I'm writing a panorama stitcher in C++ and I use slerp (spherical linear interpolation) and Quaternions to estimate the rotation matrix from adjacent images (using the Eigen library) when I can't get enough features to match. I can make a qualified guess that an image should be between the two others in a multi-row when let's say only one image can't be connected to others. It works kind of ok to this but I would like to improve it by using slerp (or something else) to do the estimate from more adjacent images.
Anyone that knows such a way? it doesn't need to be using the Eigen lib.
Currently, I do like this
But would like to improve with doing something like this using 1,3 and 5 (and even 4 and 5 if that is better)
Edit 1: From the comment by @nico-schertler I got an idea how would I go about to get the slerp fraction for 5 between 4 and 6? I would then use this fraction as my input to the slerp for 2 between 1 and 3 which currently is 0.5