A method for rapid detection of contrast sensitivity threshold
By using perturbation operator and Bayesian law to optimize stimulus intensity in contrast sensitivity threshold detection, the problems of long detection, low accuracy and high equipment requirements in the prior art are solved, and fast and accurate contrast sensitivity is achieved. Threshold detection, and reduces equipment demand, improves detection efficiency and convenience.
Patent Information
- Application Number
- CN202110121742.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-01-28
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2041-01-28
AI Technical Summary
The existing contrast sensitivity threshold detection method is too long, has low accuracy, is complex in operation and high equipment requirements, which leads to inconvenient detection.
Using a perturbation operator-based method, the stimulus intensity is optimized through Bayesian law and entropy calculation, adaptively adjust the detection process, quickly predict the contrast sensitivity threshold, and reduce equipment demand.
Fast and accurate contrast sensitivity threshold detection is achieved, reducing equipment thresholds, and improving detection efficiency and convenience.
Smart Images

Figure CN113158138B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of contrast sensitivity threshold detection, and in particular relates to a method for rapidly detecting a contrast sensitivity threshold. Background Art
[0002] The contrast sensitivity function describes the change of the inverse of the contrast threshold with spatial frequency. It is a tool in visual science to measure the degree of clarity of the boundary of objects that the human eye can distinguish. As one of the important indicators for measuring human vision, contrast sensitivity can not only be used to predict visual neuropathology, such as amblyopia, glaucoma, optic neuritis and diabetic retinopathy, but also can be used for other visual perception research, such as high dynamic range video coding based on contrast sensitivity function.
[0003] As early as 1760, contrast sensitivity was tested by illuminating a screen and an opaque rod using two candles. Later, with the development of printing technology, the bar graph method and the alphabet method emerged. These methods introduced modulation transfer functions to create corresponding bar graphs or alphabet graphs, and contrast sensitivity was tested by judging these graphs. With the continuous development of computer programming and hardware, many solutions have been proposed by domestic and foreign academic and industrial circles, including several technical solutions:
[0004] The invention patent (application number: CN200910192292.2, name: A method and system for visual contrast sensitivity test based on network) discloses a method for visual contrast sensitivity test based on network, including the following steps: the client downloads the contrast sensitivity test program from the server, which can generate a series of grating images of contrast threshold changes at a predetermined spatial frequency, and obtains the contrast sensitivity test results of the subject at all spatial frequencies by executing the contrast sensitivity test program. The invention also discloses a system for visual contrast sensitivity test based on network, wherein the server includes a sending unit storing a contrast sensitivity test program for generating a series of grating images of contrast threshold changes at a predetermined spatial frequency; the client includes: a receiving unit for downloading the contrast sensitivity test program from the server; an execution unit for executing the contrast sensitivity test program to obtain the contrast sensitivity test results of the subject at all spatial frequencies. The grating image of this method is predetermined, and can only detect the corresponding contrast grating at the corresponding spatial frequency, so the efficiency is not high; at the same time, the detection clients are all computers, the detection equipment requirements are too high, and it is not convenient enough.
[0005] Lesmes LA, Lu ZL, Baek J, et al (Bayesian adaptive estimation of the contrast sensitivity function: the quick CSF method. J Vis, 2010, 10 (3): 17.) This method first determines the contrast sensitivity model, and then directly estimates and analyzes the four parameters of the contrast sensitivity model after each test through Bayesian analysis and multiple test information acquisition technology, and calculates and gives stimulation before the next test. The contrast sensitivity function can be obtained quickly after about 50 tests. However, when the model and parameter range are not selected properly, this method will introduce a lot of errors into the measurement results. At the same time, it lacks local information and cannot make adjustments based on local damage to contrast sensitivity. The detection equipment is also a computer, and the detection is not convenient enough.
[0006] In summary, these methods all have the following problems:
[0007] 1. The detection time is too long and the accuracy is low. Some methods improve the detection speed, but the accuracy also decreases accordingly; some methods strike a balance between detection speed and accuracy.
[0008] 2. The operation is too complicated and the detection is not convenient enough. The detection of these methods requires the support of computer equipment, which greatly increases the detection conditions. Summary of the invention
[0009] In view of the above problems existing in the prior art, an object of the present invention is to provide a method for quickly detecting a contrast sensitivity threshold, which method has short detection time, high detection result accuracy and low requirements on detection equipment.
[0010] The present invention provides the following technical solution: a method for quickly detecting contrast sensitivity threshold, characterized in that it comprises the following steps:
[0011] S1: Select the measurement function Ψ λ (x) and the maximum number of iterations N, initialize the current number of iterations, the prior probability density p 0 (λ) and a lookup table of conditional probabilities p(r|λ,x);
[0012] S2: Calculate the probability p of getting feedback next time t (r|x);
[0013] S3: Predict the probability p of each measurement function t (λ|x, r);
[0014] S4: The entropy H of the probability density function of the predicted measurement function in the entire parameter space t(x, r);
[0015] S5: Predict the expected entropy E[H for each stimulus t (x)];
[0016] S6: Find the stimulus intensity x that minimizes the expected entropy t+1 ;
[0017] S7: Stimulus intensity x′ after adding the designed perturbation operator t+1 ;
[0018] S8: The feedback obtained after presenting the stimulus with the perturbation operator added to the subject is calculated to obtain the posterior distribution p t+1 (λ);
[0019] S9: Calculate new parameter prediction λ t+1 At the same time, the current iteration number increases automatically;
[0020] S10: When the current iteration number is equal to the input maximum iteration number, the threshold is output, otherwise jump to step S2.
[0021] The method for quickly detecting contrast sensitivity threshold is characterized in that in step S1, the larger the maximum number of iterations selected, the more accurate the detected threshold is, and the longer the measurement time is.
[0022] The method for quickly detecting contrast sensitivity threshold is characterized in that in step S2, the probability of the predicted feedback obtained after the next stimulus presentation being correct or wrong is calculated, and all possible probability situations can be obtained.
[0023] The method for quickly detecting contrast sensitivity threshold is characterized in that in step S3, the posterior probability is predicted by the Bayesian rule to obtain the probabilities of all possible parameters based on the current stimulus and the subject's feedback.
[0024] The method for rapidly detecting contrast sensitivity threshold is characterized in that in steps S4 and S5, the expected entropy of each possible stimulus is calculated by predicting the entropy of the measurement function in the entire parameter space, wherein the measurement function is a function with stimulus intensity as an independent variable, which describes the instability of the prediction.
[0025] The method for quickly detecting contrast sensitivity threshold is characterized in that in step S6, the contrast stimulus when the expected entropy reaches the minimum value is obtained by calculating the expected entropy.
[0026] The method for rapidly detecting contrast sensitivity threshold is characterized in that in step S7, a perturbation operator is designed to adaptively optimize the stimulus presented next time, thereby accelerating the prediction of the threshold.
[0027] The method for quickly detecting contrast sensitivity threshold is characterized in that in step S8, the stimulus after adding the perturbation operator is presented to the subject, and after obtaining real feedback, the posterior distribution, that is, the probability of all possible parameters, is calculated.
[0028] The method for rapidly detecting contrast sensitivity threshold is characterized in that in step S9, new parameter prediction is obtained by performing expectation calculation on the posterior distribution.
[0029] The method for quickly detecting contrast sensitivity threshold is characterized in that in step S10, the execution of the method for detecting contrast sensitivity threshold is controlled by judging whether the current iteration number reaches the maximum iteration number until the maximum iteration number is reached, which can effectively judge the consumption time of the method for detecting contrast sensitivity threshold.
[0030] By adopting the above technology, compared with the prior art, the beneficial effects of the present invention are as follows:
[0031] The present invention is based on a perturbation operator. On the basis of analyzing the feedback information of the current subject, it adaptively optimizes and adjusts the stimulus, speeds up the threshold prediction, ensures the accuracy, and constructs a system for rapid detection of contrast sensitivity. It is transplanted to various platforms, reduces the demand for equipment, allows users to perform detection simply and conveniently, lowers the threshold for user detection, and improves the efficiency of prediction. BRIEF DESCRIPTION OF THE DRAWINGS
[0032] Figure 1 The present invention is a flow chart of the method.
[0033] Figure 2 Schematic diagram of the rapid detection contrast sensitivity threshold of the present invention. DETAILED DESCRIPTION
[0034] 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 of the specification. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.
[0035] On the contrary, the present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention as defined by the claims. Further, in order to make the public have a better understanding of the present invention, some specific details are described in detail in the detailed description of the present invention below. Those skilled in the art can fully understand the present invention without the description of these details.
[0036] See also Figure 1-2 , a method for quickly detecting contrast sensitivity threshold, comprising the following steps:
[0037] S1: Select the measurement function Ψ λ (x) and the maximum number of iterations N, initialize the current number of iterations, the prior probability density p 0 (λ) and the lookup table p(r|λ, x) of conditional probability. The formula of the lookup table p(r|λ, x) is as follows:
[0038] p(success|λ,x)=Ψ λ (x)
[0039] p(failure|λ,x)=1-Ψ λ (x)
[0040] The measurement function defaults to the Gumbel equation, and the prior probability density p 0 (λ) is obtained by multiplying the normal distributions of the parameters.
[0041] The larger the maximum number of iterations selected, the more accurate the threshold detection will be, but the longer the measurement time will be.
[0042] S2: Calculate the probability p of getting feedback next time t (r|x), the specific formula is as follows:
[0043]
[0044] S3: Predict the posterior probability p of each measurement function through Bayes' rule t (λ|x, r), and obtain the probability of all possible parameters based on the current stimulus and the subject's feedback. The specific probability formula is as follows:
[0045]
[0046] S4: The entropy H of the probability density function of the predicted measurement function in the entire parameter space t (x, r), the formula is as follows:
[0047]
[0048] S5: Predict the expected entropy E[H for each stimulus t (x)], as follows:
[0049] E[H t (x)]=H t (x,success)p t (success|x) +H t (x,failure)p t (failure|x)
[0050] S6: The stimulus intensity x when the expected entropy reaches the minimum value is calculated by the expected entropy t+1 , the expression is as follows:
[0051]
[0052] S7: Stimulus intensity x′ after adding the designed perturbation operator t+1 , by designing a perturbation operator, adaptively optimizing the stimulus presented next time, speeding up the prediction of the threshold, the specific formulas of the designed perturbation operator and the added perturbation operator are as follows:
[0053]
[0054]
[0055] S8: After presenting the stimulus with the perturbation operator to the subject and obtaining real feedback, the posterior distribution p is calculated. t+1 (λ), that is, the probability of all possible parameters, is calculated as follows:
[0056] p t+1 (λ) = p t (λ|x′ t+1 , r t+1 )
[0057] S9: By performing expected calculation on the posterior distribution, we get the new parameter estimate λ t+1 At the same time, the current iteration number increases automatically, and the parameter is expected to be t+1 The specific formula is as follows:
[0058]
[0059] S10: If the current iteration number is equal to the input maximum iteration number, the threshold is output, otherwise jump to step S2, which controls the execution of the method for detecting the contrast sensitivity threshold by judging whether the current iteration number reaches the maximum iteration number. The loop is continued until the maximum iteration number is reached, which can effectively judge the time consumed by the method for detecting the contrast sensitivity threshold.
[0060] See also Figure 2 , a system for quickly detecting contrast sensitivity, mainly includes building a method for quickly detecting contrast sensitivity threshold into an independent system, and then porting it to various platforms, which lowers the equipment threshold and facilitates user detection.
[0061] 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 method for rapidly detecting contrast sensitivity threshold, Features: The following steps are involved: S1: Select the measurement function Ψ λ (x) and the maximum number of iterations N, initialize the current number of iterations, the prior probability density p 0 (λ) and a lookup table of conditional probabilities p(r|λ,x); S2: Calculate the probability p of getting feedback next time t (r|x); S3: Predict the probability p of each measurement function t (λ|x, r); S4: The entropy H of the probability density function of the predicted measurement function in the entire parameter space t (x, r); S5: Predict the expected entropy E[H for each stimulus t (x)]; S6: Find the stimulus intensity x that minimizes the expected entropy t+1 ; S7: Stimulus intensity x′ after adding the designed perturbation operator t+1 : In S7, the stimulus intensity x′ after adding the designed perturbation operator t+1 , by designing a perturbation operator, adaptively optimizing the stimulus presented next time, speeding up the prediction of the threshold, the specific formulas of the designed perturbation operator and the added perturbation operator are as follows: S8: Present the stimulus with the perturbation operator added to the subject, and after obtaining the real feedback, calculate the posterior distribution p t+1 (λ), i.e., the probability of all possible parameters; S9: Calculate new parameter prediction λ t+1 At the same time, the current iteration number increases automatically; S10: When the current iteration number is equal to the input maximum iteration number, the threshold is output, otherwise jump to step S2.
2. A method for rapidly detecting contrast sensitivity threshold according to claim 1, Features In step S1, the larger the maximum number of iterations selected is, the more accurate the detection threshold is, and the longer the measurement time is.
3. A method for rapidly detecting contrast sensitivity threshold according to claim 1, Features In step S2, the probability of the predicted feedback obtained after the next stimulus presentation being correct or wrong is calculated, and all possible probability situations can be obtained.
4. A method for rapidly detecting contrast sensitivity threshold according to claim 1, Features In step S3, the posterior probability is predicted by Bayes' rule to obtain the probabilities of all possible parameters based on the current stimulus and the subject's feedback.
5. The method for rapidly detecting contrast sensitivity threshold according to claim 1, Features In the steps S4 and S5, the expected entropy of each possible stimulus is calculated by predicting the entropy of the measurement function in the entire parameter space, wherein the measurement function is a function with stimulus intensity as an independent variable and describes the instability of the prediction.
6. A method for rapidly detecting contrast sensitivity threshold according to claim 1, Features In step S6, the contrast stimulus that makes the expected entropy reach the minimum value is obtained through expected entropy calculation.
7. A method for rapidly detecting contrast sensitivity threshold according to claim 1, Features In step S9, new parameter predictions are obtained by performing expectation calculation on the posterior distribution.
8. A method for rapidly detecting contrast sensitivity threshold according to claim 1, Features In step S10, the execution of the method for detecting the contrast sensitivity threshold is controlled by judging whether the current iteration number reaches the maximum iteration number, until the maximum iteration number is reached, so that the consumption time of the method for detecting the contrast sensitivity threshold can be effectively judged.
Citation Information
Patent Citations
Method and system for inspecting visual contrast sensitivity based on network
CN101647694A
System And Method For The Rapid Measurement Of The Visual Contrast Sensitivity Function
CN107224262A
Optimization method and device
CN107844860A
Adversarial sample generation method based on Bayesian optimization
CN110276377A
Methods and Devices for Rapid Measurement of Visual Sensitivity
US20100007851A1
