|
Chris,
The data is linear so long as it is linearly proportional to the original signal. That relationship is destroyed when you perform a non-linear stretch on the data, typically with HistogramTransfer or MaskedStretch.
Deconvolution and ColorCalibration should always be done with linear data. Some processes can be done on linear or non-linear data (e.g. noise reduction with MLT/MMT/ATWT/TGVDenoise, albeit carefully if the data is linear) and a lot of other processes work best on non-linear data, e.g. HDRMT and LHE.
Cheers,
Rick.
|