Forum

Author Topic: How to reduce size of generated DEM geotiff  (Read 1024 times)

c86

  • Newbie
  • *
  • Posts: 2
    • View Profile
How to reduce size of generated DEM geotiff
« on: February 08, 2023, 07:35:20 PM »
Hi,

I am trying to export a DEM that I can put into Relief Visual Terrain (RVT) to generate further visualisations. However the geotiff that is generating in agisoft is quite large at 2.66gb compared to the 5.41mb LiDAr files I put in RVT.

RVT appears to be really struggling with generating visualisations from my DEM taking quite some time. I have had to do each visualisation individually and it appears that I cant even get the multi direction hillshade and Simple Local relief map to generate.

Is there away to make the DEM geotiff generated by agisoft smaller and see if RVT can cope better with it?

Thanks,
« Last Edit: February 09, 2023, 03:38:36 PM by c86 »

dpitman

  • Full Member
  • ***
  • Posts: 246
    • View Profile
Re: How to reduce small of generated DEM geotiff
« Reply #1 on: February 09, 2023, 03:39:10 AM »
You could export a point file instead of a raster DEM.

c86

  • Newbie
  • *
  • Posts: 2
    • View Profile
Re: How to reduce size of generated DEM geotiff
« Reply #2 on: February 09, 2023, 08:21:51 PM »
Hi, I used the export points and then the .txt sa the file type because it was the only one I could see that RVT accepts. It's still quite a big file and RVT is struggling. not sure if it is possible to make them any smaller or it is just because the size of my model.

DavidD

  • Newbie
  • *
  • Posts: 22
    • View Profile
Re: How to reduce size of generated DEM geotiff
« Reply #3 on: February 10, 2023, 04:07:13 PM »
Thanks for bringing up this discussion!

As far as I understand it, the DEM geotiff basically stores a 32bit float value for every "pixel".
An Ortho in contrast has 24bit per pixel (8bit for R, G, B each + sometimes alpha) - but orthos can be compressed with extremely efficient lossy codecs like jpeg(XL) or webp.
The DEM in contrast can only be compressed losslessly.

In order to save more space, it would be amazing to have the option to store the data in half-precission 16bit-floats for example. Or even quantizise further down and map it to 8 bit integer.

Actually JPEG XL could also be the solution here...
It allows an arbitrary number of channels 1 to 4000+ and each of them can have either binary or integers and floats up to 64 bit. It also allows lossy compression even on floats...