Switch to using uints rather than float for internal image representation #28
Description
Most of our algorithms rely on using 32-bit floats for representing grayscale images, which a poor choice given that we only allow values between 0 and 1. We should switch to using uints instead.