CebraEM: Quantile normalization terribly inefficient when mask has objects with large bounding boxes

[jhennies]

CebraEM: Quantile normalization terribly inefficient when mask has objects with large bounding boxes

[josemiserra]

What about downsampling using NN instead of interpolation? José Miguel Serra LletiOn 30 Mar 2023, at 15:36, jhennies ***@***.***> wrote: It might be quicker to just retrieve n = 1024(?) pixels randomly within each mask object. The question being: does that yield the same normalization result? Using down-sampled layers of the EM dataset would not yield the same result as down-sampling is usually performed with mean interpolation such that the pixel values will have a different grayscale distribution. —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[jhennies]

The issue is that I would like to use the image pyramid already present in the big data viewer files that the user supplies. And there I can't enforce NN interpolation

[jhennies]

Doing the actual down-sampling on these large datasets is quite slow as well, so I would like to avoid computing that within the CebraEM workflow

[jhennies]

solved with #24