Hi Sjors,
Thank you for helpful reply.
I do always look at tracks and in our case B they seemed to be quite random before the polishing and still rather random with Dist <= 200 polishing (more random than in our case A). They look much better (i.e. correlated within an area) with Dist = 400, but that is of course inevitable due to algorithm, so hard to say if it is closer to reality - final resolution is similar in either case.
Could it be additional problem that movie processing is done on images which are heavily downsampled at the end of autorefine? (Because we have particles in the big box with a lot of background, the resolution is very heavily underestimated in autorefine run). It seems like that, because movements before polishing often go in straight lines (i.e. along identical coordinate with non-integral pixel value). In that case any residual movements after motioncorr are may be too small to be captured with big pixel size on downsampled images (and may be even exaggerated due to big pixel size)? Or is movie processing done on original images?
Thanks for any info!
|