are the best....? Hello. I'm starting out on Image Processing. My first task is to classify binarized images of digits 0-9: where these are scanned images of those digits that are machine-printed, written by a single person, and written by multiple persons, and the pixels of the images are only White (RB(0,0,0)) and Black (RGB(255,255,255)). I should be classifying the digits based on the count of number of black pixels of the features selected.

Now,
What 2 features do you think are the best to use to identify such? I've tried middle vertical and middle horizontal of the images, but they are not that accurate. Thanks,

BTW, these are the images.

http://www.drivehq.com/file/df.aspx/publish/jiverlalica/data_for_exer1
Oh, in the images the numbers appear twice. We only consider the left part. :-)