Background and motivation
Images in space
Human vision works so well that it gives us the
impression that it's like touching -- what we see is THERE, in the
sense of existing away from us. Often we confirm this impression by
reaching out with our fingers, and are rarely surprised. But we use
vision so often to form and confirm our mental map of the world, and
this
works so well for us, that it obscures an important fact: the images we
see are NOT THERE.
Images are patterns of light at our eyes,
specifically at our pupils.
Many people have noted this fact, in many different
contexts. But it can be forgotten in the blink of an eye as those
wonderful patterns, at those small spots, are transformed by our
lens/retina/brain into what we see. We don't perceive what's at our
pupil,
we perceive a rich, full color world. But it is a fact that is worth
remembering and reimagining. The implications are astonishing.
A rich, full color world can be reconstructed from a
pattern of light passing through any few millimeter spot of space, over
a fraction of a second of time. And every space-time spot contains a
rich,
full color world representation. In principle, and nearly in practice,
eyes and
cameras can be made as small as the pupil, or aperture, and they are
used to capture information about the pattern of light at that spot.
Vision is the process of constructing a worldview
from images at spots, with pupils being the spots of choice.
So while we percieve the visual world as spread out around us,
perhaps light-years away in the night sky, in fact the physical and
information content of the visual world is
concentrated in tiny space-time spots.
[[ weak reciprocity (invariance) between an image
and a surface, conjugate nature of surfaces and light fields, and image
space ]]
Observation of one spot at one time, a "snapshot",
only gives a
weak impression of what's going on in the wide world. There's a range
of techniques we use to fully explore the richness of the light field
in space-time:
- Watch for change with time. Pay close attention to
correlated changes.
- Look at things from a different point of view.
Move your eye with respect to the things in the world.
- Closely compare images from two nearby viewpoints.
Most, but not all, people and many other animals are hardwired to make
this comparison in exquisite detail, using their two eyes
simultaneously (stereo vision).
- Change the lighting.
- Change focus, range of focus, aperture.
- Use a filter.
- Touch the surface, move the surfaces.
- Artificial color shifting.
Images that form our visual world can be
captured in degraded
form with an artificial eye, usually a camera, stored/transmitted, and re-presented to your eye giving the
impression that you are there, where the camera was/is. [[illustrate
diagramatically]] By far the most
common and convenient method for recreating images is some kind of flat
picture; a photographic print, a television or computer monitor, a
projected movie on a "silver" screen. But the recapitulation of the
image happens at the eye. Flat pictures are are an excellent
duplication method for technical and geometric reasons. It's difficult
to recreate complex images near or at the pupil, without having a kind
of copy well away from the eye. And a uniformly
luminous/scattering flat picture makes a similar image available to a
wide range of viewpoints, effectively multiplying the image -- many
people can view the same picture, with only a minor perspective
distortion in the image each observes. [[illustrate]]
Sampling image space
Neighboring images in image space
are not independent, lending coherence to visual experience.
[[illustrate in 2D space and time]]
Several methods have been exploited to present
multiple images in such a way to illustrate the world, either in
pseudo-realistic ways or in new ways.
- Movies. Pan, glide, zoom.
- High speed flash composite
- High magnification, scale
- Fish eye, wide angle composites
- Temporal integration
- Random access arrays
- Virtual reality, telepresence
- Simultaneous, multiple viewpoint synthetic video
Some of these methods can be generalized and
extended to create novel methods of experiencing the same essential
information
content; they all involve sampling of image space.
[[ riff about movies ]]
Dense Image Space (DIS)
DIS is a method of sampling image space at regular
intervals and retrieving/displaying the samples in a manner similar to
the way we naturally explore
the visual world. Modern display technology and computer random access
memory (RAM) capacity and speed allow reasonably fast access to
multiple images derived from the same subject matter (surfaces, or
locations
in space). The goal of DIS is to facilitate acquisition of
spatio-temporal arrays of images and their subsequent navigation and
rendering.
Dense Image Space
description
A
simple dense image space;
acquisition, image, and display
geometry
Example
data set and rough demo