Writing 3 channels to 8-bit TIF in Python using gdal

Question

I have some Python code to write 3 channels to file. Here, channels is an array of 3 numpy 2D-arrays (RGB), outfile is the filename, and rows and cols are the image dimensions.

def save_tiff(channels, outfile, rows, cols):
 outdriver = gdal.GetDriverByName("GTiff")
 outdata = outdriver.Create(str(outfile)+".tif", rows, cols, len(channels), gdal.GDT_Byte)
 # write the arrays to the file
 i = 1
 for c in channels:
 outdata.GetRasterBand(i).WriteArray(c)
 outdata.GetRasterBand(i).FlushCache() 
 i += 1

However, this results in 24-bit output images (probably because of 3xGDT_Byte channel). How do I get a single 8-bit image from 3 channels (R,G,B) in Python using gdal?

• Steven Kay: Do you mean 8-bit paletted? You could use gdal's rgb2pct.py.
• appel: I'm unsure what paletted means, in the end I want to re-attach geodata to my TIF file using listgeo/geotifcp but it doesn't seem to work with 24-bit images. I tested it succesfully with an example 8-bit image so that's why I'm trying to convert it.

Answer 1

I always use this function to create 3 channel images (or more, just changing the parameters):

def array2raster2(fname, matriz, geot, proj): #filename, image, geoTransformation, Projection
 drv = gdal.GetDriverByName("GTiff")
 dst_ds = drv.Create(fname, matriz.shape[1], matriz.shape[0], 3, gdal.GDT_UInt16) #Change the 3 by the number of channels you want and gdal.GDT_Uint16 can be changed to type of data you want
 dst_ds.SetGeoTransform(geot)
 dst_ds.SetProjection(proj)
 dst_ds.GetRasterBand(1).WriteArray(matriz[:, :, 0]) 
 dst_ds.GetRasterBand(2).WriteArray(matriz[:, :, 1]) 
 dst_ds.GetRasterBand(3).WriteArray(matriz[:, :, 2])
 dst_ds.FlushCache()
 dst_ds=None

• Marcelo Villa: This creates a 16-bit output