I'm new to using GDAL and need to process hyperspectral data. I'm trying to get rid of bands within atmospheric absorption windows, but I'm note sure how to do that in GDAL or other python tools. My dataset has 284 bands, but I want to retain only 233 of them (remove bands 1-3, 130-141, 191-213, and 272-284).
Is there a way to do this in GDAL or are other methods better?
You could use the gdal.Translate function and pass a list of bands to keep (bandList argument):
from osgeo import gdal source_dataset = "input1.tif" output_dataset = "output1.tif" bad_bands = ( (1, 2, 3), range(130, 142), # note 142 as range func excludes the last value, so this will give you 130-141 range(191, 214), range(272, 285) ) src = gdal.OpenEx(source_dataset) bad_bands = [b for bands in bad_bands for b in bands] # flatten nested tuples and ranges into a single list all_bands = range(1, src.RasterCount+1) # Note GDAL band numbers start at 1 keep_bands = [b for b in all_bands if not b in bad_bands] out = gdal.Translate(output_dataset, src, bandList=keep_bands) Or with rasterio, you could write each band out separately (to the same raster):
import rasterio as rio source_dataset = "input1.tif" output_dataset = "output1.tif" bad_bands = ( (1, 2, 3), range(130, 142), # note 142 as range func excludes the last value range(191, 214), range(272, 285) ) with rio.open(source_dataset) as src: profile = src.profile.copy() bad_bands = [b for bands in bad_bands for b in bands] # flatten nested tuples and ranges into a single list all_bands = range(1, src.count+1) # Note Rasterio band numbers start at 1 keep_bands = [b for b in all_bands if not b in bad_bands] profile.update(count=len(keep_bands)) with rio.open(output_dataset, "w", **profile) as out: for i, b in enumerate(keep_bands, start=1): out.write(src.read(b), i) 