[Insight-users] ICP complexity

[Luis Ibanez]

Hi Felix,

The computation time should be in the order

a) O(N*M) : For the section that search for nearest point

                and

b) O(N) for the section that computes the actual metric value



Where

N = number of points in the Fixed Point Set
M = number of points in the Moving Point Set


BTW: please not that ICP is not expected to be
      a symmetrical problem. Therefore, although
      selecting the set of points with the least
      number of points as the Fixed Point Set
      may give you faster computation, it may
      not necessarily give you the best registration.


    Regards,


       Luis


----------------------
Felix Bollenbeck wrote:
> Dear ITK User,
> 
> I am currently working o a problem, involving a 3D correspondence 
> problem identical to the ICP setting.
> 
> For further directions I have question regarding the complexity of the 
> (ITK) ICP algorithm:
> 
> How does the computation time relate to the number of points roughly?
> 
> Regards,
> 
> Felix Bollenbeck.
>