Figure 6
Reconstructions for Tomobank dataset ID 86, nano-CT data with sparse projection angles using 202 projections, with (a) tomoCAM, (b) SVMBIR, (c) TIGRE (IRN-TV-CGLS) and (d) gridrec. Panels (e), (f), (g) and (h) are zoomed-in regions of interest represented by the boxes in (a), (b), (c) and (d), respectively. Panel (i) displays the line profiles for (d), (e), (f) and (h). A circular mask was applied to all the reconstructions. While tomoCAM and SVMBIR both do an excellent job at suppressing the noise when compared with gridrec, tomoCAM is approximately 15× faster. Also, tomoCAM is 22× faster than TIGRE (IRN-TV-CGLS). |