> Most are 500x500px
> 
> > How big do you want the output image? (All in
> > pixels pls.)
> 
> 101000x149000px

Oooouch, nice blowjob. Unless you have software which specifically
operates in a sparse-matrix environment, i.e. not storing elements which
are 0 or non-existant, you'll need (assuming 8bit/pixel) a byte for each
of those - that's 15GB. I've heard of quad AMD64 with 4GB of RAM per
processor (the sort of thing R�diger will be looking at), but you don't
want to know the price.

I think you should re-examine the problem and find a solution not
involving an outragously sized bitmap... ;) Think further - even if that
bitmap compressed very well and could be stored in 10MB disk space, each
time you access it you'd have to decompress it into memory. And unless
you can access it, there's little point in having it. If you really
can't do without my guess is you'd have to program your own image
processing routines, implementing the sparse type tricks, which doesn't
sound efficient.

Given you're at NIWA, do you have access to matlab and can check whether
your files can be read as binary data in a sparse matrix? Octave is a
2*free replacement for matlab but it's not as advanced.

Can you break down the problem into chunks of say 10000x10000 pixels
max?

In any case it looks like a grunty PC would be in order.

Let us know how you get on!

Volker

-- 
Volker Kuhlmann                 is possibly list0570 with the domain in header
http://volker.dnsalias.net/             Please do not CC list postings to me.

Reply via email to