Search code examples
pythonimagegisgdal

gdal translate scale and save to jpg using python


I know how to use gdal translate to scale and save to jpg via cmd line:

gdal_translate image.bsq image.jpg -of JPEG -outsize 10% 10% -scale

This produces (what I call a nice image):

enter image description here

I would like to produce similar image via python, something like:

from osgeo import gdal
img_bsq  = 'image.bsq'
img_jpg  = 'image.jpg'  
gdal.Translate(img_jpg, img_bsq, format='JPEG', width=1024, height=0, scaleParams=[[500,1000,10,20]])

The issue I think is how to properly select scaleParams. It seems the -scale on the cmd line compute values automatically, as per man gdal_translate:

-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.

Any tip on how to select scaleParams (or other relevant options)?


Solution

  • You should be able to leave it blank here as well, like:

    gdal.Translate(img_jpg, img_bsq, format='JPEG', width=1024, height=0, scaleParams=[[]])
    

    This causes GDAL to make a guess itself, as described in the docs:

    -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.

    http://www.gdal.org/gdal_translate.html

    Alternatively, you could also retrieve the statistics (per band) and make something up yourself.

    Getting the statistics:

    ds = gdal.Open('img_bsq')
    stats = [ds.GetRasterBand(i+1).GetStatistics(True, True) for i in range(ds.RasterCount)]
    ds = None
    
    vmin, vmax, vmean, vstd = zip(*stats)
    

    With those statistics you should be able to come up with some desired stretching. If you want to scale between the min and max for each band you could do:

    scaleParams = list(zip(*[vmin, vmax]))
    

    Or if you want to use the absolute highest and lowest (over all bands)

    scaleParams = [[min(vmin), max(vmax)]]
    

    etc.