|
chance situations and few of the models seem to include an intentional
allowance for the imperfections of the retinal image.
If all that one required to know about visual performance was concerned with
detection of simple, isolated objects in plain fields, use of one or more of these
models would suffice for all situations. However, vision in real life usually is not
as simple as that. Many scenes are highly structured, many objects have textured
surfaces, many objects have multiple contrasts against their immediate back-ground
(indeed what is the meaning of contrast when an object of interest has
areas of different luminance or is seen against a non-uniform immediate
background? ) In addition natural objects often have ‘frilled’ edges (e.g. trees and
bushes) which tend to blur the outline (see for instance Chapter 11). Yet again
some objects are characterised by a line profile with no solid centre (as in any
outline drawing) or are merely a discontinuity in the field of view, unbounded
on one or more sides and hence incapable of being represented as an area (e.g. a
horizon, particularly between sea and sky).
In order to supplement existing models, where there is a difficulty in treating
such conditions, an attempt was made by the author at modelling visual
performance based on known general physical properties of the eye and an
assumed logical form of simple, first stage processing by the central nervous
system (i.e. processing aimed at detecting existence of local stimuli rather than
recognition of any detail). The outcome was the progressive development of a
model of the visual process which is presently able to predict, using a simple set
of constants, a very wide range of simple detection thresholds, to relate many
sets of isolated data one to another, and by its very nature to allow extrapolation
into the complex regimes of practical vision. As will be seen later,
|
