Dear Chao, Thanks for the question and sorry for the delay in answering it. The boolean allows to determine if the box goes from itk::Image::Origin to the coordinate of the last pixel or whether you want to add a half voxel border, see code here <https://github.com/SimonRit/RTK/blob/master/src/rtkBoxShape.cxx#L147-L168>. In SART, you want to divide the difference between the projection and the measure by the projection of a volume filled with one. Our projectors consider that the intensity is 0 before Origin and after the coordinate of the last pixel so I think the current version is the correct one. The rtkforwardprojectiontest.cxx <https://github.com/SimonRit/RTK/blob/master/test/rtkforwardprojectiontest.cxx> tests this and you can check in it that there is indeed no half pixel border. So I believe this is the right thing to do, yes. Best regards, Simon
On Mon, Aug 19, 2019 at 10:15 PM Chao Wu <wucha...@gmail.com> wrote: > L.S., > > I have noticed a behaviour change in SART after commit 5dc1d78. > In SARTConeBeamReconstructionFilter, > m_RayBoxFilter->SetBoxFromImage(this->GetInput(0), > false) attempts to pass false to bWithExternalHalfPixelBorder, although > before this commit this variable was not used in > RayBoxIntersectionImageFilter::SetBoxFromImage and then when > rtk::BoxShape::SetBoxFromImage is called the default half pixel border flag > was set to true. After this commit a false can finally be passed through. > This gives slight different reconstruction results than before. I was just > wondering which one is more correct, with or without the half pixel border > for the volume used for the normalization. I tried to follow the external > link of explanation given in the code of SetBoxFromImage but found it > broken. Could anyone give a brief explanation? Thanks. > > Best regards, > Chao > _______________________________________________ > Rtk-users mailing list > Rtk-users@public.kitware.com > https://public.kitware.com/mailman/listinfo/rtk-users >
_______________________________________________ Rtk-users mailing list Rtk-users@public.kitware.com https://public.kitware.com/mailman/listinfo/rtk-users