Abstract
Advances in imaging technology have to a large extent shaped
scientific progress in the last 200 years. While progress in imaging
technology originated in, and forced the development of, the field of
optics, the design paradigm for optical instruments has always placed
the human observer at the center of its efforts. With the advent of
electronic computation in the second half of the 20th century, optical
design could be elevated to a new level by exploiting computer-aided
design and automated optimization procedures.
However, only in recent years have computers become so powerful, and
at the same time so small and inexpensive, that imaging technology,
storage and transmission have become completely digitized. This move
has not yet reached its full potential since the human observer is
still considered the target of optimization, whereas in fact, today's
primary observers are computers. It is this insight that enables an
entirely new approach to optics and measurement
instrumentation. Images no longer have to mimic what the human brain
is accustomed to interpret as an image of the world, i.e. integrals
over ray bundles of a restricted subset of the electro-magnetic
spectrum. Instead, sensing mechanisms can be designed that
re-distribute directional, spatial, temporal and wavelength
information to essentially agnostic sensor elements serving as simple
photon collectors.
The questions of how such redistribution can be arranged for, which
performance characteristics are to be expected of such devices, and
how these novel sensing means can be used for measurement purposes
form the basis and contents of this thesis. In particular examples,
investigations into these larger questions are explored in
detail. Through these studies we arrive at a larger picture of the
current state of affairs: We have glimpsed at the exciting
possibilities of computational optical imaging, however, we have seen
a mountain of formidable size and difficulty, the ascend of which will
require significant effort. This is how the subtitle of this thesis is
to be understood.
