Purchase this article with an account.
Eline Kupers, Marisa Carrasco, Jonathan Winawer; Towards a computational observer model of perceptual performance fields. Journal of Vision 2018;18(10):212. doi: https://doi.org/10.1167/18.10.212.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Background: Visual performance depends on polar angle when eccentricity is held constant; on many psychophysical tasks observers perform best on the horizontal meridian, worst on the upper vertical, and intermediate on the lower vertical meridian. This variation in performance with polar angle, called 'performance fields', can be as pronounced as that of doubling the target eccentricity. The cause of these asymmetries in performance is largely unknown. Goal: To investigate the extent to which factors in the eye contribute to performance fields, we implemented a first stage computational observer model focusing on the front-end of the visual system. Methods: The stimuli were static Gabor patches at 4.5° eccentricity, 4 cpd, 1° SD, oriented 20° clockwise or counterclockwise from vertical, with contrast varying from 0% to 20%, presented for 200 ms. Our computational observer model began with the full spectral representation of the stimulus, which was jittered by small fixational eye movements (tremor and drift), and transformed by the human optics and spatial and spectral sampling of the cone mosaic. The time-varying 2D array of cone absorptions was used as input to a linear support vector machine classifier, which performed a 2AFC discrimination (clockwise / counterclockwise). We asked how changes in defocus and cone density affect model performance. Results: To account for the nearly doubling of thresholds for upper vertical compared to horizontal meridian, as observed psychophysically on the same task, our computational observer model would require either an increase of ~5 diopters (D) of blur or a 50% reduction in cone density. Because cone density at 4.5 deg eccentricity varies only by about 10% across polar angles, and blur at 4.5 deg is typically less than ~0.25 D, these factors account for a small fraction of performance fields. Substantial additional asymmetries must arise in retinal or cortical processing.
Meeting abstract presented at VSS 2018
This PDF is available to Subscribers Only