Eye Tracking Analysis For Visual Quality And Visual Health Assessment

The eye is known as the window of the mind,it is the most important information input system in human information processing system.In order to see an object clearly,our eyes must maintain a certain orientation in order to make the object clearly imaged on the retina.Human eye movement can reflect people's interest,psychological activity and physical state.The development of eye tracker technology allows us to easily obtain human eye movement information.Eye trackers are widely used in human-computer interaction,fatigue assessment,and measurement of psychological and physiological indicators.Based on the eye tracker,this thesis deeply investigates and solves three major problems that people care about: the design and the evaluation of subjective test of image quality assessment,the assessment of eye fatigue when people are watching electronic screens,and the prediction of the effects of amblyopia treatment.First,Subjective quality assessment serves as the foundation for almost all visual quality related researches.The size of the image quality databases has expanded from dozens to thousands in the last decades.Since each subjective rating therein has to be averaged over quite a few participants,the ever-increasing overall size of those databases calls for an evolution of existing subjective test methods.Traditional single/double stimulus based approaches are being replaced by multiple image tests,where several distorted versions of the original one are displayed and rated at once.This naturally brings upon the question of how to arrange those multiple images on screen during the test.In this thesis,I answer this question by performing subjective viewing tests with an eye tracker for different types of arrangements.My research indicates that isometric arrangement imposes less duress on participants and has a more uniform distribution of eye fixations and movements and therefore is expected to generate more reliable subjective ratings.Second,eye fatigue and associated eye disease have become increasingly severe as the digital device progresses in the last decade.If someone knows how fatigued his eyes are,he can relax and protect his eyes timely.In this thesis,I provided this issue into two distinct application scenarios: liquid-crystal display(LCD)users and head-mounted display(HMD)users.For the LCD users,I provided a novel definition of eye fatigue and a personalized real time eye fatigue assessment model.Questionnaire is an inaccurate and inconvenient way to assess eye fatigue.I proposed an objective assessment in real time.Many previous studies define eye fatigue by only one optometry characteristic.In this research,eye fatigue is defined by seven characteristics.The assessment model is based on an unobtrusive eye tracker.Previous eye-tracking-based methods can judge eye fatigue but can not give a fatigue level.While in this thesis,I selected3 eye movement characteristics based on which different performance of assessment operators are investigated.The proposed eye fatigue assessment model provides a computational measure with results that are well correlated to true eye fatigue values.In addition to the LCD,HMDs and virtual reality(VR)have been frequently used in recent years,and user's experience and computation efficiency could be assessed by mounting eye trackers.However,in addition to visually-induced motion sickness(VIMS),eye fatigue has increasingly emerged during and after the viewing experience,highlighting the necessity of quantitative assessment of the detrimental effects.In this thesis,I showed the differences between VIMS and eye fatigue,and presented a new scheme to assess eye fatigue of HMDs users by analysis of parameters of the eye tracker.As no measurement method for eye fatigue has been widely accepted.I detected parameters related to optometry test.I proposed a novel computational approach for estimation of eye fatigue by providing various verifiable models.I also implemented three classifications and two regressions to investigate different feature sets,which led to present two valid assessment models for eye fatigue by employing blinking features and eye movement features with the ground truth of indicators for optometry test.As a result,three graded results and one continuous result were provided,respectively by each model,which caused the whole result to be repeatable and comparable.Finally,I proposed an algorithm to predict the treatment effect of the children with amblyopia.Gazing following training is a widely used adjuvant therapy for amblyopia.Researchers found that the therapeutic effect of amblyopia was closely related to the patient's compliance.I propose an objective algorithm to quantify the performance of the children with amblyopia in the gaze following training.In addition,I propose to predict the patients' treatment effect according to their performance at the first time training.We recruited 30 children with anisometropic amblyopia and 30 healthy children.Every participant was asked to complete a 2-minute gaze following training in front of a monitor.They needed to follow the red dot moving on the screen.A high-speed eye tracker recorded their gaze locations in the whole process.I extracted 20 eye movement features which might suggest the performance of gaze following.I accumulated each feature in 2 minutes.We compared the difference of each feature between the children with and without anisometropic amblyopia using significance test.Then,we accumulated each feature in 20 seconds and supervised individual's variation trend of the performance.We compared the corrected visions of the subjects at the beginning and three months later.Among the 20 eye movement features,only 6 features did not show a significant difference between the amblyopia group and the control group.The other 14 features are sensitive to the children with anisometropic amblyopia.Based on the 14 eye movement features,I found that the gaze following performance of some subjects varied greatly throughout the training process.Four features can be the indicators to distinct whether the child could get his/her corrected vision improved or not after treatment.In summary,this thesis based on eye trackers can mainly be separated into three parts: the measurement of human eye workload for subjective image quality assessment,the eye fatigue assessment caused by LCD and HMD,and the prediction of amblyopia treatment.