Python2k's csv module doesn't support Unicode input. This is described in the document. (cf. csv)
To read Unicode input, you should write wrapper and re-code the encoding. For instance, if you want to read a file encoded in cp932, you should do cp932 -> unicode -> utf8 -> csv -> utf8 -> unicode conversion. The csvwrapper library will do this conversion for you.
Although the csvwrapper library reads any file properly, the reading speed is much decreased. (10 times slower in my PC)
This module aims to port Python3k's csv module, which supports Unicode input, to Python2k.