[Rtk-users] Question about the geometry

[Chao Wu]

Well, that's just some guess.
I assume the central ray goes towards the center of your detector, so the
origin should be denoted as somewhere near (- half_width, - half_height) in
millimeter. Then I just fine-tune it manually a little bit to get a little
bit better results...

Regards, Chao


2015-10-26 19:44 GMT+01:00 Arvid Piehl Lauritsen Böttiger <
bottiger at gmail.com>:

> Dear Chao.
>
> I am very thankful for the help you provided me some days ago on the
> RTK mailing list. However, I have a quick followup question for you,
> which I have not been able to learn by reading the (very long) manual
> another user provided me with.
>
> I am now working on another dataset, and get similar artifact when
> reconstructing. Could you please help understand why you picked the
> parameters you did? I guess the "spacing" is pixel size, but why did
> why you pick "neworigin" to be -20,-20 ?
>
> best
>
> Arvid
>
>
> On Fri, Oct 16, 2015 at 2:03 PM, Chao Wu <wuchao04 at gmail.com> wrote:
> > Hi Arvid,
> >
> > Since you know how important a correct geometry is, you should understand
> > why you need projection origin and spacing.
> > CT geometry has more parameters than sid and sdd. How big the detector
> pixel
> > size (spacing) and how the detector is places w.r.t. the central ray
> > (origin) also play a role in the geometry. And in-plane/out-of-plane
> > rotations etc.
> > In your dataset the correct spacing and origin information is not stored
> > correctly in the projection data, so the total geometry is wrong.
> >
> > I have tested your data with the following command line and it gives
> > meaningful image. The newspacing and neworigin arguments force rtkfdk to
> > accept spacing and origin information of projections from command line
> > instead of from the data. Also I change the spacing of the volume from 2
> to
> > 0.2 because your object is small.
> >
> > rtkfdk -p c:\Work\out_dir\ -r cylinder-a.mha -o
> > c:\Work\out_dir\cylinder_a_recon.mha -g c:\Work\out_dir\geometry_real.xml
> > --spacing=0.2 --dimension=256  --newspacing=0.2 --neworigin=-20,-20
> >
> > Apparently there is still some misalignment between this guessed geometry
> > and the real geometry, which you need to figure out with your own
> hardware.
> >
> > Regards,
> > Chao
> >
> >
> > 2015-10-16 13:15 GMT+02:00 Arvid Piehl Lauritsen Böttiger
> > <bottiger at gmail.com>:
> >>
> >> Hi Chao.
> >>
> >> Thank you for the reply. However, I'm not quite sure what you mean by
> >> origin and spacing.
> >> A pixel on the detector is ~200 microns, and the sample is magnified
> >> around 1.3 times.
> >>
> >> I do understand that having a correct geometry is important to get an
> >> output image with correct dimensions, but I do not understand it's so
> >> difficult for me to make a single successful reconstruction (possibly
> >> with a dummy geometry).
> >>
> >> best
> >>
> >> Arvid
> >>
> >> PS: Just to validate the data I'm testing on I installed Octave and
> >> created a sinogram I then ran through its iradon implementation. It
> >> seemed to work quite nicely (see attachments).
> >>
> >> On Fri, Oct 16, 2015 at 12:09 PM, Chao Wu <wuchao04 at gmail.com> wrote:
> >> > Hi Arvid,
> >> >
> >> > Is the origin and spacing of your projection data correct? What is
> your
> >> > projection pixel size?
> >> >
> >> > Regards, Chao
> >> >
> >> > 2015-10-16 9:12 GMT+02:00 Arvid Piehl Lauritsen Böttiger
> >> > <bottiger at gmail.com>:
> >> >>
> >> >> Dear all. Hello again.
> >> >>
> >> >> I writing to you again because I need some help to understand how to
> >> >> declare the geometry of the system.
> >> >>
> >> >> Cyril redirected me towards this document (
> >> >> http://www.creatis.insa-lyon.fr/~srit/geometry.pdf ) which helped a
> >> >> lot, but unfortunately it has not been enough to make me reconstruct
> a
> >> >> tomogram.
> >> >>
> >> >> I have attached a drawing of my setup (setup.jpg). The
> >> >> source-to-detector distance is around 1600 mm. I just rotate the
> >> >> sample instead of the source, and the sample-detector distance is 15
> >> >> cm which makes the isocenter distance 1450. The field of view is
> >> >> around 2 cm.
> >> >>
> >> >> The aligned projections are located here:
> >> >> https://www.dropbox.com/s/ycg41qkeym61kec/cylinder-a.mha?dl=0
> >> >> and the raw tiffs are here:
> >> >> https://www.dropbox.com/s/hb67pkueyb3zcc3/tiffs.zip?dl=0
> >> >>
> >> >> I would say the input data looks as nice, but when reconstructing I'm
> >> >> still just getting noise:
> >> >>
> >> >> This is how I perform my "reconstruction"
> >> >>
> >> >> # 361 projections from 0 to 360 degrees. The source-detector distance
> >> >> is 1.6 m and the
> >> >> # sample-detector distance is 15 cm
> >> >> # I have tried with many variations of --ssd and --sid but I think
> >> >> they all look the same
> >> >> rtksimulatedgeometry.exe -f 0 -n 361 -a 360 --sdd 1600 --sid 1450 -o
> >> >> c:\Work\out_dir\geometry_real.xml
> >> >>
> >> >> # make a mha file out of the tiff files
> >> >> rtkprojections.exe --path c:\Work\tomography_data\tiffs\ --output
> >> >> c:\Work\out_dir\cylinder-a.mha --regexp .tif
> >> >>
> >> >> # "Reconstruct" the data.
> >> >> rtkfdk.exe -p c:\Work\out_dir\ -r cylinder-a.mha -o
> >> >> c:\Work\out_dir\cylinder_a_recon.mha -g
> >> >> c:\Work\out_dir\geometry_real.xml --spacing 2 --dimension 256
> >> >>
> >> >> The result can be viewed in the recon_result.PNG attachment.
> >> >> The reconstrcution can be found here:
> >> >> https://www.dropbox.com/s/cbwwnbbzvhqiyvy/cylinder_a_recon.mha?dl=0
> >> >>
> >> >> Can someone please help me what I am doing wrong?
> >> >>
> >> >> best
> >> >>
> >> >> Arvid
> >> >>
> >> >> _______________________________________________
> >> >> Rtk-users mailing list
> >> >> Rtk-users at public.kitware.com
> >> >> http://public.kitware.com/mailman/listinfo/rtk-users
> >> >>
> >> >
> >
> >
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/rtk-users/attachments/20151026/35110cfe/attachment-0010.html>