Analysis

Every function works in physical units and records itself in the dataset's history (operation, parameters, input and output content hashes, software versions, person, time).

from nuscope import analyze as an

Function	Returns	Notes
stats(ds)	dict	mean, std, min, max, 1st/99th percentiles, noise σ (wavelet estimate), SNR
crop(ds, y=(y0, y1), x=(x0, x1))	Dataset
bin_image(ds, factor=2)	Dataset	pixel size scales with the factor
calibrate(ds, pixel_size_nm, reason="")	Dataset	the reason (e.g. a calibration standard) is kept in the history
denoise(ds, method="nlm", strength=1.0)	Dataset	nlm (non-local means), gaussian, median
subtract_background(ds, sigma_nm=None, sigma_px=50)	Dataset	removes slow shading
fft(ds, window=True, log=True)	Dataset	power spectrum with axes in 1/nm
radial_profile(ds, bins=200)	DataFrame	azimuthal average: q (1/nm), intensity, pixels, period (nm)
line_profile(ds, start, end, width=1)	DataFrame	distance in nm along the line
threshold(ds, method="otsu", value=None, dark=False)	Dataset (mask)
particles(ds, method="otsu", min_area_nm2=0, exclude_border=True, separate=False)	DataFrame	area, equivalent diameter, Feret, aspect ratio, circularity, orientation, mean intensity, in nm

table = an.particles(an.denoise(ds), min_area_nm2=500, separate=True)
table.attrs["area_fraction"]        # fraction of the field covered
table["eq_diameter_nm"].median()

Your own functions can join the history with the tracked decorator:

from nuscope.analyze.core import tracked

@tracked("my_lab.flatten")
def flatten(ds, order=2):
    ...
    return ds.derive(new_array)