Unsupervised classification and i.cluster

[Harini Nagendra]

Hello - does anyone know the algorithm based on which i.cluster was 
written? I am trying out unsupervised classifications by fiddling with 
the various parameters specifiable in i.cluster - except that I don't 
really understand what they mean. Specifically,

1]  How is a cluster determined?i.e. which clustering algorithm is used? 
IDRISI uses a "histogram peak technique of cluster analysis" - does GRASS 
use this or something different? Any references I can read up as to the 
method used?

2] Are the number of initial clusters used before the 1st pass is made 
only, or do they count in subsequent passes also?

3] What are the units that the minimum class seperation is to be 
specified in - the 4.0 imagery tutorial says that users commonly specify 
a value between 0.5 and 1.5 - doesn't make too much sense unless one 
knows 0.5 of what- can anyone help?

4] The manual for i.cluster says that both the number of initial clusters 
and the min. class seperation will be used to calculate the number of 
final clusters - that is why I wanted to know whether the no. of initial 
clusters is only used n the clustering alg. before the passes, and the 
min. dist used to rearrange pixels subsequently, in each pass. This is 
what seemed likely to me, but I would like to have it confirmed by 
someone who knows better.....

Sorry, this is a rather "questiony" mail - I hope it isn't too long and 
involved.

Thanks a lot for all help
Harini  

------------------------------------------------------------------------
| Harini Nagendra               E-MAIL : harini at ces.iisc.ernet.in      |
| Center for Ecological         PHONE  : 91 80 309 2639 (Hostel: LR-94)|
| Sciences                             : 91 80 309 2506 (Department)   |
| Indian Institute of Science   FAX    : 91 80 334 1683                |
| Bangalore 560 012 India       TELEX  : 845 8349 IISc IN              |
------------------------------------------------------------------------