Archive for raster data

FME vs Global Mapper

Posted in FME, Software with tags , , on July 9, 2009 by hansvandermaarel

So for a long time I’ve been saying I generally don’t use FME for raster work because it’s slow. This was all based on some bad experiences way back with FME 2006 (the first version to support raster data if I recall correctly) and the infamous 37-Gigabyte temporary file that was left on my hard drive after a mosaicking operation crashed. So I’d mostly been using FME for dynamically making a clip out of a larger raster file, which I would invariably produce with Global Mapper.

Recently, Dale Lutz of Safe Software had suggested that I’d give FME another chance, because they had made a lot of progress in the handling of raster data. So I decided to give it another go, pitting the FME 2010 “UC” beta against Global Mapper 10.02.All tests were done in Windows Vista x64 running through Bootcamp on my Mac Pro with 10 Gb of ram. All data was present on the local hard drive.

Test 1 : tiling a raster

I took a TIFF file of Tom Patterson’s Natural Earth. Specifially, I had already produced a combined version. Tom offers it in “west” and “east” files, I put those together to get the entire earth in a single, almost 2 Gb, GeoTIFF. I set up both FME and Global Mapper to tile this file into 90 x 90 degree chunks and output each of them as GeoTIFF. FME blazed through this in a mere 32 seconds. Global Mapper took close to 18 minutes.

Test 2 : mosaicking the tiles back to an ECW

Another common procedure, mosaicking a number of raster tiles and creating a single large image out of them. I did try to set the compression factor in both FME and Global Mapper to the same value but since they specify them in different ways, it’s difficult. I got close, but not quite. In any case, compression generally isn’t going to give very nice results on map images, where you often have areas with solid colors. Also, I did struggle a bit with FME to get it to deal with the raster bands (the alpha channel) in the GeoTIFF’s correctly. Still, I only measured actual processing time, not how much time it took me to set it up (which, to be honest, is roughly the same in both FME and Global Mapper). I fed both applications their own tiles that were output in test 1. FME got through this in about 1:30, Global Mapper took 3:30

Final verdict

Easy one really, FME took a combined total of 2 minutes of processing, against 21 minutes for Global Mapper. Need I say more?

I do think the two products each have their unique strengths though, so I will certainly not stop using Global Mapper. But I will use FME much more often for this kind of work.