Could the transform functions handle non-PIL data?

my project is to handle some 2 channel pictures(Real and imaginary),I read the docs ,all of them says the functions processing PIL image(RGB?), and my data is even not picture format,i transform them into numpy ndarray,so are these functions working?

Depending on the data format you could maybe split the channels and apply the transformations on them separately.
What rand do your values have?