[vtk-developers] Re: vtkProcrustesAlignmentFilter

[Rasmus Reinhold Paulsen]

On Wed, 16 Jan 2002, Will Schroeder wrote:

> I want to make sure my understanding is correct: you want to process a 
> group of data sets as a single data set?

Yes. Tim and I plan to implement the Procrustes alignment which basically
is an algorithm that takes as input a set of shapes with corresponding
vertices (for example anatomical landmarks on biological objects as faces
etc). The algorithm then calculate an average shape in an iterative
fashion
(align all shapes to the current average shape estimate, calculate new
estimate
of average shape, repeat until covergence). The output is the set of
shapes
aligned with the estimated average shape and offcourse the estimated
average shape.

Later these aligned shapes can be analysed by a principal component
analysis to give an impression of the major modes of shape
variation. This could also be implemented as a multi-shape in/multi-shape
out class.

So it would be convenient to pipe around a collection of "similar" shapes.

Regards
Rasmus


BTW. I am now on the developer list, so you do not need to CC me anymore.