Is your feature request related to a problem? Please describe.
Currently, we have original_shape, original_affine, etc. in the meta_dict, I suggest to also store the max, mean, std, etc. of the intensity of input image in the meta dict, then the following transforms can dynamically use it as parameters to scale/shift/clip values instead of fixed value for all the images.
For example:

[
    LoadImaged("image")
    EnsureChannelFirstd("image"),
    IntensityStatsd("image", key_prefix="orig", ops=["max", "std"]),
    RandScaleIntensityd("image"),
    RandShiftIntensityd("image", offset=0.1, src="orig_std"),
    ThresholdIntensityd("image", threshold="orig_max", above=True, cval="orig_max"),
]