These modifications allow valid URL strings to be passed directly into imread.
URL recognition is done using the standard library function urlparse. If the
string contains a valid url scheme, it will be used to download and process
the data using a combination of urlopen and BytesIO (StringIO in python 2).

The Pillow docs state that Image.open can handle either a filename string or a
file object. That simplifies the pilread function substantially.

It seems that the identity test for the parsed URL scheme was not sufficient.
I changed this to an equality test, and things are now working properly.