Back to publications
Learned denoising with simulated and experimental low-dose CT data
Paper Details
Published: 2024/08/15
DOI: 10.48550/arXiv.2408.08115
ARXIV ID: 2408.08115v1
This paper analyses what happens when artificial noise is added to experimental data at the same level of real noise, and cross-compare AI methos trained in each. The authors showcase that even the most sophisticated of the “fast” X-ray noise simulators do not capture the complexity of real noise, and thus any model trained with such simulated noise may not generalize to real data.
Authors
M.B. Kiss
C.B. Schonlieb
K.J. Batenburg