Apodization implementation

[harshamarupudi56]

I was using the Ex_2D_Edge.m function to compute the MTF for a slanted edge phantom. The edge identification and segmentation work well, but I can see the edge spread function often comes out very noisy. I see there in the LSF calculation there is zeroing. I think there can be a function for apodization as this was useful for me. Potentially implementation of a Hamming/Hanning window. I assume this can be an issue others run into.

[pengweiwu]

Hi Harsha, thanks for your interest in our toolbox!!
An apodization window can definitely be implemented (one just need to remove the effect in the end by dividing the corresponding MTF). And I agree that LSF calculation is sensitive to noise considering the gradient operation.
By ESF being noisy, do you mean before or after the binning process? If it is noisy before the binning, it should be fine in my opinion (as the binning removes a lot of noise, especially with a small over sampling factor).

[harshamarupudi56]

Are you referring to the binning in the Ex_2d_Edge_BinningMatch.m script? If that is the case I don't think I am doing any binning my ESF calculation as I just used the Ex_2d_Edge.m script. I applied the apodization window myself however and didn't see any issues in the MTF.

[pengweiwu]

Got it. I meant that if you are using Ex_2d_Edge.m (or almost any script within this toolbox), internally it calls discretizeCurveMaxFreq, which does binning to the oversampled voxel size (details are in our paper).
Could you try reduce maxFreq in Ex_2d_Edge.m? Maybe reduce from 6 to 1.5, so it only samples to 1.5x Nyquist frequency (therefore more binning).

[harshamarupudi56]

Ah I see what you are referring to. I can try playing around with that as well.