Detectability of anomalies from a few noisy tomographic projections
| Authors: | Fillatre Lionel, Université de Technologie de Troyes, France, Metropolitan
Retraint Florent, Université de Technologie de Troyes, France, Metropolitan
Nikiforov Igor, Université de Technologie de Troyes, France, Metropolitan |
|---|
| Topic: | 1.1 Modelling, Identification & Signal Processing |
|---|
| Session: | Image Processing and Biomedical Applications |
|---|
| Keywords: | computer tomography, parametric model, statistical tests, faultdetection, incomplete data, invariance, signal detection |
|---|
Abstract
The problem of detecting an anomaly from a limited number of noisy tomographic projections is addressed from the statistical point of view. An unknown two (or three)-dimensional scene is composed of a background, considered as a nuisance parameter, with a possibly hidden anomaly. A parametric approach is proposed to reduce the lack of a priori information and an optimal test is designed. To decide between two hypotheses (absence or presence of the anomaly), the statistical test eliminates the background, which can hide the anomaly. New results on anomaly detectability are proposed and discussed in this paper. It is shown that a size-limited anomaly is better detectable with several projections.
