view article

Figure 4
The effect of varying SNR on algorithms for finding the center of rotation (COR) for the tomo_00064 data set. In (a), we show the COR along the [\hat{t}]-axis as determined by three methods (phase symmetry, phase correlation, and sinogram FFT) for real data with simulated noise. Center finding was repeated for 100 instances of simulated Poisson noise for each value of [\bar{n}] and resulting SNR as summarized in Table 1[link]. From these 100 instances, the mean value of the center of rotation, as well as the standard deviation of the results, are shown for each SNR value. In (b), the power spectral density for the reflection image pair is shown for a smaller subset of SNR values, with all curves normalized to the same power at zero spatial frequency; as expected from Fig. 3[link](d), the lowest spatial frequencies are least affected by noise. The sinograms versus SNR are shown in (c). In (d), a reconstruction of the sinogram is done for the COR as determined by the phase symmetry method (left), and also for a 2 pixel offset in the COR (right); as can be seen, the image on the left shows no ring or tuning-fork artifacts, while the image on the right shows ring artifacts as expected for this 360° data set.

Journal logoJOURNAL OF
ISSN: 1600-5775
Follow J. Synchrotron Rad.
Sign up for e-alerts
Follow J. Synchrotron Rad. on Twitter
Follow us on facebook
Sign up for RSS feeds