search

carnisj / bcdi

BCDI: tools for pre(post)-processing Bragg coherent X-ray diffraction imaging data
Other
23 stars 17 forks source link

[FEAT] Resolution in workflow #265

Closed DSimonne closed 1 year ago

DSimonne commented 2 years ago

Hi,

I would like to try to implement at least one way to compute the spatial resolution with bcdi that is not the voxel size. It should also be clear what the voxel size is for us, I believe it is computed from the avg_obj shape at line 608.

So is it based on amplitude_threshold=0.05 ? Line 191. I don't think the phase unwrap or phase ramp removal changes the shape.

It would be nice to have a clear description of this to put in the docs.

Then what about automating the FDC ? Or the PRTF but is it rather a measure of the quality of the reconstruction than the resolution.

Does the good method exist ? What about deriving a method that uses the phase, which is much better determined than the amplitude.

mirichard-S commented 2 years ago

Hi,

Up to now, there is no good method. The PRTF gives the consistency between the reconstructed and measured intensities. If you take one solution, the PRTF is very close to one and it is difficult to estimate the resolution. What is the FDC? Is it the Fourier Shell correlation? It could be good the find a new method!

carnisj commented 2 years ago

The voxel sizes can be fixed manually or are calculated using the geometry in the method setup.voxel_sizes (not the shape of the reconstructed crystal). They are not related to the resolution. Currently you can estimate the resolution in the package using the PRTF or linecuts through the reconstructed edges, see #246

DSimonne commented 2 years ago

Yes it was a typo I meant FSC.

I saw the new method in postprocessing, it's nice ! It quantifies well what we see! I will add these values to the cxi file.

How can we well define the area where our SNR is good ? It has high impact as well. In terms of $\delta q_x$, $\delta q_y$, $\delta q_z$. That would actually be quite meaningful.

Of course the voxel size only means something if the shape of the array is also meaningful.

DSimonne commented 2 years ago

The resolution that you get from the linecut is acutally not saved in the h5 file, so I'll try to modify that

carnisj commented 2 years ago

Yes, and that was on purpose, because the cut is made along the basis axes, and not perpendicular to the facets, which would be more meaningful. What would be excellent is to calculate the direction of the linecut during facets recognition. Currently it is just an indication.

수요일, 6월 8, 2022, 4:33 오후에 David Simonne @.***>이(가) 다음과 같이 작성:

The resolution that you get from the linecut is acutally not saved in the h5 file, so I'll try to modify that

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were assigned.Message ID: @.***>

DSimonne commented 2 years ago

Yes it would be nice I agree ! But even just as an indication, it's nice to save =)