Current code sets auto_sigma_p to be 1.0 * auto_sigma_prior(sino, delta_channel, sharpness) (1/5 of auto_sigma_x).
In the cone beam code we found out that 2.0 * auto_sigma_prior(sino, delta_channel, sharpness) is generally (in terms of MACE reconstruction quality of two datasets) better.
Current code sets auto_sigma_p to be
1.0 * auto_sigma_prior(sino, delta_channel, sharpness)(1/5 of auto_sigma_x). In the cone beam code we found out that2.0 * auto_sigma_prior(sino, delta_channel, sharpness)is generally (in terms of MACE reconstruction quality of two datasets) better.