Computational techniques for fitting a 3-D rotation to 3-D data are re
capitulated in a refined form as minimization over proper rotations, e
xtending three existing methods-the method of singular value decomposi
tion, the method of polar decomposition, and the method of quaternion
representation. Then, we describe the problem of 3-D motion estimation
in this new light. Finally, we define the covariance matrix of a rota
tion and analyze the statistical behavior of errors in 3-D rotation fi
tting.