Gdal sieve filter in python

Question

I want to sieve a classified image in py, and I try to use the gdal sieve function, but it will not work:

Syntax:

SieveFilter(Band srcBand, Band maskBand, Band dstBand, int threshold, int connectedness=4, char ** options=None, GDALProgressFunc callback=0, void * callback_data=None) -> int

My code:

gdal.SieveFilter(1 "C:/X/testX.tif",None,1 "C:/X/testX_sieved.tif",100, 8)

What's the error here?

Answer 1

I do not believe there is a working python binding for the gdal sieve command. I am able to run it successfully by calling it from the command line by using this:

import os, sys, subprocess
sys.path.append(r'C:\Users\*****\AppData\Local\conda\conda\envs\****\Scripts')
gm = os.path.join('C:\\','Users','*****','AppData','Local','conda','conda','envs','****','Scripts','gdal_sieve.py')
sieve_command = ["python", gm,'-st','10','-8','-nomask','-of','GTiff','to_be_sieved.tif','sieved.tif']
subprocess.call(sieve_command,shell=True)

Of course, you'll need to change some paths in the first couple lines to point to the location where your gdal scripts are stored.

Answer 2

To apply the sieve in Python, you must provide raster bands for the first and third arguments. I suspect the error arises because the destination band (third argument) is not writable (i.e. you must open the destination image in read-write mode).

For example, to overwrite the original image:

from osgeo import gdal
Image = gdal.Open('SomeImageName.tif', 1)  # open image in read-write mode
Band = Image.GetRasterBand(1)
gdal.SieveFilter(srcBand=Band, maskBand=None, dstBand=Band, threshold=100, connectedness=8, callback=gdal.TermProgress_nocb)
del Image, Band  # close the datasets.

P.S. This works on Linux, gdal=3.1, python=3.9.