The same image raster in 3 formats, TIF, COG, JP2. COG have fast GET. Is the fast 'get' represented in fast display when panning? Am I supposed to see COG out perform the others when: if all 3 are on cloud? if all 3 are on network? if all 3 are on local drive?
- COG should 'generally' be faster as the client is only requesting the part of the file it currently needs, and can utilize pyramids if included. But that assumes your client is COG friendly, otherwise you're 'get'ing the entire filetomtomnz– tomtomnz2024-06-28 03:11:12 +00:00Commented Jun 28, 2024 at 3:11
Add a comment |
1 Answer
Cloud-optimised geotiffs (COGs) do provide benefits, even when not involved with the cloud, especially when considering displaying the data.
Imagine a case where you are loading a multi-GB image into a GIS system - the first thing that is shown is a zoomed out version of the image. In the case of an old school TIF, your computer has to generate a low resolution version of the data on-the-fly, to visualise the imagery. To do this, it has to load, and do calculations on the full file. Meanwhile, with a geotiff, your computer can directly load the recalculated overview. This difference does greatly speed up the process of displaying the data. Similar benefits can be seen from a JP2, if it is built for it, but JP2 has many variants - so it is not clear cut.
As such - COGs are great and should be consistently be used over old-school un-optimised geotiffs, as they are faster to work with in all environments, be they cloud, network, or local. The only situation where a un-optimised geotiff should be used, is when the data is a temporary file, where the next step in the processing chain will work on the whole file, and is not going to be visualised. With regards to JP2 - it is a good format, and has very efficient compression algorithms available. However, it is a broad definition, so it can also be inefficient, if insufficient processing is done.
Lastly, it should be noted that, cropping and otherwise working with the data is also sped up a bit, by using COGs.
- 1GeoTIFFs can also be tiled and have overviews. Let's call them GIS-friendly. What is special for COG compared to GIS-friendly is that from COG data can be accessed nicely with HTTP range requests. In local use that is not important and GIS-friendly is as fast.user30184– user301842024-06-27 07:01:41 +00:00Commented Jun 27, 2024 at 7:01
- @user30184 - COG is just a subset of GeoTIFF, so it is a bit of a circular argument. And the good HTTP range request performance comes as a result of the tiling, and compression.Mikkel Lydholm Rasmussen– Mikkel Lydholm Rasmussen2024-06-28 05:23:49 +00:00Commented Jun 28, 2024 at 5:23
- All in all, my answer is mainly focused on a raw default geotiff vs COG vs JP2. Given how modular both geotiff and JP2s are, covering all the variability is impossible. However, using COG as the default output format can generally be considered a good choice, for the last step in a processing chain, as it'll overall provide benefit for most users. Is it always the perfect choice - no, but it is a good place to start.Mikkel Lydholm Rasmussen– Mikkel Lydholm Rasmussen2024-06-28 05:27:44 +00:00Commented Jun 28, 2024 at 5:27