leads to numerical instability and poor solutions - oh goody, we're in the land of 'here be dragons' - the insight I'm trying to play now is that if you take all the images and rip them in half and duplicate them, then you can come up with optimal solutions per camera per half - the problem is each half field view is perceived by two cameras and they don't agree (better to say they may violently disagree). However, it is an affine transform at the end, and I know that the image breaks are perfect, i.e. if I map the transforms they should refuse - the question is can I drop the depth information out that way.........stay tuned..... wish me luck if you would.