Re: I need to locate pore clusters
- From: "Harris" <xgeorgio@xxxxxx>
- Date: 23 Feb 2006 15:16:06 -0800
In the general case, you need a K-means algorithm with a fixed
threshold T for "closure" and arbitrary number of final clusters. In
some software packages, these routines are included under the name
"cluster analysis" or "linkage". In any case, if you use Euclidean
distance, you always need to start off by calculating all the pair-wise
distances of N points, i.e. a NxN/2 distances matrix. The formation of
clusters is negligible in complexity/time compared to this task. This
means that any time constraints mostly refer to the "closure" measure.
If a 2x"1D" measure is used as previously proposed, your computation
time should drop almost by a factor of N.
--
Harris
.
- References:
- I need to locate pore clusters
- From: vforvasanth
- Re: I need to locate pore clusters
- From: aruzinsky
- I need to locate pore clusters
- Prev by Date: Re: How to know the maximum frequency of an image to apply a filter.
- Next by Date: Re: Is there any way to solve the aliasing of an image in the spatial domain
- Previous by thread: Re: I need to locate pore clusters
- Next by thread: how to save the trained pattern in cognex vision pro software
- Index(es):
Relevant Pages
|
