|
Figure 9
Comparison of inference time between our method and classical method followed by post-processing. Our method involves reconstructing the raw projection, the processed projection, and the processed sinogram, requiring three passes through CNNs at each stage. In contrast, post-processing involves pre-processing the raw projection, reconstructing and applying it to the image domain CNN, exemplified here with a wavelet-based technique for clarity. Notably, when considering methods by Vo et al. (2018 ![]() ![]() |