Hi Simon,
I am afraid I was no clear enough. Please find a picture of the real
object and the reko at that link:
https://www.dropbox.com/sh/ul0oy9kv3us4ey7/AABQ5Y4R1PR-jcRawGFKOUK4a?dl=0
So you can see that on the part, the serial number is on the "head" side
while it is on the "tail" side on the reconstruction, using the "direct"
geometry. That is what I call the mirror image. The rotation axis is
along the vertical direction of the image. I could easily reorder the
reconstructed slice to get it in the right orientation, but I was
wondering where the issue comes from.
Best regards,
Vincent
On 28.08.20 12:13, Simon Rit wrote:
Mirror in which direction? Depending on the direction, it can also be
a 180° offset of the angle. If it reconstructs well, I would assume
that the direct direction is the correct one but there is something
else you need to understand...
On Fri, Aug 28, 2020 at 10:44 AM Vincent Libertiaux <[email protected]
<mailto:[email protected]>> wrote:
Hi Simon,
thank you for testing my dataset.
I get the same results you describe and I am quite happy with the
first result. However, the reconstructed volume is a "mirror" view
of the real object, and my guess was that the rotating plate was
going in the opposite direction assumed by rtk. Is it the wrong
assumption?
Thank you again for your help,
best regards,
Vincent
Thanks for the dataset. When I run
rtkfdk -p . -r ^proj.mha$ -g direct.xml --spacing 0.5 -d 300
--hardware cuda -o fdk.mha
The result looks good to me. Obviously, when I run
rtkfdk -p . -r ^proj.mha$ -g inverse.xml --spacing 0.5 -d 300
--hardware cuda -o fdk.mha
the result is bad since the correct rotation direction seems to be
the direct one. Did you expect the second line to produce the
correct result? Or is the first line not producing a good enough
result in your opinion?
_______________________________________________
Rtk-users mailing list
[email protected]
https://public.kitware.com/mailman/listinfo/rtk-users
