1.1  The Visual System as an Engineering Compromise

The purpose of this chapter is to outline the architecture and properties of the human visual system, but only to the extent required for a better understanding of its role as a substrate for visual prostheses. By sketching the properties of the healthy human visual system, we intend to provide the reader with an appreciation of the challenges one encounters in trying to reconstruct vision to the blind, even to a very modest level. Readers interested in more detailed or specific information regarding the visual system in health and disease are referred to some of the many excellent reviews in this area [12, 21, 22, 24, 49] and specifically to Chaps. 2–5 in this volume.

Evolution of the vertebrate visual system over several hundred million years has provided the human eye and higher visual processing centers with ingenious compromises to allow sharp central vision, a wide field of view, color perception, and an enormous range of light-to-dark adaptation. Note the following benchmarks, unparalleled by any single man-made system:

•The optic nerve, connecting the eye to the visual centers of the midbrain, has only approximately 1.2 million fibers [37] to represent the entire visual field (over 140° horizontally and 120° vertically, or roughly 3.6 × 107 arcmin²), in full color; a digital color camera with similar output bandwidth would provide 333,000 pixels, i.e., about 630 pixels across the field or 13.3 arcmin resolution. Yet the human eye achieves 1 arcmin resolution in the center of the visual field by

­combining variable cone photoreceptor spacing – from 0.4 arcmin (i.e., 1/150th of a degree, or 0.0067 mm; foveola) to 3 arcmin (far periphery)[1, 2, 17, 38] – with variable post-receptoral convergence – from (on average) three ganglion cells per cone in the fovea to one ganglion cell per 6 cones in the far periphery [16, 56].

•The three color filters used in digital cameras have narrower bandwidths and wider color separation than the three human cone types. Yet the post-receptoral interactions in human vision allow discrimination over a wider range of color space than can be physically created with common light sources and pigments (pp. 306 ff. in [66]).

•Both traditional cameras and the human eye employ mechanical apertures to adjust to a limited range of light levels (over 100 to 1 in cameras, about 15 to 1 for the human pupil). This, however, represents only a fraction of the dark adaptation range required by changes in natural lighting conditions. Low noise properties of CCD chips and a variety of automatic gain control mechanisms allow modern cameras to function over a brightness range from less than 1 to over 100,000 lux. Rod photoreceptors in the human eye, however, extend the downward range by at least a factor of 1,000 [38], while cone dynamics extend the upward range by at least a factor of 10. The rod system transmits its information to the brain using the same optic nerve fibers used by the cones.

•Compared to man-made detection and shape-recognition systems, the human visual system works quickly and with great precision: Attentional shift mechanisms