SPATIAL AND FORM VISION: EARLY PROCESSING

The optical system of the eye produces a 2-dimensional image that is sampled by a huge array of photoreceptors. In capturing, processing, and encoding the spatial information at the inital levels (within the eye), the visual system must solve two major problems and operate with three serious constraints. Many aspects of our spatial vision reflect the way in which the visual system goes about solving these problems with these constraints.

One major problem is that of transducing the tiny energy of single photons into neural activity which involves the relatively massive flow of ions across cell membranes. A second problem is that of separating the tiny amount of information in the image about reflective objects (the useful information) from the massive variations in light due to the illuminant (relatively useless information), when the two are totally confounded at each point in the image. One serious early constraint on spatial vision is the bottleneck of the optic nerve, which severely restricts how much information can be sent from the eye to the brain. A second constraint is that the system must simultaneously process and encode color and spatial information, with different requirements for the two and the need to send both kinds of information up through the optic nerve bottleneck. A third constraint is the very limited capacity of axons to carry information through spike firing rates while operating in real time.

The receptors amplify the varying photon-capture signal with an enzymatic cascade that provides temporal integration at the cost of a slow response and low temporal resolution. To start the separation of illuminant from reflective object, the visual system attenuates very low temporal and spatial frequencies, in the retinal processing. It requires many more neurons to encode high than low spatial frequency information, so the visual system processes and transmits high spatial frequency information only over a small retinal region. Furthermore, spatial information is largely multiplexed with color information to maximize information transmission through the optic nerve bottleneck, and there are parallel systems encoding increments and decrements (and color changes in opposite directions) around the mean illumination level.