Converting PNG to GeoTIFF using GDAL Python

Question

I want to ask how to convert PNG to GeoTIFF using GDAL in Python. I ran this code but it's not work and I want to export the output into my computer.

How can I fix my code?

ds = gdal.Open('input.png')
gt = gdal.Translate(ds, 'output.tif', projWin = [-180, 90, 180, -90])
gt = None

Answer 1

I'm assuming that your PNG file doesn't have geolocation information embedded but you know what the corners of the data. I think you just had the wrong option for gdal.Translate. Check this out using a random example image that should span the same extent as you had in your example. I also assume that it is in WGS84 Latitude/Longitude (not quite for my example)

inp = "/vsicurl/https://upload.wikimedia.org/" + \ 
      "wikipedia/commons/thumb/2/23/" + \ 
      "Blue_Marble_2002.png/" + \
      "450px-Blue_Marble_2002.png"
ds = gdal.Open(inp)
gt = gdal.Translate(ds, 'output.tif',
                    outputBounds = [-180, 90, 180, -90],
                    outputSRS="EPSG:4326"
                    )
gt = None

Answer 2

I know that the OP specifically asked for a gdal solution but may I offer another solution that works better for me ?

Here I'm using the rasterIo lib (available on both pipy and conda) to read the png, extract band information and build the output file. As it's more pythonic, I think if something goes wrong you'll have more error logs to help you.

import rasterio as rio 

dataset = rio.open('input.png')
bands = [1, 2, 3] # I assume that you have only 3 band i.e. no alpha channel in your PNG
data = dataset.read(bands)

# create the output transform
west, south, east, north = (-180, -90, 180, 90)
transform = rio.transform.from_bounds(
    west, 
    south, 
    east, 
    north, 
    data.shape[1], 
    data.shape[2]
)

# set the output image kwargs
kwargs = {
    "driver": "GTiff",
    "width": data.shape[1], 
    "height": data.shape[2],
    "count": len(bands), 
    "dtype": data.dtype, 
    "nodata": 0,
    "transform": transform, 
    "crs": "EPSG:4326"
}

with rio.open("output.tiff", "w", **kwargs) as dst:
    dst.write(data, indexes=bands)