Pupil swim is a kind of distortion resulting from imperfect lens distortion correction of Virtual Reality Head-Mounted Displays (VR-HMDs) and eye motion. Because of pupil swim, there is a sensation of movement or optic flow in parts of the virtual environment that the users expect to be stationary, which can be disturbing. New VR-HMD prototypes need to be tested for pupil swim, however psychophysical modelling of pupil swim perception is preferable as it is more efficient. Existing studies on the psychophysical effects of different optic flow patterns cannot fully explain pupil swim as these studies mainly focused on optic flow perceived in self-motion. This concept was validated by decomposing pupil swim into a linear sum of common components. The difference between the original pupil...[ Read more ]

Pupil swim is a kind of distortion resulting from imperfect lens distortion correction of Virtual Reality Head-Mounted Displays (VR-HMDs) and eye motion. Because of pupil swim, there is a sensation of movement or optic flow in parts of the virtual environment that the users expect to be stationary, which can be disturbing. New VR-HMD prototypes need to be tested for pupil swim, however psychophysical modelling of pupil swim perception is preferable as it is more efficient. Existing studies on the psychophysical effects of different optic flow patterns cannot fully explain pupil swim as these studies mainly focused on optic flow perceived in self-motion. This concept was validated by decomposing pupil swim into a linear sum of common components. The difference between the original pupil swim optic flow map and an artificial optic flow map reconstructed using the components was an inverse metric for the ability of these patterns to represent pupil swim. Components were algorithmically selected from the basis set of Zernike vector polynomials, which include common lens aberrations. It was found that the identified components could represent pupil swim with higher accuracy compared to optic flow from self-motion alone. Moreover, the linear coefficient fitted to a component represented its magnitude in the pupil swim. By determining the quantitative relationship between component magnitude and perception, a predictive model for pupil swim perception was developed. This predictive model was trained and validated through a user study of pupil swim from real VR-HMD lens designs. This perception model showed sufficient accuracy in predicting subject-reported disorientation scores for different pupil swim cases.