Tomographic reconstructions using map algorithms - application to the SPIDR mission
The spectral image of an astronomical scene is reconstructed from noisy tomographic projections using maximum a posteriori (MAP) and filtered backprojection (FBP) algorithms. Both maximum entropy (ME) and Gibbs prior are used in the MAP reconstructions. The scene, which is a uniform background with a localized emissive source superimposed on it, is reconstructed for a broad range of source counts. The algorithms are compared regarding their ability to detect the source in the background. Detectability is defined in terms of a contrast-to-noise ratio (CNR) which is a Monte Carlo ensemble average of spatially averaged CNRs for the individual reconstructions. Overall, MAP was found to yield improved CNR relative to FBP. Moreover, as a function of the total source counts, the CNR varies distinctly different for source and background regions. This may be important in separating a weak source from the background.
- Research Organization:
- Lawrence Berkeley National Lab. (LBNL), Berkeley, CA (United States)
- Sponsoring Organization:
- USDOE Director. Office of Science. Office of Biological and Environmental Research (US)
- DOE Contract Number:
- AC03-76SF00098
- OSTI ID:
- 838183
- Report Number(s):
- LBNL-54934; R&D Project: 445401; TRN: US200507%%265
- Resource Relation:
- Other Information: PBD: 21 Jan 2004
- Country of Publication:
- United States
- Language:
- English
Similar Records
Noise properties of filtered-backprojection and ML-EM reconstructed emission tomographic images
Comparing lesion detection performance for PET image reconstruction algorithms: A case study