Table 1. Average signal-to-noise ratio (SNR) after attenuation of streaks in the Shepp–Logan phantom subject to mixed streak and Poissonian noise as in (14), with different combinations of {{\rm{std}}(\eta_{{}_{{\rm{P}}}})} and peak values of A, with {\rm{peak}} = \infty being the limiting case for which {\pi\! = \!0}.
As all of the algorithms aim to remove streak noise only, the SNR values are calculated with Y = \ln[A+\pi/(1\!+\!\eta_{{}_{{\rm{P}}}})] as {{\rm SNR}(\hat{Y}) = 10\,{\rm{log}}_{10}({\rm{svar}}_{X}\{Y^{2}\}/{\rm{smean}}_{X}((\hat{Y}\!-\!Y)^{2})), where svar and smean denote sample variance and sample mean, respectively. Each value of the table is the average SNR over ten different noise realizations.
| SNR | |||||
|---|---|---|---|---|---|
| Peak | std(ηP) | Noisy | Münch et al. (2009 ▸) ▸ | Vo et al. (2018 ▸) ▸ | Proposed |
| ∞ (π = 0) | 0.005 | 32.61 | 11.80 | 28.97 | 44.05 |
| 0.01 | 26.59 | 11.78 | 28.52 | 39.19 | |
| 0.02 | 20.58 | 11.72 | 27.48 | 34.29 | |
| 0.05 | 12.77 | 11.49 | 24.62 | 27.24 | |
| 2560 | 0.005 | 32.66 | 11.85 | 28.32 | 38.41 |
| 0.01 | 26.64 | 11.82 | 27.89 | 35.90 | |
| 0.02 | 20.63 | 11.77 | 26.90 | 32.63 | |
| 0.05 | 12.82 | 11.54 | 24.21 | 26.67 | |
| 1280 | 0.005 | 32.71 | 11.90 | 27.76 | 36.51 |
| 0.01 | 26.69 | 11.87 | 27.36 | 34.31 | |
| 0.02 | 20.68 | 11.82 | 26.45 | 31.55 | |
| 0.05 | 12.86 | 11.59 | 23.92 | 26.21 | |