A screen dynamic solid color picture measurement method based on pure visual perception

Abstract

The application discloses a screen dynamic pure color picture measuring method based on pure visual perception, which comprises the following steps: obtaining a standard sample image for testing; collecting standard sample image data through a visual sensor and performing self-adaptive learning to obtain an algorithm model for the visual sensor of the category; performing refresh frequency sensing of a display screen through the algorithm model; and automatically adjusting sampling parameters of the visual sensor according to the refresh frequency of the display screen to realize synchronization of screen refresh and visual sensor sampling. The application can automatically adjust the sampling rate and sampling duration of the visual sensor through an algorithm and software, and even for a display screen device with unknown refresh frequency, the application can search for the frequency to realize frequency synchronization and then accurately perform visual sampling.

Description

Technical Field

[0001] This invention relates to the field of image measurement, and more specifically, to a method for measuring dynamic solid color images on a screen based on pure visual perception. Background Technology

[0002] The measurement of optical properties of displays includes time stability, brightness and color uniformity, color gamut, chromaticity constancy, channel independence, and color temperature. Traditional display measurements require multiple devices such as luminance meters, colorimeters, and vision sensors, which typically need to be deployed in specialized optical laboratories. Building automated testing production lines for displays often involves multiple processes, resulting in complexity, long measurement times, and high instrument costs, making large-scale replication and application difficult. With the development and widespread adoption of camera technology, the accuracy and resolution of brightness and color measurement using vision sensors have significantly improved. If only vision sensors are used for brightness and color measurement of displays, the traditional measurement work can be concentrated into a single process step, significantly improving measurement efficiency and reducing costs.

[0003] Unlike photographing natural scenery, relying solely on visual sensors to measure displays can lead to frequency synchronization issues, resulting in flickering noise in the sampled image. Frequency synchronization refers to the fact that the screen image is dynamically refreshed, rather than a continuous, unchanging image reflecting light from a natural scene. Visual sensors also have a frame rate, representing the number of times the image is captured per second. Assuming the screen refresh rate is α and the visual sensor sampling rate is β, when α ≠ β or the two are out of sync, black streaks of varying size and speed will appear in the captured image—the so-called "flickering phenomenon." This is especially problematic in VR devices, where screen refresh rates can reach as high as 90Hz to create a more natural visual experience. Furthermore, the refresh rate of these devices is difficult to control through software, placing even higher demands on the synchronization of the visual sensor. Therefore, relying solely on visual sensors is limited by this frequency synchronization issue, making it impossible to accurately measure the brightness and colorimetry of a display's optical properties. Dedicated luminance and colorimeters are also necessary. Summary of the Invention

[0004] To address at least one of the aforementioned technical problems, this invention proposes a method for measuring dynamic solid color images on a screen based on pure visual perception.

[0005] The first aspect of this invention provides a method for measuring dynamic solid color images on a screen based on pure visual perception, comprising the following steps:

[0006] Obtain standard sample images for testing;

[0007] Standard sample image data is collected by a visual sensor, and adaptive learning is performed to obtain an algorithm model for this type of visual sensor.

[0008] The refresh rate of the display screen is perceived through an algorithmic model;

[0009] The sampling parameters of the vision sensor are automatically adjusted according to the refresh rate of the display screen to achieve synchronization between screen refresh and vision sensor sampling.

[0010] In a preferred embodiment of the present invention, the standard sample image is divided into 5 standard colors, namely red, green, blue, black and white.

[0011] In a preferred embodiment of the present invention, standard sample image data is acquired through a visual sensor, and adaptive learning is performed to obtain an algorithm model for this type of visual sensor; specifically including:

[0012] Set the minimum and maximum refresh rates for the display screen;

[0013] Read the sample image and display screen frequency range, set the screen to different refresh rate points in a certain step and display a standard image of 5 colors;

[0014] Image data is acquired via a visual sensor under each display condition;

[0015] Brightness change waveform signals were acquired using a photoelectric sensor;

[0016] The visual sensor extracts and calculates features from the images it acquires, and adjusts the visual sensor's sampling parameters.

[0017] The learning process stops when the frequency output by the algorithm matches the waveform of the photoelectric sensor.

[0018] In a preferred embodiment of the present invention, the algorithm model is denoted as Network, which is a deep learning model. The algorithm model includes two parts: Q-Network and Action-Network, specifically including:

[0019] Input the model image parameters St to generate a set of predicted actions. ;

[0020] Network in the current environment Using the action as input, the system obtains the maximum probability image pattern prediction for the next time step through a nonlinear mapping of a neural network. ;

[0021] L(w) is the L2 norm of the difference between the system's predicted reward and the actual reward, representing the error in the system's prediction. The training objective of the Q-Network is to make L(w) equal to 0, thereby obtaining data that eliminates out-of-step stripe noise from the sampled images.

[0022] In a preferred embodiment of the present invention, the network model of Q-Network samples a multilayer perceptron and performs a nonlinear mapping on the input current image, the current sampling rate, and the sampling duration to obtain the desired image pattern.

[0023] In a preferred embodiment of the present invention, the Action-Network employs an LSTM network to perform convolution and accumulation calculations on images from the past N-1 time steps.

[0024] In a preferred embodiment of the present invention, the input model image parameter St includes sampled images from the visual sensor over the past N-1 time moments, with the sampling frequency and sampling duration for each frame at each time moment.

[0025] In a preferred embodiment of the present invention, the predicted action set includes the sampling frequency and the sampling duration per frame that should be set for the visual sensor at the next time step.

[0026] The technical solution of the present invention has the following advantages compared with the prior art:

[0027] (1) This application can adaptively adjust the sampling rate and sampling duration of the visual sensor through algorithms and software. Even for display screen devices with unknown refresh rates, it can automatically search for frequencies to achieve frequency synchronization and thus accurately perform visual sampling.

[0028] (2) This application supports adaptive learning for different screen types, different visual sensor types and different frequency ranges, and learns automatically as the scene changes, with high robustness; the measurement of screen attributes takes short time and is highly efficient. Attached Figure Description

[0029] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, some of the drawings in the following description are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0030] Figure 1 This is a schematic diagram of texture noise caused by sampling misalignment in an embodiment of the present invention;

[0031] Figure 2 This is the overall process for measuring dynamic solid color images on a screen using pure visual perception in this embodiment of the invention;

[0032] Figure 3 This is a schematic diagram of the adaptive learning system for visual sensor sampling parameters in an embodiment of the present invention;

[0033] Figure 4 This is a schematic diagram of an adaptive learning neural network for sampling parameters in an embodiment of the present invention;

[0034] Figure 5 This is a schematic diagram of a purely visual screen attribute measurement system in an embodiment of the present invention. Detailed Implementation

[0035] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0036] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0037] Example 1

[0038] See Figure 1-5 As shown, this invention proposes a method for measuring dynamic solid color images on a screen based on pure visual perception, comprising the following steps:

[0039] Obtain standard sample images for testing;

[0040] Standard sample image data is collected by a visual sensor, and adaptive learning is performed to obtain an algorithm model for this type of visual sensor.

[0041] The refresh rate of the display screen is perceived through an algorithmic model;

[0042] The sampling parameters of the vision sensor are automatically adjusted according to the refresh rate of the display screen to achieve synchronization between screen refresh and vision sensor sampling.

[0043] In other words, the overall process for measuring dynamic solid color images on a screen based on pure visual perception is as follows: Figure 2As shown, the process is divided into three stages: Stage 1 involves preparing a display screen with a configurable refresh rate and standard sample images for testing; Stage 2 involves adaptive learning of the visual sensor sampling parameters to obtain an algorithm model for this type of visual sensor. This algorithm model can perceive the screen's refresh rate using only the visual sensor image and the current sampling parameters. Based on this, it automatically adjusts the visual sensor sampling parameters to achieve synchronous sampling with the screen refresh action; Stage 3 involves using the visual sensor for attribute measurement and calculation of other screens based on the algorithm model obtained in the previous step. In this stage, only the visual sensor and computer hardware are required, without the need for dedicated equipment such as luminance meters and colorimeters.

[0044] Furthermore, the system includes a set of standard sample images, a standard display screen, a photoelectric signal sensor, a vision sensor, screen refresh rate range setting software, and a learning algorithm. The standard sample images are divided into five standard colors: red, green, blue, black, and white. The screen refresh rate setting software is used to set the minimum and maximum refresh rates of the screen. The learning algorithm software reads the sample images and the screen frequency range, and then sets the screen to different refresh rate points and displays the five standard colors in a certain step manner. Under each display condition, the algorithm software simultaneously reads the image data collected by the vision sensor and the brightness change waveform signal collected by the photoelectric sensor. Based on the extraction and calculation of image features, the algorithm software adjusts the sampling parameters of the vision sensor. When the frequency result output by the algorithm is consistent with the waveform of the photoelectric sensor, the learning stops.

[0045] Secondly, a deep learning-based algorithm model was designed, denoted as Network, which is a deep learning model consisting of two parts: Q-Network and Action-Network. Figure 4 As shown in Table 1, St represents the model image parameter input for the current time and the past N-1 time steps, including the sampled images from the visual sensor over the past N-1 time steps, the sampling frequency at each time step, and the sampling duration per frame.

[0046] Serial Number Input variables Remark 1 F(t) Visual sensor images 2 S Sampling rate 3 Tf Single frame sampling duration

[0047] Table 1. Types of input data.

[0048] As the set of predicted actions, and the output of the Action-Network, the Action-Network maps actions from model input to the next time step. This includes the sampling frequency and sampling duration per frame that should be set for the visual sensor at the next moment, as shown in Table 2.

[0049] Serial Number Input variables Remark 1 S’ Sampling rate 2 Tf' Single frame sampling duration

[0050] Table 2 shows the types of output data.

[0051] Q-Network in the current environment Using the action as input, the system obtains the maximum probability image pattern prediction for the next time step through a nonlinear mapping of a neural network. The reward is the image value actually sampled by the visual sensor at the next moment, which is calculated from the measured error index. The system reward represents the maximum benefit that the system can obtain in the long term. It is the noise level after adjusting the visual sensor parameters. The larger the reward, the smaller the noise. A negative reward indicates increased noise. Its rule design directly affects the model's self-learning ability.

[0052] L(w) is the L2 norm of the difference between the system's predicted reward and the actual reward, representing the error in the system's prediction. The training objective of the Q-Network is to make L(w) equal to 0, thereby obtaining an accurate calculation of the effect of eliminating out-of-step stripe noise in the sampled image.

[0053] According to an embodiment of the present invention, the network model of Q-Network samples a multilayer perceptron and performs a nonlinear mapping on the input current image, the current sampling rate, and the sampling duration to obtain the desired image pattern.

[0054] According to an embodiment of the present invention, Action-Network employs an LSTM network to perform convolution and accumulation calculations on images from the past N-1 time steps.

[0055] According to an embodiment of the present invention, the input model image parameter St includes sampled images from the visual sensor over the past N-1 time moments, the sampling frequency at each time moment, and the sampling duration per frame.

[0056] According to an embodiment of the present invention, the predicted action set includes the sampling frequency that should be set for the visual sensor at the next moment and the sampling duration value per frame.

[0057] In summary, this application can adaptively adjust the sampling rate and sampling duration of the visual sensor through algorithms and software. Even for display screen devices with unknown refresh rates, it can automatically search for frequencies to achieve frequency synchronization and thus accurately perform visual sampling. This application supports adaptive learning for different screen types, different visual sensor types, and different frequency ranges, and automatically learns as the scene changes, exhibiting high robustness. It also has a short measurement time for screen attributes and high efficiency.

[0058] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0059] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to the above embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0060] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for measuring dynamic solid color images on a screen based on pure visual perception, characterized in that, Includes the following steps: Obtain standard sample images for testing, collect standard sample image data, and perform adaptive learning to obtain the algorithm model; The algorithm model is used to sense the refresh rate of the display screen and automatically adjust the sampling parameters. Standard sample image data is acquired through a visual sensor, and adaptive learning is performed to obtain an algorithm model specific to this type of visual sensor; specifically including: Set the minimum and maximum refresh rates for the display screen; Read the sample image and display screen frequency range, set the screen to different refresh rate points in a certain step and display a standard image of 5 colors; Image data is acquired via a visual sensor under each display condition; Brightness change waveform signals were acquired using a photoelectric sensor; The visual sensor extracts and calculates features from the images it acquires, and adjusts the visual sensor's sampling parameters. The learning process stops when the frequency output of the algorithm matches the waveform of the photoelectric sensor. The algorithm model is denoted as Network, and it is a deep learning model. The algorithm model consists of two parts: Q-Network and Action-Network, specifically including: Input the model image parameters St to generate a set of predicted actions. ; Network in the current environment Using the action as input, the system obtains the maximum probability image pattern prediction for the next time step through a nonlinear mapping of a neural network. ; L(w) is the L2 norm of the difference between the system's predicted reward and the actual reward, which represents the error of the system's prediction. The training objective of the Q-Network is to make L(w) equal to 0, thereby obtaining data that eliminates out-of-step stripe noise from the sampled image. The Q-Network network model samples a multilayer perceptron and performs a nonlinear mapping on the input current image, current sampling rate, and sampling duration to obtain the desired image pattern; Action-Network uses an LSTM network to perform convolution and accumulation calculations on images from the past N-1 time steps; The input model image parameters St include the sampled images from the visual sensor over the past N-1 time steps, the sampling frequency at each time step, and the sampling duration per frame. The predicted action set includes the sampling frequency and the sampling duration per frame that should be set for the vision sensor at the next time step.

2. The method for measuring dynamic solid color images on a screen based on pure visual perception according to claim 1, characterized in that, The standard sample image is divided into 5 standard colors, namely red, green, blue, black and white.

Images