Perform processing operations in calibrated units where possible

[cmeyer]

Processing operations requiring multiple inputs should have sensible result intensity units:

• add, subtract should use a common intensity unit
• multiply, divide should use a common unit to both
• cross correlate?

Processing operations requiring multiple inputs should also use common spatial reference frames:

• disallow add, subtract, cross correlate if spatial reference frame does not match
• require one input to be uncalibrated for multiply, divide
• add explicit commands to shift and output the overlapped portions for n inputs which can subsequently be used in processing

For instance, calibrated subtraction...

Emails from 2019-07-02, AM, TL.

import numpy
from scipy.ndimage import shift
import uuid
from nion.data import xdata_1_0 as xd
shift_amount = (src2.xdata.dimensional_calibrations[0].offset - src1.xdata.dimensional_calibrations[0].offset) / src1.xdata.dimensional_calibrations[0].scale
shifted_data2 = shift(src2.data, shift_amount, order=1)
shifted_xdata2 = api.create_data_and_metadata(shifted_data2, intensity_calibration=src1.xdata.intensity_calibration, dimensional_calibrations=src1.xdata.dimensional_calibrations)
target.xdata =  shifted_xdata2 - src1.xdata

[cmeyer]

See also https://github.com/nion-software/niondata/blob/a0e1ee7580df5e6615bc76f5740a9be88b9d3cff/nion/data/Core.py#L2038-L2057

[cmeyer]

ND 2021-05-07:

if I add two images, the result will not be the sum of the images if either of the images has in intensity calibration.

So if I create a shifts image from an SI, any operation done to it will divide the data by the dispersion in user terms.

[cmeyer]

Also see ND 2021-06-14 email.