Analysis of discrete-to-discrete imaging models for iterative tomographic image reconstruction and compressive sensing
Year: 2011
Type: preprint
Abstract: Discrete-to-discrete imaging models for computed tomography (CT) are becoming increasingly ubiquitous as the interest in iterative image reconstruction algorithms has heightened. Despite this trend, all the intuition for algorithm and system design derives from analysis of continuous-to-continuous models such as the X-ray and Radon transform. While the similarity between these models justifies some crossover, questions such as what are sufficient sampling conditions can be quite different for the two models. This sampling issue is addressed extensively in the first half of the article using singular value decomposition analysis for determining sufficient number of views and detector bins. The question of full sampling for CT is particularly relevant to current attempts to adapt compressive sensing (CS) motivated methods to application in CT image reconstruction. The second half goes in depth on this subject and discusses the link between object sparsity and sufficient sampling for accurate reconstruction. Particularly, it is pointed out that, because CS motivated image reconstruction is object dependent, there is a need to consider the imaging task so that test phantoms are employed with a similar sparsity level as what might be encountered.
Source: arXiv (Cornell University)
Institutions: Technical University of Denmark, University of Chicago
Language: English
Subfield: Computational Mechanics
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W290896398