392
How closely does the perceived depth correspond with the real depth?
Over how large a range of depth may objects be seen simultaneously as single?
Since stereoscopy depends upon the fusion of dissimilar images, what are the
limits of exaggeration before stereoscopy breaks down?
What movements do the eyes make during the observations?
The first question is perhaps the best known, but the others have suggested
equally interesting lines of investigation. Let us consider some of the answers in
their relation to photogrammetry.
In photogrammetry the first question may be asked in the form, “Within what
limits may the floating mark be set on the terrain?” Indeed, on this precision
depends the ultimate accuracy of the human side of photogrammetry. There is,
however, a distinction between precision and accuracy, which will be discussed
later.
It is found that parallax differences of only 2 seconds of arc can be detected
under the best conditions/ 43 which means for example that two objects about
1 metre from the observer need be separated by only 0T5 mm. in order to appear
at different depths. It is strange that stereoscopic acuity should be some thirty
times better than monocular acuity, which is about 1 minute of arc/ 43 but a partial
explanation is that monocular acuity depends on a process of boundary recognition,
whilst stereoscopic acuity relies on pattern recognition over large areas/ 5,63 Thus,
although the clarity of the targets is very important for their resolution in the plane
at right angles to the line of sight, it is comparatively easy to separate the most
vague patterns into planes of depth. This is, of course, very important to photo-
grammetrists who may have to work with extremely hazy pictures because of bad
aerial conditions. They may not be able to recognise the nature of the details on the
terrain, but they will see precisely the differences in level.
Unfortunately, although observers can set precisely on a blurred photograph,
they may not always set very accurately. They may in fact set the floating mark
consistently below the surface, especially if they are fatigued. This seems to be
due to the psychological difficulty of recognising that the mark has gone through
an apparently solid surface, especially if the latter is not well defined/ 73 Therefore
observers will sometimes be content to leave the mark just “underground”, thinking
it still to be on the surface, whilst immediately detecting when it is just in front of
the surface. Hence the average of the readings will be weighted too low. The
inaccuracy might not be very great, the parallax error being less than 1 minute of arc
under very bad conditions, but the readings might still look consistent to within
± 20 seconds of the false mean, and so give an unrealistic impression of their accuracy.
Another interesting optical “illusion” is the false impression of depth produced
by objects of highly saturated colour. Thus, red objects might appear to be nearer
than blue, or vice versa, although all the objects are in fact at the same distance.
This is obviously relevant to the use of colour photographs in photogrammetry.
One explanation is that the optical axis of the eye does not usually coincide with its
visual axis, and that the eye is not achromatic, as discussed by Vos. [8] Consequently,
images of differently coloured objects may fall on the retinas with slight shifts caused
by the dispersion. These shifts may then be interpreted as due to changes in depth.
The directional sensitivity of the retina may also play a part in this effect,
for the retina is constructed rather like an inverted carpet with its “pile” of rods
and cones lying towards the back of the eye, and the almost transparent “backing”
of nerves and blood vessels facing the light. Radiation is most successful in penetrat
ing into the pile when it falls normally on the retina, that is from some point in the
central region of the lens. Thus rays from the margin of the lens are not so effective