search

arcadelab / deepdrr

Code for "DeepDRR: A Catalyst for Machine Learning in Fluoroscopy-guided Procedures". https://arxiv.org/abs/1803.08606
GNU General Public License v3.0
208 stars 60 forks source link

Questions regarding the article #43

Closed tomeramit closed 3 years ago

tomeramit commented 3 years ago

Hi guys,

I was wondering how you combine those 3 parts (attenuation, scatter, noise) can you elaborate on that? also when playing with the projection I saw that the output xray is somehow normalized between 0 and 1, can you elaborate on that as well?

btw very nice article and project! Thanks

mathiasunberath commented 3 years ago

Photon signal and scatter are additive and noise is computed based on that signal (both Poisson noise and correlated noise to simulate detector readout). You can find those in https://github.com/arcadelab/deepdrr/blob/main/deepdrr/projector/projector.py#L577 and https://github.com/arcadelab/deepdrr/blob/main/deepdrr/projector/projector.py#L591.

The signal that we simulate may be within 0,1 but there is no guarantee. We do not specifically simulate I(x)/I_0, so you cannot assume [0,1].

tomeramit commented 3 years ago

Thanks!