, aiming to solve the technical problems of low precision and complex operation of the existing display screen correction based on hardware parameters.

[0007] To achieve the above object, according to one aspect of the present invention, a display method of an electronic device is provided. The electronic device includes a display screen, and the display method includes:

[0008] Input the initial image;

[0009] Acquire a first image correction parameter data set, the first image correction parameter data set comprising image correction parameter data for correcting non-uniformity within the display screen and / or inconsistency between the display screen and a reference screen;

[0010] Correcting the initial image using the first convolution kernel neural network and the first image correction parameter data set to obtain a corrected image of the initial image;

[0011] The corrected image data is output to a display screen for display, so that the display screen displays an image after correcting the unevenness within the display screen and / or the inconsistency between the display screen and a reference screen.

[0012] As a further improvement of the present invention, the first convolution kernel neural network is a fully convolutional neural network.

[0013] As a further improvement of the present invention, the display screen has intra-screen non-uniformity at least in one gray scale, or there is inconsistency between the display screen and the reference screen;

[0014] As a further improvement of the present invention, the display screen is a wall splicing screen, and the wall splicing screen includes a plurality of display units which are independently photoelectrically controlled.

[0015] Or the display screen is a liquid crystal screen.

[0016] As a further improvement of the present invention, correcting the initial image includes:

[0017] Image decoding, including extracting feature parameters from the input initial image;

[0018] Image encoding includes compensating the initial image through a first image correction parameter data set, and outputting the compensated image data to a display screen.

[0019] As a further improvement of the present invention, obtaining the first image correction parameter data set includes:

[0020] Using the first image data set to perform model training to obtain a first image correction parameter data set, where the first image data set includes the first category of pictures and / or the second category of pictures;

[0021] The second type of images are used to characterize the presence of intra-screen non-uniformity in at least one grayscale of the display screen, or the inconsistency between the display screen and the reference screen, or the inconsistency between multiple display units in the display screen;

[0022] The first category of pictures are pictures that are not the second category of pictures.

[0023] As a further improvement of the present invention, obtaining the first image correction parameter data set includes:

[0024] Using the second image data set to perform model training to obtain a second image correction parameter data set, where the second image data set includes at least the first category of pictures;

[0025] Using the second type of pictures to correct the second image correction parameter data set to obtain the first image correction parameter data set;

[0026] The second type of images are used to characterize the presence of intra-screen non-uniformity in at least one grayscale of the display screen, or the inconsistency between the display screen and the reference screen, or the inconsistency between multiple display units in the display screen;

[0027] The first category of pictures are pictures that are not the second category of pictures.

[0028] As a further improvement of the present invention, the second type of pictures are display pictures that are obtained by inputting the first type of pictures or other test pictures to a display screen and by imaging.

[0029] As a further improvement of the present invention, a computing server is used to perform computing to obtain the first image correction parameter data set.

[0030] As a further improvement of the present invention, the display screen is a wall splicing screen, which is spliced ​​by a plurality of display units independently controlled by photoelectricity. The corresponding screen adjustment process includes:

[0031] Acquire a third image data set, wherein the size of each image in the third image data set is adapted to a single display unit;

[0032] Obtaining a third image correction parameter data set according to the third image data set;

[0033] The third image data set includes pictures of the first type suitable for a single display unit;

[0034] The third image correction parameter data set is used as the second image correction parameter data set applicable to the entire wall splicing screen.

[0035] As a further improvement of the present invention, all display units of the reference screen have consistent display brightness.

[0036] As a further improvement of the present invention, the display characteristics of the reference screen are preset values ​​within the variation range of the display characteristics of the display screen, and the preset values ​​are average values ​​obtained by collecting electro-optical performance data of multiple sample screens.

[0037] As a further improvement of the present invention, the first image correction parameter data set is used to correct the unevenness within the display screen and / or the inconsistency between the display screen and the reference screen, the inconsistency includes the inconsistency of one or more display parameters among gamma, color and brightness, and the unevenness includes the unevenness of one or more display parameters among gamma, color and brightness.

[0038] To achieve the above object, according to another aspect of the present invention, a display system is provided, the system comprising:

[0039] A processing module including a processing chip, used for receiving an initial image and a first image correction parameter data set, wherein the first image correction parameter data set includes image correction parameter data for correcting unevenness within a display screen and / or inconsistency between the display screen and a reference screen; and further used for correcting the initial image using a first convolution kernel neural network and the first image correction parameter data set to obtain a corrected image of the initial image, and outputting the corrected image data to the display screen;

[0040] The display screen is used to receive the corrected image data and display the image after correcting the unevenness in the display screen and / or the inconsistency between the display screen and a reference screen.

[0041] As a further improvement of the present invention, the type of processing chip includes one or more of ISP, GPU and NPU.

[0042] To achieve the above-mentioned purpose, according to another aspect of the present invention, there is provided an electronic device, comprising at least one processing unit and at least one storage unit, wherein the storage unit stores a computer program, and when the program is executed by the processing unit, the processing unit executes the steps of the above-mentioned method.

[0043] To achieve the above objective, according to another aspect of the present invention, a computer-readable medium is provided, which stores a computer program executable by an electronic device, and when the computer program runs on the electronic device, the electronic device executes the steps of the above method.

[0044] In general, the above technical solution conceived by the present invention has the following beneficial effects compared with the prior art:

[0045] The present invention provides a display method, a display system and an application of an electronic device, which obtains a first image correction parameter data set suitable for a display screen and uses the image correction parameter data set to correct an input image, thereby avoiding the traditional method of correcting by adjusting a driving voltage through hardware.

[0046] The present invention provides a display method, a display system and an application of an electronic device. By acquiring a first image correction parameter data set suitable for a display screen, the display screen measurement and correction parameter generation can be completed in the same functional module. There is no need to store the non-uniformity pattern and verification parameters separately or have them completed by different manufacturers. At the same time, through correction, a hardware screen with large non-uniformity can still obtain a good image effect, thereby relaxing the binning requirements of LEDs, improving production capacity and reducing costs.

[0047] The present invention provides a display method, display system and application of an electronic device, which uses a convolutional kernel neural network to calibrate the image quality after correction, thereby reducing the burden of manual visual verification and improving accuracy and reliability. Therefore, the professional requirements for test site engineers can be greatly reduced, and many tasks of image quality assessment can be completed by quantitative calculation.

[0048] The display method, display system and application of an electronic device of the present invention have a fundamental change in the design of a display screen driver chip, reducing the hardware design requirements of the driver chip for characteristic adjustment parameters and compensation algorithms of a specific display screen, but greatly increasing the computational requirements for the generation of a corrected image. The corresponding image calculations are performed by an ISP or NPU or GPU computing unit in a processing module main chip, and algorithm development and evaluation can be completed through simulation, which can greatly shorten the development cycle. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 A schematic diagram of a process for display screen correction in the prior art;

[0050] Figure 2 A flow chart of a display method of the technical solution of the present invention;

[0051] Figure 3 It is a schematic diagram of one specific embodiment of the technical solution of the present invention. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0053] In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other. The present invention is further described in detail below in conjunction with specific embodiments.

[0054] Figure 2 FIG. 1 is a flow chart of a display method of the technical solution of the present invention. Figure 2 As shown, in one embodiment, a display method is provided, which uses an application processor (AP) to perform image processing. Preferably, the application processor is provided with a processing chip of an ISP or a GPU or an NPU, which inputs the processed image data to a display screen for image display, specifically comprising:

[0055] The application processor acquires an initial image through a camera connected thereto or a network connected thereto, and may also acquire a corresponding first image correction parameter data set through the above connection, wherein the first image correction parameter data set includes image correction parameter data for correcting non-uniformity within a display screen and / or inconsistency between the display screen and a reference screen;

[0056] On the basis of obtaining the above input data information, the application processor can use the convolution kernel neural network to perform image data calculation on the above data, generate image data after correcting the initial image, and output it to the display screen, so that the display screen displays the image after correcting the unevenness in the display screen and / or the inconsistency between the display screen and the reference screen, thereby obtaining the correction of the corresponding input image, and transmitting it to the display screen to correct the unevenness of the screen or the variation between screens, so that the displayed image is closer to the original input image. As a preferred method, the convolution kernel neural network is a full convolution neural network. Of course, the application processor is only a specific implementation form of the processing module, and other electronic devices including processing chips that can achieve the same function can be used to perform the above image processing, and the specific selection can be made according to the needs. Since the above method directly uses the processing device to perform the corresponding image processing, the correction of the display screen can be completed without the corresponding modulation of the hardware parameters of the display screen or the display device.

[0057] Preferably, the display screen may be a liquid crystal screen with intra-screen non-uniformity at least in one grayscale, or the display screen may be inconsistent with a reference screen; or the display screen may be a wall splicing screen, which includes a plurality of display units that are independently photoelectrically controlled; or the display screen may be a liquid crystal screen with backlight partition control (2D backlight). In short, any display device that needs to correct intra-screen non-uniformity or inter-screen inconsistency can be corrected using the above method, so that its corresponding display image is closer to the original input image.

[0058] Preferably, the first image data set includes:

[0059] The first category of pictures may be one or more of natural landscapes, daily life and texts, or may be test pictures used to characterize image quality, such as uniform pictures of monochrome or mixed colors, but are not suitable for reflecting the existence of inhomogeneity within the display screen at least in one grayscale, or reflecting the inconsistency between the display screen and a standard screen, or reflecting the inconsistency between multiple display units in the display screen;

[0060] The second type of pictures is used to reflect the existence of in-screen non-uniformity in at least one grayscale of the display screen, or the inconsistency between the display screen and a standard screen, or the inconsistency between multiple display units in the display screen.

[0061] Preferably, the application processor uses a convolutional kernel neural network to perform image data calculation to obtain a first image correction parameter data set suitable for the display screen and performs image correction, and the application processor includes at least one encoder and at least one decoder, specifically including:

[0062] Using a decoder to perform image decoding, including extracting feature parameters from an input initial image;

[0063] An encoder is used to perform image encoding, including compensating each input image through a first image correction parameter data set, and outputting the compensated input image data to a display device. The above process can correct the unevenness within the display screen and / or the inconsistency between the display screen and the reference screen.

[0064] Obtaining a first image correction parameter data set includes: using the first image data set to perform model training to obtain a first image correction parameter data set, the first image data set includes a first type of image and / or a second type of image; using a learning algorithm such as a neural network to learn from the first image data set to obtain feature parameters for image correction to obtain a first image correction parameter data set, the first image data set here includes image data for characterizing unevenness within a display screen and / or inconsistency between a display screen and a reference screen. As an example, the learning process can be to adjust the initial first image correction parameter data set by the difference between the image data used for training and its corresponding original image data, and continuously adjust it through the corresponding training set and verify it through the corresponding verification set, so as to obtain a trained first image correction parameter data set; the second type of image can be obtained by taking a test image displayed on the imaging display screen. Of course, the first image data set here can also include some general image data, and specific image selection can be performed according to needs.

[0065] As another implementation manner, obtaining the first image correction parameter data set includes:

[0066] The model is trained using the second image data set to obtain a second image correction parameter data set, the second image data set includes at least the first type of pictures; as an example, the learning process can be to adjust the initial second image correction parameter data set by the difference between the picture data used for training and the corresponding original picture data, and continuously adjust the corresponding training set and verify the corresponding verification set to obtain the trained second image correction parameter data set, the second picture data set is a general picture set; the general picture set here refers to not including image data used to characterize the unevenness in the display screen or the inconsistency between the display screen and the reference screen; the second image correction parameter data set here can be used for image correction of the same type of display screen or the same batch of display screens, but because the second picture data set does not carry image data that characterizes the unevenness in the display screen or the inconsistency between the display screen and the reference screen, it cannot correct the unevenness in the display screen or the inconsistency between the display screen and the reference screen. Wherein, all display units of the reference screen here have consistent display brightness, or the display characteristics of the reference screen are preset values ​​within the display characteristics variation range of the display screen, and the preset value is the average value obtained by collecting electro-optical performance data of multiple sample screens.

[0067] Based on the processing of the second image correction parameter data set, the process of acquiring the first image correction parameter data set includes:

[0068] By inputting a test picture to the display screen and acquiring a corresponding display picture by taking an image, the display picture is input as a training picture to the processing module to obtain a first image correction parameter data set suitable for the display screen. Preferably, the display picture can be added to the second picture data set to generate a corresponding first picture data set. Of course, the corresponding first picture data set can also be generated by combining the display picture with other general pictures according to the requirements.

[0069] Preferably, the above-mentioned operation of obtaining the first image correction parameter data set and / or the second image correction parameter data set can be performed on a computing server, which can reduce the local computing workload and improve the local image processing efficiency.

[0070] Preferably, the display image here is used to characterize the presence of intra-screen non-uniformity in the display screen at least in one grayscale; or to characterize the display deviation caused by the inconsistency between the display screen and a reference screen; or to characterize the display deviation caused by the inconsistency between multiple display units in the display screen.

[0071] As a specific example, for a 2D backlight LCD screen, by generating an algorithm of a fully convolutional neural network, a corrected image suitable for a LCD screen can be generated for any input image after a given 2D backlight lighting pattern. After this corrected image is combined with the backlight lighting pattern, the display effect is very close to the original input image to the viewer of the display screen. In this processing process, each block of the 2D backlight is assumed to be the same and each block is uniform. In fact, the blocks used for display are not the same and may be non-uniform, so it is necessary to verify and correct this non-uniformity. The verification and correction scheme proposed on this basis is as follows:

[0072] Figure 3 FIG. 1 is a schematic diagram of a specific embodiment of the technical solution of the present invention. Figure 3 As shown, the original input image I in The backlight neural network LDNN generates the corrected image I CP The image on the screen is combined with the backlight shape presented on the screen to obtain the output image I seen by the observer. out In order to make I out Reproduce the original image I in , taken with the camera outAfter the image is input to the server, the difference loss is compared and the LDNN is used to adjust the parameters for optimization. When the loss is less than the pre-given threshold, the image optimization is completed. The generated model parameters define the LDNN backlight model M0; in the above process, it is first assumed that each block is uniform and uniform, and the algorithm model M0 generated thereby is used as the initial model, and on this basis, the correction model M1 is further established. The process is as follows: first input a number of test image sets A on the display screen to be verified, and then use a camera to shoot the display screen image to obtain a verification image set A′ showing the unevenness of the display screen; the test image set A and the verification image set A′ are respectively used as I in and I out ; The model is trained again to obtain a correction model M1 for the unevenness of the display screen. The model parameters corresponding to M1 are derived from the training server and stored in the storage unit of the electronic device on which the display screen is installed. When the display screen is used, the main chip of the electronic device retrieves the model parameters of M1 from the storage unit, runs the model on the input picture to be displayed, first decodes the image and then re-encodes it with the model parameters to generate a corrected picture, which is sent to the display screen to obtain a corrected display picture. This picture has corrected the unevenness of the display screen, and the picture presented to the observer on the display screen reproduces the original input picture.

[0073] In order to test the effectiveness of the above method, a computer model method can be used to generate chromaticity non-uniformity from the original uniform 2D backlight block; calculate the impact of this non-uniformity on several different images presented on the display. When the correction model M0 is used, without considering the non-uniformity of each backlight block, obvious color and brightness deviations can be seen, especially for a uniform white screen, the block display with chromaticity non-uniformity has obviously shifted to magenta, and after using the two correction methods M1 and M2 described above, it can be seen that the display of the block with chromaticity non-uniformity has been significantly improved.

[0074] For the special case where the display screen is a wall splicing screen, the wall splicing screen here includes multiple display units that are independently photoelectrically controlled, and the corresponding screen adjustment process includes:

[0075] Acquire a third image data set, wherein the size of each image in the third image data set is adapted to a single display unit;

[0076] Obtaining a third image correction parameter data set according to the third image data set;

[0077] The third image data set includes pictures of the first type suitable for a single display unit;

[0078] The third image correction parameter data set is used as the second image correction parameter data set applicable to the entire wall splicing screen. For example, the display brightness of the reference display unit is consistent, or the display characteristic of the reference display unit is a preset value within the display characteristic variation range of the display unit, and the preset value is an average value obtained by collecting electro-optical performance data of multiple display units.

[0079] As a specific example, similar to the above, first generate M0 from the original image library as the initial model, but because the spliced ​​screen does not need to consider the backlight-related light transfer diffusion function and backlight lighting pattern, the M0 training is greatly simplified. In theory, it is only necessary to reproduce the original image of the image library without transformation. Then repeat the method of the first embodiment, input a number of test image sets A on the large screen to be spliced ​​to be verified, and use a camera to shoot the display screen image to obtain a verification image set A′ showing the unevenness of the display screen; the test image set A and the verification image set A′ are respectively used as I in and I out , and train the model again to obtain a correction model M1 for the unevenness of the display screen. The model parameters corresponding to M1 are derived from the training server and stored in the storage unit of the electronic device that controls the display screen. When the display screen is used, the main chip of the electronic device retrieves the model parameters of M1 from the storage unit, runs the model on the input picture to be displayed, first decodes the image and then re-encodes it with the model parameters to generate a corrected picture, which is sent to the display screen to obtain a corrected display picture. This picture has corrected the unevenness of the display screen, and the picture presented to the observer on the display screen reproduces the original input picture.

[0080] In the process of generating M0, you can choose screens with different resolutions. For the full convolutional network, the parameters of the final model are independent of the selected resolution. As an optimization solution, you can choose a smaller resolution, such as the resolution of a certain splicing module to generate M0. In the subsequent M1 generation process, the resolution of the final spliced ​​large screen is selected for verification.

[0081] Preferably, the processing module is used to correct the non-uniformity within the display screen and / or the inconsistency between the display screen and the reference screen, the inconsistency between the display screen and the reference screen includes the inconsistency of one or more display parameters including gamma, color and brightness, and the non-uniformity within the display screen also includes the non-uniformity of one or more display parameters including gamma, color and brightness.

[0082] In another embodiment, a display system is provided, the system comprising:

[0083] A processing module is provided with an ISP or a GPU or an NPU, and is used to obtain an initial image and a first image data set, wherein the first image data set includes image data for characterizing the non-uniformity within the display screen and / or the inconsistency between the display screen and a reference screen; and is also used to perform image data calculation using a convolution kernel neural network, and generate image data after correcting the initial image and output it to the display screen;

[0084] The display screen is used to receive the corrected image data and display the image after correcting the non-uniformity in the display screen and / or the inconsistency between the display screen and a reference screen.

[0085] The implementation principle and technical effects of this system are similar to those of the above-mentioned display method and will not be described here in detail.

[0086] Corresponding to the above display method, an electronic device is also provided, comprising at least one processing unit and at least one storage unit, wherein the storage unit stores a computer program, and when the program is executed by the processing unit, the processing unit executes the steps of the above method.

[0087] Corresponding to the above display method, a computer-readable medium is also provided, which stores a computer program executable by an electronic device. When the computer program runs on the electronic device, the electronic device executes the steps of the above method.

[0088] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A display method for an electronic device, characterized in that: The electronic device includes a display screen, and the display method includes: Input the initial image; Acquire a first image correction parameter data set, wherein the first image correction parameter data set includes image correction parameter data for correcting non-uniformity within the display screen and / or inconsistency between the display screen and a reference screen; Correcting the initial image using the first convolution kernel neural network and the first image correction parameter data set to obtain a corrected image of the initial image; Outputting the corrected image data to the display screen for display, so that the display screen displays an image after correcting the unevenness within the display screen and / or the inconsistency between the display screen and a reference screen; Wherein, obtaining the first image correction parameter data set includes: Performing model training using a second image data set to obtain the second image correction parameter data set, where the second image data set includes at least the first category of pictures; Using the second type of pictures to correct the second image correction parameter data set to obtain the first image correction parameter data set; The second type of pictures are used to characterize that the display screen has inhomogeneity within the display screen at least on one gray scale, or inconsistency between the display screen and a reference screen, or inconsistency between multiple display units in the display screen; The first category of pictures is not the second category of pictures.

2. A display method for an electronic device according to claim 1, wherein: The first convolutional kernel neural network is a fully convolutional neural network.

3. The display method of an electronic device according to claim 1, wherein: The display screen has intra-screen non-uniformity at least in one gray scale, or there is inconsistency between the display screen and a reference screen.

4. The display method of an electronic device according to claim 1, wherein: The display screen is a wall splicing screen, and the wall splicing screen includes a plurality of display units which are independently photoelectrically controlled. Or the display screen is a liquid crystal screen.

5. The display method of an electronic device according to claim 1, wherein: The correcting of the initial image comprises: Image decoding, including extracting feature parameters from the input initial image; Image encoding includes compensating the initial image using the first image correction parameter data set and outputting the compensated image data to a display screen.

6. The display method of an electronic device according to claim 1, wherein: Obtaining the first image correction parameter data set includes: Performing model training using a first image data set to obtain the first image correction parameter data set, where the first image data set includes first-category pictures and / or second-category pictures; The second type of pictures are used to characterize that the display screen has inhomogeneity within the display screen at least on one gray scale, or inconsistency between the display screen and a reference screen, or inconsistency between multiple display units in the display screen; The first category of pictures is not the second category of pictures.

7. The display method of an electronic device according to claim 1, wherein: The second type of pictures are display pictures that are obtained by inputting the first type of pictures to the display screen and displaying them through imaging.

8. The display method of an electronic device according to claim 1, wherein: The computing server is used to perform computing to obtain the first image correction parameter data set.

9. The display method of an electronic device according to claim 1, wherein: The display screen is a wall splicing screen, which is composed of a plurality of display units independently controlled by photoelectricity. The corresponding screen adjustment process includes: The size of each picture in the second image data set is adapted to a single display unit; Using the second image data set to perform model training to obtain the second image correction parameter data set applicable to a single display unit; The second image correction parameter data set applicable to the single display unit is used as the second image correction parameter data set applicable to the entire wall spliced ​​screen.

10. A display method for an electronic device according to any one of claims 1, wherein: All display units of the reference screen have consistent display brightness.

11. The display method of an electronic device according to claim 1, wherein: The display characteristic of the reference screen is a preset value within the variation range of the display characteristic of the display screen, and the preset value is an average value obtained by collecting electro-optical performance data of multiple sample screens.

12. The display method of an electronic device according to claim 1, wherein: The first image correction parameter data set is used to correct the non-uniformity within the display screen and / or the inconsistency between the display screen and a reference screen, the inconsistency includes the inconsistency of one or more display parameters among gamma, color and brightness, and the non-uniformity includes the non-uniformity of one or more display parameters among gamma, color and brightness.

13. A display system, characterized in that: The system comprises: A processing module including a processing chip, configured to receive an initial image and a first image correction parameter data set, wherein the first image correction parameter data set includes image correction parameter data for correcting unevenness within a display screen and / or inconsistency between the display screen and a reference screen; and further configured to correct the initial image using a first convolution kernel neural network and the first image correction parameter data set to obtain a corrected image of the initial image, and output the corrected image data to the display screen; A display screen, used for receiving the corrected image data and displaying an image after correcting the unevenness within the display screen and / or the inconsistency between the display screen and a reference screen; Wherein, obtaining the first image correction parameter data set includes: Performing model training using a second image data set to obtain the second image correction parameter data set, where the second image data set includes at least the first category of pictures; Using the second type of pictures to correct the second image correction parameter data set to obtain the first image correction parameter data set; The second type of pictures are used to characterize that the display screen has inhomogeneity within the display screen at least on one gray scale, or inconsistency between the display screen and a reference screen, or inconsistency between multiple display units in the display screen; The first category of pictures is not the second category of pictures.

14. A display system according to claim 13, wherein: The types of the processing chip include one or more of ISP, GPU and NPU.

15. An electronic device, characterized in that: The method comprises at least one processing unit and at least one storage unit, wherein the storage unit stores a computer program, and when the program is executed by the processing unit, the processing unit executes the steps of the method according to any one of claims 1 to 12.

16. A computer readable medium, characterized in that It stores a computer program executable by an electronic device. When the computer program runs on the electronic device, the electronic device executes the steps of the method according to any one of claims 1 to 12.

Images