The general problem considered in the om
nt+ project is to best
solve a system of integral equations
with a lattice constraint of the form where
and
are potentially infinite, but measurable,
functions. Examples abound, including spectral estimation,
spectroscopy, and of particular interest to the
om
nt+ project,
tomography.
Since the unknown is a function, residing in a known function
space
, the integral equations (1) are not
sufficient to uniquely determine
. If the equations are
consistent, they are underdetermined: there are an infinity of
solutions. To overcome this difficulty, we select the ``best''
function by minimizing some measure of the function. Mathematically,
we seek solutions to
Writing
for the system in (1), and
incorporating the lattice constraints in the objective function
by
adding infinities as needed, we have the exactly constrained problem
If the data vector
, or the measurement process
, is known
to be inexact, we may relax the constraints to the form