Skip to content

A "cluster" package question and request for images

1 message · Bio7

Original
Bio7
# 
Dear R developers,

i'm developing a Java application with a very efficient image transfer api
between ImageJ and R.
In my next release i can transfer bytes to R very fast with the help of the
Rserve application.
Furthermore i added an easy to use interface to the clustering algorithm
"clara" in the cluster package.
At the moment i can cluster an R,G,B image with size 4000*4000 without any
problem on a 32-bit
package (with 3GB ram) with a good speed . For the clustering i have to
convert the byte vectors to integer vectors
and add them to a matrix. My question is if it is possible to change the
package in that way that the function clara 
which i use for clustering also accepts smaller datatypes so that i can
cluster bigger images (or more dimensions) without running out of memory. In
addition is there a better
way to transfer the vectors to this function maybe without the need to bind
the vectors to a
matrix (which costs little bit more time)?

Any answers or comments are appreciated.

With kind regards

M.Austenfeld