On 23 Apr., 12:02, Wox <nom...@[EMAIL PROTECTED]
> wrote:
> On Wed, 23 Apr 2008 00:42:07 -0700 (PDT), VP <peter.vonto...@[EMAIL PROTECTED]
>
> wrote:
>
> >Hi
>
> >please compare your sinogram with the following:
>
> >estsinogram=radon(tomogram,rho=rho,theta=theta,ntheta=nangles)
>
> >compare the rho and theta values and read the IDL radon help pages.
>
> >HTH
>
> >Peter
>
> So basically what you're saying is: don't undersample.
>
> However, now there are more zero's in the sinogram. Let me explain
> what I want to do. The algorithm I was talking about (OSEM, although I
> think it's really called MLEM, I'm not sure) goes like this
> (BP=backprojection, FP=forward projection):
>
> ========Pseudo-code========
> sino1=sinogram with all 1's
> tomo1=BP(sino1)
>
> tomo=tomogram with all 1's
> for i=0,niter-1 do begin
> estsino=FP(tomo)
> tomo = tomo * BP(sino_measured/estsino)/tomo1
> endfor
> ========================
>
> You see that "estsino" and "tomo1" can't have zeroed pixels. Off
> course, I tried replacing the zeroed pixels by 1, max(rest),
> min(rest), etc... But this gives some artifacts in the resulting
> tomogram. Any ideas?
Hi
I cannot comment your effort to implement an iterative reconstruction
algorithm. Simply comparing the two sinograms, i claim that your
estsinogram is not the full sinogram of a square. You first have to
make sure to start with the sinogram of a square !
HTH
Peter
|
