Did you define and enable a no data value for all three rasters in the band properties?
Aha！Thank you very much! It’s what confuses me.
A little related to this topic…I wanted to ask if ESA SNAP is a good program to make unsupervised classification and if it is possible with images which have no bands. My picture is a screenshot fom google earth. Thanks.
SNAP offers two algorithms for unsupervised classification, k-means clustering and EM cluster analysis. To my experience, the second takes incredibly long, so I recommend k-means for a first try.
Even if you make a screenshot from Google Earth, your image has three bands (red, green, blue), and SNAP will read the screenshot as a three band raster. So you have at least a three-dimensional feature space. But the quality of the image will probably not be very good due to compression of the image and the reduced spatial resolution.
Technically, this should work but I’m not sure if it will give you good results.
Do you know how big the image could be so it is possible to do it with SNAP? The reason why I am trying it in the first place is because QGIS and his SCP-plugin can´t analyse it because of its size > 1GB.
Thank you very much again.
saved as a TIF file it could probably work even when it is larger than 1 GB