Parameter estimation in tomography
- Speaker:
- Joost Batenburg (CWI)
- Date / time:
- Tuesday, April 27, 2010, 11:00
- Location:
- Buys Ballot Laboratory, Room 165, Utrecht (BBL 165)
Abstract
Tomography is a powerful technique for three-dimensional imaging that
deals with image reconstruction from a series of projection images,
acquired along a range of viewing directions. In recent years,
significant progress has been made in reconstructing images from limited
projection data (i.e., few projections, truncated projections, limited
angular range), by incorporating various forms of prior knowledge.
Most of the available theory on tomographic image reconstruction is based on
the assumption that certain properties of the acquisition geometry
(angles, alignment) and possibly also of the object (densities, shape
properties) are known in advance to good accuracy. In practice, however,
these properties are often known only with limited accuracy, or are even
completely unknown. Incorporating these unknown parameters in the
classical reconstruction problems yields a wide range of new research
questions, as well as significant computational challenges.
In this talk I will introduce the basic concepts of tomography and
tomographic reconstruction and present some recent theoretical and
practical results. Subsequently, I will provide an overview of parameter
estimation problems and the possibilities for solving them, either by
elegant mathematical manipulation or by large-scale scientific
computing, exploiting the power of modern parallel hardware such as
GPUs.