I have written a version in bash that uses the pbmtools to do the hard lifting work. It seems to run quite fast too.
16000x16000 jpeg converted in about 30 mins on a 2.4G p4 with 512 M ram.
A 8000x2800 image was converted in approx 30 secs.
It is available from http://nosm.helix.net.nz/pano/zoomifyimage.sh
It needs the packages (debian names) of
libjpeg-progs for djpeg and cjpeg and
netpbm for the scaling and cutting out.
Very interesting, I never thought of doing this in a shell script. (Probably because I wouldn't have known about pbmtools.
Can I add this to my documentation/software. In the most recent release, I described two other approaches in the readme file and put the code and documentation in a 'contrib' folder. I am not sure when I will do another release (probably a while...) but I can at least add it to CVS for now.
If you want to do this, do you want to write up a little more about it, or can I just use the information from your pose?
Thanks for letting us know about this--it might be a lifesaver for someone browsing these forums.
You are welcome to add it to your next release, and to cvs.
I can write up how it works, how to use it, and what is needed to use it for you.
Just let me know of the most effective way to get the information to you.
Thanks. You can send files to me at email@example.com .
I just added Fran's code and documentation to the 'contrib' folder of the ZoomifyImage package in CVS. Its available there and will be included in the next release. (I will probably wait until I make more upgrades or have more contributions before doing another release.) If you don't like CVS, the link provided above by Fran should get you going.
Thanks again Fran!
Sorry to bring up this old thread.
I tested your script on a 41736x64713 file. It took only 14 hour an a p4 :-)
It created 216 TileGroup directory's.
I had to change the first line to #!/bin/bash from #!/bin/sh
and I also had to recompile the latest cvs version of netpbm to handle such large files. It also look like pnmfile are replaced by pamfile.
It was by the way a map exported from arcgis in two sessions and joined again with nip2/vips.
Thank you very much it worked very well.