Neural net (machine learning) is a bit different in this regard. Its behavior depends on the training used to make it. So the difference in behavior for before and after stretching is not something you can predict and say as you do for example there is more of a loss of detail.
I have not found the NXT, when used appropriately, looses any significant detail either in linear images or non-linear. If you are concerned, however, you can easily make luminance masks to protect any detail that is above a given threshold of brightness.
Comments