[gdal-dev] Gdal_edit and TIFFTAG_EXTRASAMPLES

Even Rouault even.rouault at spatialys.com
Sat Sep 30 14:45:16 PDT 2017


On samedi 30 septembre 2017 21:01:57 CEST Rahkonen Jukka (MML) wrote:
> Hi,
> 
> I have been searching an easy way to change the interpretion of 4-channel
> images between RGBA and RGB+undefined. Wouldn't if be useful to make
> gdal_edit to support also changing the TIFFTAG_EXTRASAMPLES metadata? Or do
> we have some other easy way for performing that task?

Jukka,

I couldn't find a clean working way.

I had though that calling ds.GetRasterBand(4).SetColorInterpretation(gdal.GCI_AlphaBand or 
gdal.GCI_Undefined) would work (which could have been done by a new option of gdal_edit 
to modify the color interpretation). But for some (not clear, and likely unneeded) reason, this 
is restricted on a fresh new file, not yet "crystalized" (if you do it on an existing file, the 
change will only go to the .aux.xml file). I don't see a strong reason for this restriction, so it 
could probably be removed.


I've also tried the tiffset utility, but for some reason it absolutely wants to set at least 2 
values in the ExtraSamples tag. Sounds like a bug of tiffset/libtiff

So to set alpha:

tiffset -s 338 2 2 0 test.tif

338 = code for ExtraSamples
2 = 2 values
2 = alpha unassociated
0 = undefined (ignored, and somehow invalid since that sample doesn't exist)

(should normally be tiffset -s 338 1 2 test.tif if that worked properly)

or to set undefined

tiffset -s 338 2 0 0 test.tif
(should normally be tiffset -s 338 1 0 test.tif if that worked properly)

GDAL will be happy, but technically the TIFF is not so correct with that 2nd extra sample, 
since for a Photometric=RGB with 4 samples, you should just have 1 extra sample.


I would also have suggested to just take your favorite hexadecimal editor, and just alter the 
proper byte, but I don't feel this will be a popular option ;-)

Even

-- 
Spatialys - Geospatial professional services
http://www.spatialys.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/gdal-dev/attachments/20170930/1228340c/attachment.html>


More information about the gdal-dev mailing list