Abstract
Acuity is the most widely used measure of visual function employed in both research and clinical settings. It is an estimate of the minimal size at which a particular set of symbols (optotypes) can be identified reliably. To understand the role of optical and neural contributions, we have developed a computational model of visual acuity.
Our model includes rendering of the retinal image by an optical point-spread function, anisoplanatic filtering of the retinal image by an array of midget retinal ganglion cells, perturbation by ganglion cell noise, and classification using an optimal template-matching procedure. We call this the Neural Image Classifier (Watson & Ahumada, 2015).
This model builds on ideas from optical simulation (Artal et al., 1989), ideal observer models (Geisler, 1989), and letter identification (Beckmann & Legge, 2002; Chung et al., 2002; Dalimier & Dainty, 2008; Gold et al., 1999; Nestares et al., 2003; Parish & Sperling, 1991; Watson & Fitzhugh, 1989).
For a given optical and neural configuration, acuity values can be estimated by conducting psychophysical trials using Monte-Carlo simulation. The model relies on other models we have developed of pupil diameter (Watson & Yellott, 2012), optical point-spread (Watson, 2013), and distribution of retinal ganglion cells (Watson, 2014).
The model has been used to predict effects on acuity of particular wavefront aberrations (Watson & Ahumada, 2008), to predict acuity for optotypes varying in complexity (Watson & Ahumada, 2012), and to predict the effects of size on contrast thresholds for letter identification (Watson & Ahumada, 2015).
Here we describe elements of the model and illustrate how it is used to compute predictions of acuity.
This work supported by the NASA Space Human Factors Research Project WBS 466199.