Purchase this article with an account.
Jared Abrams, Chris Bradley, Wilson S. Geisler; A model of detectability across the visual field. Journal of Vision 2013;13(9):273. doi: 10.1167/13.9.273.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Accurately characterizing peripheral vision is a fundamental goal of vision science. Additionally, predicting the detectability of a peripheral target in healthy adults has important implications for understanding visual dysfunction, as well as in designing displays. Here, we propose a model of detectability across the visual field. Our model filters the target image by the optics, samples the image with an array of model ganglion cells with Difference of Gaussians receptive fields, applies a rule similar to d’-summation, and then converts the result into a detection threshold. If the background is not uniform, that threshold is adjusted by taking the local luminance, spatial uncertainty, and background contrast power affecting the target template into account. This model (1) takes the physiology and anatomy of early vision into account, (2) is consistent with the extant psychophysical literature, (3) has the flexibility to account for any target/background combination, and (4) is computationally fast. In this study, we measured contrast detection thresholds for three targets (Gabor, Gaussian, Edge) at four eccentricities (0, 2.5, 5, & 10°) along the horizontal meridian on a uniform background. In this case, the modeled parameters are the foveal center size for the ganglion cell receptive fields, the size of the center-surround scalar, the center weight, the exponent for summation, and the internal noise. The first three parameters have direct analogs in the physiological literature, while the remaining two have been examined psychophysically. All thresholds are well fit by the model. Additionally, we found that our model provides a good fit for the ModelFest data set. The parameter estimates are within established limits defined by previous literature. Thus, we are able to predict detection thresholds for multiple targets on a uniform background. This model can serve as a framework for models of other phenomena such as visual search, attention, or learning.
Meeting abstract presented at VSS 2013
This PDF is available to Subscribers Only