GDAL Translate – Convert TIFF to PNG with Added Alpha Channel

cloud-optimized-geotiffgdalgdal-translate

I am using gdal_translate to create a Cloud Optimized GeoTIFF:

  gdal_translate ${lvl2_cub} ${tifname} \
                          -co COMPRESS=LZW -co TILED=YES -b ${gdalidx} -co BIGTIFF=YES \
                          -strict -r bilinear -co BLOCKXSIZE=512 -co BLOCKYSIZE=512
  gdaladdo -r lanczos --config COMPRESS_OVERVIEW LZW \
                          ${tifname} 2 4 8 16 32 64
  gdal_translate ${tifname} ${cogname} -co COMPRESS=LZW -co TILED=YES \
                    -co PREDICTOR=2 -co COPY_SRC_OVERVIEWS=YES \
                    -co BLOCKXSIZE=512 -co BLOCKYSIZE=512 --config GDAL_TIFF_OVR_BLOCKSIZE 512

The resulting COG is a 32-bit file. I am then attempting to create an 8-bit thumbnail (JPG or PNG are both workable) where the no data value is mapped to zero and is transparent. To do this, I am pretty sure that I need to get an alpha channel added to the PNG, but the driver does not have a -co ALPHA=YES like the GeoTIFF driver.

Here is my attempt at a gdal_translate call:

gdal_translate -ot Byte -scale $minval $ maxval 1 255 -outsize 300 0 -a_nodata 0 "${cogname}" "${outsubdir}/${name_wo_extension}.png"

How can one add an alpha channel to a PNG using gdal_translate when the input data are single band, grayscale COGs?

Best Answer

By the driver documentation https://gdal.org/drivers/raster/png.html grayscale with alpha is not supported even the PNG format does support it http://libpng.org/pub/png/spec/1.2/PNG-DataRep.html

I checked also that the GDAL autotests for PNG have some tests for RGBA but not any for grayscale+alpha. However, setting a nodata value should be enough, nodata and alpha are alternative ways for defining transparency/nodata.

Related Solutions

gdal – How to Add Alpha Channel to MBTiles Using gdal_translate

For the issue "turned pngs", you may be able to use https://github.com/roblabs/gdal2tilesp

which has support for xyz, from the notes: Generates Leaflet layer for use with the XYZ default format. Where the tms parameter is set to false

I've also tried gdal2tiles.py and mb-util utilities. they made good mbtile from point of transparency but all the pngs are turned upside down see figure2 (gdalinfo shows that origin has wrong Y coordinate with "-")

python gdal2tilesp.py -h

-o OUTPUT_CACHE, --output=OUTPUT_CACHE
                    Format for output cache. Values allowed are tms and
                    xyz, being xyz the default value

[GIS] gdal_translate – raw to png

You have an 8 bit 3 band binary file, interleaved by pixel. Assuming you don't already have a header file that GDAL understands and you know the number of columns and rows, you can create (in code or by hand) a VRT file that describes the layout of the binary file so that GDAL can read it. Write something like the following to a file (can be called anything you like)

<VRTDataset rasterXSize="400" rasterYSize="300">
  <VRTRasterBand dataType="Byte" band="1" subClass="VRTRawRasterBand">
    <SourceFilename relativetoVRT="0">/path/to/file.raw</SourceFilename>
    <ImageOffset>0</ImageOffset>
    <PixelOffset>3</PixelOffset>
    <LineOffset>1200</LineOffset>
  </VRTRasterBand>
  <VRTRasterBand dataType="Byte" band="2" subClass="VRTRawRasterBand">
    <SourceFilename relativetoVRT="0">/path/to/file.raw</SourceFilename>
    <ImageOffset>1</ImageOffset>
    <PixelOffset>3</PixelOffset>
    <LineOffset>1200</LineOffset>
  </VRTRasterBand>
  <VRTRasterBand dataType="Byte" band="3" subClass="VRTRawRasterBand">
    <SourceFilename relativetoVRT="0">/path/to/file.raw</SourceFilename>
    <ImageOffset>0</ImageOffset>
    <PixelOffset>3</PixelOffset>
    <LineOffset>1200</LineOffset>
  </VRTRasterBand>
</VRTDataset>

From the GDAL VRT tutorial:

SourceFilename: The name of the raw file containing the data for this band. The relativeToVRT attribute can be used to indicate if the SourceFilename is relative to the .vrt file (1) or not (0).

ImageOffset: The offset in bytes to the beginning of the first pixel of data of this image band. Defaults to zero.

PixelOffset: The offset in bytes from the beginning of one pixel and the next on the same line. In packed single band data this will be the size of the dataType in bytes.

LineOffset: The offset in bytes from the beginning of one scanline of data and the next scanline of data. In packed single band data this will be PixelOffset * rasterXSize.

So the only things you'll need to change in the example are rasterXSize, rasterYSize, SourceFilename and LineOffset. LineOffset will be 3 x rasterXSize.

You can then use gdal_translate to convert the VRT to PNG:

gdal_translate {appropriate options if required} /path/to/file.vrt /path/to/file.png

Related Question