[GIS] Preserving nodata values when using gdal_translate -scale

gdalgdal-translaterasterraster-conversion

I have four 12 – bit Bands in single images which I want to merge to one 8bit Geotiff using GDAL (GDAL 2.1.3)

First thing I do is building a VRT:

gdalbuildvrt -srcnodata 0 -separate bands_merge.vrt band1.jp2 band2.jp2 band3.jp2 band4.jp2

Then with gdal_translate I create the tif.

gdal_translate -of GTiff -ot Byte -scale 0 4095 0 255 bands_merge.vrt bands_merge.tif

That works just perfect. But as the nodata value of the source files is 0 and all values will be scaled, original values below ~8 will become 0.

My first thought on how to fix that issue was using -scale 1 4095 1 255 but that does not work because it assigns 1 to all previous 0 values.

How can I make sure that the -scale 0 4095 0 255 does not create false nodata values ?

Best Answer

Instead of using NODATA, in this case it is better to use a mask:

As the NODATA value is 0, I can easily create the mask with -co NBITS=1 as everything above a value of 1 will be cut to 1.

# Create the mask
gdal_translate -of GTiff -ot Byte -co NBITS=1 band1.jp2 mask.tif
# Create VRT
gdalbuildvrt -separate bands_merge.vrt band1.jp2 band2.jp2 band3.jp2 band4.jp2 mask.tif
# Create the tif
gdal_translate -of GTiff -ot Byte -b 1 -b 2 -b 3 -b 4 -mask 5 -scale 0 4095 0 255 bands_merge.vrt bands_merge.tif
# Remove the temporary files
rm bands_merge.vrt mask.tif

Unfortunately QGIS cannot handle GeoTiff mask files yet, but I need it for GeoServer.

Related Solutions

[GIS] How to create mask bands in GDAL and Python

for one particular band, here's how I create a nodata mask.

import gdal
import numpy as np

nodata = -9999#or set it based on the raster nodata value.
r = gdal.Open(rastername)
raster_arr = np.array(r.GetRasterBand(band).ReadAsArray())

nodatamask = raster_arr == nodata

#do your thing and end up with 
#a result_raster that you'd like to save

result_raster[nodatamask] = nodata
#then save.

Of course, if you're doing logarithms and you have negative values, you'd need to make a temporary nodata value.

nodatamask = raster_arr == nodata

raster_arr[nodatamask] = 100000000

#do your thing

result_raster[nodatamask] = nodata
#then save.

[GIS] Scale 16 bit to 8 bit within range using GDAL

If you don' want the values above 255 to be cut, you need to scale them down. For that purpose gdal_translate provides the option -scale:

From the Manual:

-scale [src_min src_max [dst_min dst_max]]: Rescale the input pixels values from the range src_min to src_max to the range dst_min to dst_max. If omitted the output range is 0 to 255. If omitted the input range is automatically computed from the source data.

So what you can use is the following:

gdal_translate -of GTiff -ot Byte -scale 0 65535 0 255 src_dataset dst_dataset

Why 255 and 65535 ?

You have 2^nbits values. As the min value is usually 0, the max value is 2^nbits-1.

8 bit --> 2^8 = 256 values --> 0-255
16 bit --> 2^16 = 65536 vaues --> 0-65535

Best Answer

Related Solutions

[GIS] How to create mask bands in GDAL and Python

[GIS] Scale 16 bit to 8 bit within range using GDAL

Related Question