Is it normal for `branca.utilities.write_png` function to normalize its values?

I may be wrong, but I find it weird the way branca.utilities.write_png function normalize its values.
See https://github.com/python-visualization/branca/blob/v0.8.1/branca/utilities.py#L324.

>>> import io
>>>
>>> import numpy as np
>>> import PIL.Image
>>>
>>> import branca.utilities
>>>
>>> data = np.array([[[0.74901961, 0.50196078, 1.0, 1.0], [0.25098039, 1.0, 0.74901961, 1.0]]])
>>>
>>> # Current result
>>> png_bytes = branca.utilities.write_png(data)
>>> with PIL.Image.open(io.BytesIO(png_bytes)) as image:
...     np.asarray(image)
array([[[255, 127, 255, 255],
        [ 85, 255, 191, 255]]], dtype=uint8)
>>>
>>> # What I expected
>>> np.rint(data * 255.0).astype(np.uint8)
array([[[191, 128, 255, 255],
        [ 64, 255, 191, 255]]], dtype=uint8)

As a result, folium layers like folium.raster_layers.ImageOverlay appear brighter than I expected when passing image as an array because of the normalization.
I guess this allows users to automatically display mono rasters with a scale from minimum value as black to maximum value as white. However, for non-mono and mono rasters with a colormap, the normalization seems undesirable.

Start in branca/utilities.py at write_png, using the linked line and the supplied NumPy/PIL example to reproduce the current output. Done means deciding and verifying the intended normalization behavior for mono, non-mono, and colormapped rasters, including the reported brightness effect in folium ImageOverlay layers.