Closed vilim closed 4 years ago
This is related to #16 as voxel size depends on ROI dimensions
the json file has this field (optionally)
Inter-plane pacing came come in 2 ways: The dinstance between the centre of the beam waist of subsequent planes or the distance between then end of a waist and the start of the next.
Also to calculate voxel size we need to calibrate the microscope with some sample of know size (like a ruler). Next question is, in voxel size, which of the 2 z distances do we use.
it is the right-hand one. In usual digital imaging sampling models, you sample at coordinates, how much the point spread functions blur into one another it is irrelevant (you want to know e.g how far the cell centers are, if you model the cell as any distribution, e.g. a 3D Gaussian only the right-hand distance is informative). For the xy voxel size we know it's 0.3um at 1x1 binning, but maybe it's good to double-check this. For a very detailed discussion of pixels/sampling, have a look at this: http://alvyray.com/Memos/CG/Microsoft/6_pixel.pdf
voxel_size = (z,x,y) or (x,y,z)? What is the current convention in the lab if any?
Committed to master (as they were small changes). I decided to go for (z,x,y), if I was wrong it is a matter of changing one line of code
The implementation is not correct: The voxel size should be the distance between two adjacent voxel centers, which is not influenced by waist width at all. The order is z y x as is common for all datasets.
Imagine you are trying to register a lightsheet imaging dataset to a reference brain, or want to plot a scatterplot of all the cell centers: the thing you are interested is the spacing between imaged planes, which is independent of waist width/z blurriness.
Can we close this? Note: x and y were swapped and now dims are (z, y, x)
voxel size (binning dependent) and inter-plane spacing