proj-imim:
Re:
John Blommers 2001-Jan-05 20:09:56
Richard,
Haw! Your are right. Each control point represents two variables to be
optimized. You'd expect to need at least 14 control points for 28 variables.
Curiously, when I played with this, I could reduce the number of control
points in a test script well below this number and the optimiser still
chugged along happily, not complaining about the foolishness of it all.
Needless to say the resulting optimization did not result in a good
stitch. The rules of thumb about extra control points is still true
if you want a good stitch. Still, setting up control points is a royal
pain since it takes up time, and if we can get good stitching results
with less points, this is good news. My gut feel is that you need at
least four control points per image pair for good optimization.
Additional feedback on this topic is welcomed :-)
-John