DSP seminar: October 23, 1997

Speaker: samit Basu

Title: Tomography of Moving Objects

Abstract:

In 2D parallel beam tomography, the problem is to recover the density of
an object in the plane given parallel line integrals of that
density along a set of known directions.  In certain situations, however,
these directions may be unknown.  Examples include cryo-EM based viral
imaging, and MR imaging of a moving patient.  In this talk, we will
examine the problem of tomographic reconstruction when the set
of directions at which data were collected is unknown.  Using the
consistency conditions on parallel beam tomographic data, we establish
conditions for the unique recovery of the set of directions directly from
the data.  We also calculate performance bounds that demonstrate the
direction recovery process to be feasible in the presence of noise.
Finally, as a demonstration of feasibility, we construct a simple
heuristic algorithm to estimate these directions, and present simulations
on synthetic data at moderate noise levels.