i.cluster sample increment dependence

[Richard Wagener]

I run i.cluster on AVHRR data (384 by 512 image with three channels).
When I accept the default input parameters to i.cluster, I end up with
14 clusters.  However, when I try to change the suggested sample step
size in x and y from (3,5) to (1,1), i.cluster reduces to just 1
cluster between iteration 4 and 5.  In the two iterations before that the
biggest cluster (almost half of all the points) shifted back and forth
between cluster number 9 and 10 (of 15).

I got a similar behavior if I ran i.cluster with the default step size
first and then used the output signature file as a SEED signature for
a second iteration with stepsizes (1,1).

Did anybody else have similar experiences with i.cluster and what is
the likely cause of this behavior? 
-- 
Tschuess ...rick...
_______________________________________________________________________________
Richard Wagener				| 
Bldg. 426, rm. 43			| IP: wagener at bnl.gov
Environmental Chemistry Division	| DEC: 44048::rwagener
Brookhaven National Laboratory		| phone: 1-516-282-5886
Upton, New York 11973, U.S.A.		| fax:   1-516-282-2887
_______________________________________________________________________________