Hello,
I am experiencing difficulty in performing texture analysis after mosaicing 14 high-resolution multispectral RGB raster images. Both SNAP and ENVI are very time-consuming. Is there any alternative solution to this?
Thank you
14 rasters with 10 bands are probably a really large input…
How about performing a PCA on the multispectral mosaic first and calculate the textures only on the first three components?
Also, many GLCM textures are highly redundant, for example Energy, Homogeneity, MAX and ASM aer often quite similar, so there is rarely a real need to calculate all textures on all bands.
The original rasters come with only 3 bands and I am performing texture measures only on the green band. I have also reduced the number of texture features but I am still experiencing the same.
ok, so it is only the size of the mosaic which makes problems. How large is it?
The size is 83440 x 24040. Each raster images has size of 11920 x 12020.
so to me there is no wonder why the derivation of textures is time-consuming 
This is way beyond the average raster size, I’d say.
Finally the problem is solved. Yesterday, I computed 5 texture measures on the green band using GRASS GIS. Though it took an hour to compute, but still very less than the ENVI.
1 Like
good to hear. Different software packages can drastically vary regarding accessing, manipulating and writing files.
1 Like
Excellent.
Note that virtual mosaics are possible in GRASS GIS (like GDAL VRT):
https://grass.osgeo.org/grass-stable/manuals/r.buildvrt.html
Maybe interesting for you.