Hi All,
I'm trying to implement an alternative for the FBP(filtered
backprojection) method for reconstructing objects measured in
tomography experiments. It's supposed to give less noisy tomograms.
Anyway, it's called OSEM and it's some iterative procedure using
forward and backward projection until the real sinogram and the
calculated sinogram are close.
For the projection, I use IDL's radon function. But I noticed
something strange with the forward projection. Try the code below. It
calculates the sinogram of a tomogram which is an image with all
pixels equal to 1. If you look at the result, something strange is
going on in the corners of the sinogram image. Does anyone know what
causes it and whether it is an intrinsic radon transform problem?
I would like to get rid of it, because this "estsinogram" is
calculated in each iteration of the OSEM (only in the first iteration
on an image with 1's) and used to normalize the measured sinogram
before adapting the tomogram. The resulting tomogram has some
artifacts because of it.
Thanks in advance,
Wout
pro test
; Detector
N=80
projcen=(N-1)/2.
; Angles
anglestart=0.
anglerange=180.
NAngles=anglerange/2.
angleinc=anglerange/(NAngles-1)
angles=anglestart+angleinc*findgen(NAngles)
angles*=!pi/180
; Reconstructing an object with 1's
tomogram=replicate(1.,N,N)
estsinogram=radon(tomogram,theta=angles,xmin=-projcen,RMIN=-projcen,drho=1,NRHO=N,/LINEAR)
loadct,0
window
tvscl,not bytscl(rebin(estsinogram,NAngles*3,N*3,/sample))
end
|
