I'm currently building a project that will involve using raster imagery in the field on a Windows tablet with QGIS/Intramaps Roam and haven't been able to get a satisfactory understanding of a few questions about raster imagery.
The data I'm using comes from MassGIS and is in small tiles in jpeg2000 format. I'm using about 500 tiles. In order to use the data, I need to do two things:
- Re-project the data to WGS84
- Mosaic the tiles to form a single seamless image.
There seems to be two ways to do this, one very fast the other very slow. The first is to build a virtual raster tile (vrt) and then use gdalwarp to project to another vrt. Very fast. The second is to build a vrt and then use gdalwarp (or raster project in ArcGIS) to output a new jpeg2000 raster. Very slow.
Since storage space is not really an issue, my question is, which of these will perform better in the field and why? It seems like a vrt that is simply projected from another vrt is more like doing an OTF projection and would thus be slower than a jpeg2000 raster, but I haven't confirmed this anywhere.
If this sounds incredibly dense and there is something better than both these options, please share!