BUG: change output type to float
Created by: fedorov
Cast of the output to the input type led to occasional overflows. Output of this module is inherently floating point, as it is a multiplication of the recovered bias field by the input image.
The issue discovered and reported by Kilian Pohl.