Since the change of direction on my OSS I have narrowed my project to working on just one problem, video mosaics. I mentioned this problem in my last post (it was more hand wavy then). However since then I have gained a new knowledge of the vocabulary used in the field and it turns out that the problem is most frequently referred to as video mosaics.

The first truly useful paper I found on the subject was written by Szeliski who happens to work at Microsoft research. He also seems to be the world’s expert in the field as almost every paper that I have read about the subject includes atleast one of his papers.

His method to finding out how images should overlay on top of each other involves iteratively reducing the error of intensity differences between two frames by using the Levenberg-Marquardt algorithm.

It turns out that this research involves multi-variable calculus and linear algebra which I have been slowly picking back up. I don’t quite understand projective transformations yet but I hopefully will in the near future.

In order to make some immediate progress I downloaded the solver today and am beginning to piece together parts of Szeliski’s algorithm in Eclipse.