3

Trying to create 1 mosaic dataset from 2 million files. Would it be better to break the files into regions and then create regional mosaic datasets and then make a mosaic dataset of the mosaic datasets (does that make any sense?)? Or just slug it out and do one dataset of 2 million files? BTW the datatype is CIB

Curt
  • 91
  • 2
  • 2 million images in one go? Its not a good idea for any software. Large images are very slow when thrown in to mosaic. Every time you add image to your mosaic, it will increase the size, which increase time of processing with next image. So processing time will gradually rise. Moreover, in case when your images are scattered all over the place (not in proper order) and your mosaic will start with 2 extreme images, mosaicing for all the rest images will consume maximum time. – Tomek Apr 11 '12 at 09:18

2 Answers2

2

My personal experience in mosaicking even just a few hundred tiles together is that it is much faster to do it iteratively by ganging them up in sub-regions first. Indeed I found with the arcinfo workstation mosaic command that some mosaic operations wouldn't complete at all without sub-grouping first. In the case if millions I'd expect to repeat many times (though row and column pixel dimensions may be a more significant factor than number of files).

I'd also investigate using gdal's VRT format to delay the actual computation as far into the process as possible: How can I merge geotiffs with a vrt file?. There's also the related Converting CIB data to Google Tiles with GDAL

Community
  • 1
matt wilkie
  • 28,176
  • 35
  • 147
  • 280
0

I would look into LizardTech software their software runs in a x64 bit enviornment and should be able to handle the workload. Either way make sure you have a lot of time for processing.

Ken Carrier
  • 166
  • 4