[Rtk-users] scanner development problems

[Vasiliy Nabokov]

Hi dear RTK users. Sorry, its offtop post... i just wanna tell you about my
problems if anyone can help me..
I develop scanner. Reconstruction of parallelepiped or cube (
http://i.imgur.com/9YHyfWH.jpg) caused to phantom artifact (
http://i.imgur.com/yyI184j.png). Don't look at noise, its test scan,
without averaging and big rotation step.  Yea, it seems that its mainly
misalignments problems, so i calibrated sourceX, sourceY, detectorX,
detectorY by the following method
http://www1.jinr.ru/Pepan_letters/panl_2015_5/10_gongad.pdf, got sourceX =
-215 mkm, sourceY = 650 mkm, detectorX = 50mkm. i specify this information
but without success.

I also wrote a little program in this way
    for (float sourceXOffset_ = sourceXFrom; sourceXOffset_ < sourceXTo;
sourceXOffset_ += searchStep)
    for (float sourceYOffset_ = sourceYFrom; sourceYOffset_ < sourceYTo;
sourceYOffset_ += searchStep)
    for (float projXOffset = projXFrom; projXOffset < projXTo; projXOffset
+= searchStep)
    {
        sprintf(cmd, "./rtksimulatedgeometry -o geometry.xml --n 90
--arc=360 --sdd=232.241 --sid=170 --source_x=%.6f --source_y=%.6f
--proj_iso_x=%.6f", sourceXOffset_, sourceYOffset_, projXOffset);

        int r = system(cmd);

        sprintf(cmd, "./rtkfdk --hardware=cuda -p /home/fee/Public/ -r
[0-9]+.tif$ -o dump/out.mha -g geometry.xml --verbose --dimension 800,1,800
--spacing 0.0296 --newspacing 0.0148 --neworigin -18.137,-12.128,0
--subsetsize=1 -l --origin -11.8252,-3.55,-11.8252");
        r = system(cmd);
...

to search scanner alignment but also without success.

I noticed interesting thing: when i specially offset detector horizontally
to the right by 4.5mm from the center, and specify it by --proj_iso_x=4.5
when generating geometry file, phantom disappears! (
http://i.imgur.com/OOsPsfL.png) I even take scans when the detector on the
center, reconstruct it, and here is phantom artifact, then i manually
offset picture to the left on the tifs (from http://i.imgur.com/FGxCJUo.png
to http://i.imgur.com/brbp5sd.png), and it reconstructs without phantom
with --proj_iso_x=4.5! So, for this moment i work in this way, take scans
when detector on the center, manually offset picture to the left and
specify it by proj_iso_x to avoid phantom artifact, but its HUGE WORKAROUND
=))


What do you think about it ? If its geometry issue, why real or virtual
offset of detector cleans up the phantom artifact?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/rtk-users/attachments/20160316/e8a6f795/attachment-0009.html>