This forum is closed. Thank you for your contributions.
Is any information available about the algorithm that ICE uses to match images?
Pointers to code, white papers, patents, etc?
There is no source available.
ICE consists of two main parts. Alignment and compositing.
The alignment is done based on the "image features" described here: http://research.microsoft.com/apps/pubs/default.aspx?id=79807
The compositing is done via the techniques in this paper: http://grail.cs.washington.edu/projects/photomontage/