Abstract
We report experiments that address the interplay of three factors limiting visual resolution: optical and neural low-pass filtering, and sampling by the photoreceptor mosaic. The neural passband is roughly matched to the optical one. The neural limit for S cones (much lower than for L or M cones) may however serve to block frequency components that are phase-reversed, rather than obliterated, by chromatic defocus of the S cone image. As recently noted, optical aberrations can preserve such components at greater contrast than would be the case in an ideal system, but that may only increase the need to filter them out neurally, as they represent more of a threat than a benefit to vision. Other results challenge the prevailing view that the sampling frequency of the cone mosaic is closely related to optical or neural limits on visual resolution. We show that in responding to grating targets near the resolution limit, the visual system averages inputs from several rows of cones; with an irregular cone mosaic the effective sampling frequency thus becomes several times the Nyquist frequency for a regular array of the same spacing. Thin line stimuli preclude such integration, but with these stimuli the resolution limit is only about 20 cpd, about one third of the Nyquist frequency. Thus in all cases we have examined, the photoreceptor sampling frequency greatly exceeds the resolution limit.
NIH EY01711