Open RingoIngo opened 4 years ago
Have you tried running with the most recent ODL (here at github)? The backend (ASTRA?) has changed scalings lately which could cause this.
Yes the backend I used was astra_cpu
and this scaling does not happen with skimage
. Thanks for the hint. I tried using pip install https://github.com/odlgroup/odl/archive/master.zip
but still ran into #1498.
Ok that's a bit weird. I'm sadly not the one who solved that issue the last time but perhaps @kohr-h has some input?
@RingoIngo I just tested installation via zip file, and it worked fine. Could you please give some details about your system, as well as the exact error message? As for the original issue, which version of the ASTRA toolbox did you use?
What's the status with this issue?
I am using odl 0.7.0 and ran the filtered back projection example for the parallel 2D geometry. Somehow the output of the fbp is not scaled correctly (see screenshot).![Bildschirmfoto 2019-08-01 um 15 24 44](https://user-images.githubusercontent.com/10393893/62297043-b656f600-b470-11e9-9aa9-4cbb2114e59f.png)
The phantom (not visible on screenshot) takes values in
[0, 1]
while the fbp reconstruction takes values in[0, 0.02]
(see colorbar in screenshot) the resulting error is of course large (second plot on screenshot). I think this is connected to the choice ofmin_pt
andmax_pt
forreco_space
which also influence the scale of the sinogram. Maybe I don't understand the use ofmin_pt
andmax_pt
but I think that its choice should not influence the scale of the sinogram (onlyshape
should)? This behaviour was the reason I tried the example in the first place.