I am wondering - thinking about step 2 & step 4 - for later frames, the higher resolution information will have been lost due to radiation damage. If these are compared after just filtering both according to the FSC of the reference, presumably this will mean that fits for later frames are fitting a lot of noise.
Are the initial dose weights from Patch Motion used in steps 1&2 to filter the particle images and reference projections for comparison in step 4? Or is this not necessary?
I am wondering - thinking about step 2 & step 4 - for later frames, the higher resolution information will have been lost due to radiation damage. If these are compared after just filtering both according to the FSC of the reference, presumably this will mean that fits for later frames are fitting a lot of noise.
You’re correct; we filter according to the FSC curve (technically we only filter the reference) and don’t do any filtering as a function of frame to the patches during alignment. We did some experiments which suggested that downweighting the high frequencies could cause worse results, but it is on our roadmap to revisit this. Thanks for bringing it up. I agree that in principle, some sort of suppression of the high frequency noise in the later frames might be beneficial.
