[Camino-users] Eigensystem issue monotensor fitting

[Alessandro Calamuneri]

Hi camino experts,

I got into troubles when analyzing eigensystem coming from a monotensor
fitting (with semidefinite positiveness constraints)..

When creating rotation matrices associated to eigendecomposition, I can see
that a lot of times I do not have a proper rotation, ie, determinant =-1.
I tried both by considering eigensystem coming from dteig command, either
by determining it with in-house scripts built in matlab based on dtfitting
output.

Possible explenation are:

1) it should work that way because, for instance, you reorder rotation
matrix columns to get corresponding eigenvalues sorted from biggest to
smallest;
2) it should work that way for whatever any other reason;
3) Problem is caused by bad fitting due to noise in my data.
4) May problem be related to some error in output bit representation? (I am
using voxel2image specifying number of components to get the data); however
I do not think this be the case..

I am asking because I am extracting quaternion representation out from
rotation matrices., and a correct rotation data is crucial..

Best Regards,

Alessandro
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.nitrc.org/pipermail/camino-users/attachments/20151014/a7e0c9db/attachment.html>