Select to view content in your preferred language

Holes in Mosaic Dataset After Stretch

246
6
01-21-2025 01:18 PM
lah
by
Frequent Contributor

I'm struggling trying to load a bunch of orthomosaics into an EGDB Mosaic Dataset. The original orthos look fantastic, but when I add them to the Mosaic Dataset, they are either super dark and impossible to make out, or really washed out/hazy in color.

I have played around with the stretch properties, and think I finally got the color to start looking better using PercentMinMax at 0.5 each, however, the mosaicked dataset then comes out with a bunch of holes in the darker areas (trees, water). If I remove the minimum percent clip, I don't have empty spots, but my color is bad again.

I'm using ArcGIS Pro 3.3.2, but this is also published to ArcGIS Enterprise 11.3 (image server) as a Imagery Layer (from a Referenced MD) and having the same effect there. My original orthos are in an S3 bucket.

I'm very new to working with imagery/rasters and would appreciate any insight from someone more experienced in this world on what I could be missing or anything else I should try. This is ultimately going to be a very large, ongoing collection and want to make sure I am setting this up correctly.

Thank you!

0 Kudos
6 Replies
DanPatterson
MVP Esteemed Contributor

did you calculate statistics for your data first?

Calculate Statistics (Data Management)—ArcGIS Pro | Documentation


... sort of retired...
0 Kudos
lah
by
Frequent Contributor

I was checking the boxes to Calculate Statistics and Build Raster Pyramids under Raster Processing when loading to the MD, but I guess that is not the same.

As a test, I just ran the Calculate Statistics on one of my original orthos and reloaded it, but am still seeing the same holes.

0 Kudos
JeffreySwain
Esri Regular Contributor

In the 'holes' can you identify the pixel values?  It looks as if your water pixels are being confused for NoData. Depending on the pixel values the stretch appears to be removing them during the process based on the histogram. Does the effect happen with any other stretch?

0 Kudos
lah
by
Frequent Contributor

Yes, I can still identify pixels in the empty areas. It does seem to happen to some degree with any stretch that I apply. This is just the best combination I tried that resulted in the best color/look.

0 Kudos
JeffreySwain
Esri Regular Contributor

What are the pixel values ? 

0 Kudos
lah
by
Frequent Contributor

It varies with my %MinMax stretch, but looking like legitimate values: examples - (43, 191, 32), (32, 64, 64), (128, 64, 223)

With a StDev stretch, they identify as either NoData or 255,255,255

0 Kudos