Purpose:
We previously described an ideal-observer based model of visual acuity, incorporating optical filtering, neural filtering, noise, and template matching (Watson and Ahumada, 2008, J. Vis.). We have shown how this model can be computed rapidly, to enable efficient calculation of acuity for arbitrary optotypes and optical aberrations. Here we have compared this model to acuity data for six human observers each viewing seven different optotype sets (Zhang, Zhang, Xue, Liu, & Yu, 2007, Invest. Ophthalmol. Vis. Sci.). The sets consisted of Sloan letters and six sets of Chinese characters, differing in complexity. We sought to determine the ability of the model to predict human letter identification, and to account for performance with a broad range of different optotypes.
Methods:
We drew a large population of wavefront aberrations for a 6mm pupil from a statistical model of normal human eyes (Thibos, Bradley & Hong, 2002, Ophthalmic Physiol Opt.). For each eye, we estimated acuity for each set of optotypes using the model observer and a Quest testing procedure. The final estimate of acuity was the mean of the population of eyes. The noise parameter of the model was optimized separately for each optotype set and observer.
Results:
Data for each observer and optotype set were fit well by the model, but the estimated noise parameter varied with optotype (Figure 1). Estimated noise was 1.4 times higher for more complex optotypes (sets 3-7) than for the Sloan letters or simplest Chinese characters (sets 1-2). We also compared confusion matrices for human and model observers. Correlations for off-diagonal elements ranged from 0.5 to 0.8 for different sets, and optimizing one parameter were competitive with correlations of a 13 parameter geometrical moment model (Liu, Klein, Xue, Zhang & Yu, 2009, J. Vis.).
Conclusions:
An ideal observer model of acuity gives a good account of performance of human observers with range of optotype sets varying in complexity, but the estimated noise must be adjusted for more complex optotypes. This is in close agreement with the variations in efficiency and complexity found by Pelli, Burns, Farell & Moore-Page (2006, Vis. Res.).