Cfrtjryk

I would like to convert a GeoTIFF to PNG or JPEG RGB, and my GeoTIFF is stripped of a lot of valuable data (I think..)

I am on ubuntu and I am using gdal_translate to try and make a GeoTIFF turn into a regular PNG.

The GeoTIFF I have 'works'. I know because I can view it in QGIS on my Mac.

My GeoTIFF has 8 bands and I dont know which band corresponds to RGBA or what the other 4 bands stand for. It appears none of the bands have a valid 'color table'

  gdal_translate -of PNG test.tiff output.png

gives me

  ERROR 6: PNG driver doesn't support 8 bands.  Must be 1 (grey), 2 (grey+alpha), 3 (rgb) or 4 (rgba) bands.

When I try and just use any combination of bands for RGBA like this:

 gdal_translate -of PNG -b 1 -b 2 -b 3 test.tiff output.png

I just get a black image

In QGIS if I right click on my image and go to properties > Symbology, I see that the redband is band 1 (grey) (min 0, max 439, the green band is band 2 (min 0, max 460), the blue band is band 3 (min 0, max 454). and the image appears nicely in rgb - so the GeoTIFF 'works'.

I've also tried to make a gray image (even though I want an RGB), just because it looks like QGIS is saying band 1 is gray

  gdal_translate -of PNG -expand gray -b 1 test.tiff output.png

but that gives me

  ERROR 1: Error : band 1 has no color table

I dont really know what else to try, or if my image is black because I need to do something with 'scales' or something?

When I run tiffdump on my file I get

Magic: 0x4949 <little-endian> Version: 0x2a <ClassicTIFF>

Directory 0: offset 8 (0x8) next 0 (0)

ImageWidth (256) SHORT (3) 1<2604>

ImageLength (257) SHORT (3) 1<2233>

BitsPerSample (258) SHORT (3) 8<16 16 16 16 16 16 16 16>

Compression (259) SHORT (3) 1<1>

Photometric (262) SHORT (3) 1<1>

StripOffsets (273) LONG (4) 2233<18288 59952 101616 143280 184944 226608 268272 309936 351600 393264 434928 476592 518256 559920 601584 643248 684912 726576 768240 809904 851568 893232 934896 976560 ...>

SamplesPerPixel (277) SHORT (3) 1<8>

RowsPerStrip (278) SHORT (3) 1<1>

StripByteCounts (279) LONG (4) 2233<41664 41664 41664 41664 41664 41664 41664 41664 41664 41664 41664 41664 41664 41664 41664 41664 41664 41664 41664 41664 41664 41664 41664 41664 ...>

PlanarConfig (284) SHORT (3) 1<1>

ExtraSamples (338) SHORT (3) 7<0 0 0 0 0 0 0>

SampleFormat (339) SHORT (3) 8<1 1 1 1 1 1 1 1>

33550 (0x830e) DOUBLE (12) 3<2.20356e-06 2.20356e-06 0>

33922 (0x8482) DOUBLE (12) 6<0 0 0 -82.0121 27.3184 0>

34735 (0x87af) SHORT (3) 32<1 1 0 7 1024 0 1 2 1025 0 1 1 2048 0 1 4326 2049 34737 7 0 2054 0 1 9102 ...>

34736 (0x87b0) DOUBLE (12) 2<298.257 6.37814e+06>

34737 (0x87b1) ASCII (2) 8<WGS 84|>

gdalinfo returns

Band 1 Block=2604x1 Type=UInt16, ColorInterp=Gray

Band 2 Block=2604x1 Type=UInt16, ColorInterp=Undefined

Band 3 Block=2604x1 Type=UInt16, ColorInterp=Undefined

Band 4 Block=2604x1 Type=UInt16, ColorInterp=Undefined

Band 5 Block=2604x1 Type=UInt16, ColorInterp=Undefined

Band 6 Block=2604x1 Type=UInt16, ColorInterp=Undefined

Band 7 Block=2604x1 Type=UInt16, ColorInterp=Undefined

Band 8 Block=2604x1 Type=UInt16, ColorInterp=Undefined

You can try to run a command with the options that handle the input bands in the best way possible and result in a well-constructed image similar to the one shown by qgis:

gdal_translate -ot Byte -of PNG -b 1 -b 2 -b 3 -scale_1 0 439 -scale_2 0 460 -scale_3 0 454 -a_nodata none -colorinterp_1 red -colorinterp_2 green -colorinterp_3 blue test.tiff output.png

UPDATE (based on comments):

The answer works for GDAL >= 2.3.

Previous versions of GDAL do not support the -colorinterp option.

In that case, the solution is taking advantage of the PHOTOMETRIC=RGB creation option:

gdal_translate -ot Byte -of PNG -b 1 -b 2 -b 3 -scale_1 0 439 -scale_2 0 460 -scale_3 0 454 -a_nodata none test.tiff output.png -co PHOTOMETRIC=RGB

If the Geotiff has 8 bands, these don't have to be any colour reference. RGB has values between 0 and 255 for each band. So your data does not fit to RGB channels. Gray scales could be any range.

It could also be that an undeclared NODATA value is fooling you.

Either the metadata or the author of the file should reveal what the 8 bands are about. QGIS just makes a guess for RGB, but that does not have to be right.