view article
Figure 4
Denoising of the 3-D BrainWeb phantom with noise as in (3)[link] with [{\rm{std}}\{\eta_{{}_{{\rm{P}}}}\}] = 0.02 and signal peak 2560, displaying a single sinogram extracted from the stack. Left: a sinogram of ln(A) (3)[link] and the corresponding streak-free but noisy sinogram of Y (5)[link]. Right: on top, noisy sinogram Z (4)[link] and the comparison of estimates: [\hat{Y}] for Münch et al. (2009)BB27, Vo et al. (2018)BB35, Mäkinen et al. (2021)BB23, the proposed framework using old BM4D and using improved BM4D, and [\widehat{\ln(A)}] with the proposed framework and new BM4D. Below, corresponding estimation errors. Note that the errors [|\hat{Y}-Y|] show only the effectiveness of streak removal, as the compared algorithms are designed for streak attenuation, whereas [|\widehat{\ln(A)}-\ln(A)|] shows the complete denoising error, scaled separately. Notably, the proposed algorithm offers superior performance, but only when embedded with the improved BM4D.

Journal logoJOURNAL OF
SYNCHROTRON
RADIATION
ISSN: 1600-5775
Volume 29| Part 3| May 2022| Pages 829-842
https://doi.org/10.1107/S1600577522002739
Follow J. Synchrotron Rad.
Sign up for e-alerts
E-alerts
Follow J. Synchrotron Rad. on Twitter
Twitter
Follow us on facebook
Facebook
Sign up for RSS feeds
RSS