Speed up raster symbology statistics.

1152
2
04-29-2021 10:11 AM
Status: Open
Labels (1)
NickWeil
New Contributor III

Computing stats and histogramComputing stats and histogram

I have recently been working with some hyperspectral rasters and have noticed that applying a stretch in ArcGIS Pro is very slow. I decided to perform some tests to get a speed benchmark. The rasters are 16.5 GB  with 274 bands and I apply a Standard Deviation stretch. This process takes 3 minutes. I compared the process in ArcMap and the same size raster took. 1.68 seconds. If I bring the raster into Pro after calculating statistics in ArcMap, it displays instantly with no additional calculations needed. It would be beneficial if this task could be sped up.

2 Comments
KoryKramer

Hi Nick,

We wanted to follow up on this now that the team has been able to look at the issue.  What ArcGIS Pro does is calculate the statistics for all bands and store those in an auxiliary file along with the dataset.  As you're seeing, if given a hyperspectral dataset with 247 bands, the initial calculation will take longer than ArcMap which is calculating only the 3 display bands on the fly.  ArcMap does not store these statistics, so the calculation would need to be repeated every time the dataset is used in new maps.

In ArcGIS Pro, on the other hand, once the statistics are calculated for all bands, since these are persisted, there is no need to recalculate when that dataset is used again in any other map or project, or when the user switches the bands being displayed - the calculation is already done.

We're leaving the idea Open so that we can gauge whether this is affecting other ArcGIS Pro users, but we wanted to provide the explanation of the difference you're seeing between ArcMap and ArcGIS Pro.

Thank you

RabiyaDar

Hi @KoryKramer, I'm having a similar issue where a single-band raster with 0.6m cell size and a relatively small processing extent (Charlotte Harbor estuary, FL) is taking over an hour to compute statistics and histogram in order to classify it in symbology. This is not a new issue for me where seemingly simple processes take hours to complete, while I have peers who are doing analyses on global datasets but their tools and processes run in minutes. I'm working in ArcGIS Pro. Any tips on what the issue may be would be greatly appreciated.