Polarimetric Whitening Filter (PWF)¶

This notebook shows how to apply the PWF on the S matrix.

In [ ]:

%load_ext autoreload
%autoreload 2

import os

import numpy as np
from pathlib import Path
import xarray as xr
from polsarpro.io import open_netcdf_beam
from polsarpro.speckle_filters import pwf

# optional import for progress bar
from dask.diagnostics import ProgressBar

# change to your data paths
# original dataset
input_alos_data = Path("/data/psp/test_files/SAN_FRANCISCO_ALOS1_slc.nc")

# output files from C
output_dir = Path("/data/psp/res/")

%load_ext autoreload %autoreload 2 import os import numpy as np from pathlib import Path import xarray as xr from polsarpro.io import open_netcdf_beam from polsarpro.speckle_filters import pwf # optional import for progress bar from dask.diagnostics import ProgressBar # change to your data paths # original dataset input_alos_data = Path("/data/psp/test_files/SAN_FRANCISCO_ALOS1_slc.nc") # output files from C output_dir = Path("/data/psp/res/")

Load ALOS data and C outputs¶

In [ ]:

S = open_netcdf_beam(input_alos_data)

S = open_netcdf_beam(input_alos_data)

Apply the filter¶

Important note:
this version differs from the C implementation since it allows to smooth the covariance matrix using the test_window_size prior to multiplying it with the inverse covariance estimated using the train_window size parameter. Using values [1, 1] for the test window is equivalent to the original C PolSARpro implementation.

In [ ]:

file_out = output_dir / "pwf.nc"
# netcdf writer cannot overwrite
if os.path.isfile(file_out):
    os.remove(file_out)

with ProgressBar():
    pwf(S, train_window_size=[15, 15], test_window_size=[3, 3]).to_netcdf(file_out)

file_out = output_dir / "pwf.nc" # netcdf writer cannot overwrite if os.path.isfile(file_out): os.remove(file_out) with ProgressBar(): pwf(S, train_window_size=[15, 15], test_window_size=[3, 3]).to_netcdf(file_out)

[########################################] | 100% Completed | 52.22 s

Open the output file¶

In [19]:

res = xr.open_dataset(file_out)

res = xr.open_dataset(file_out)

Display the result¶

In [20]:

# clip the intensity to handle the high dynamic range
res.pwf.plot.imshow(figsize=(7,9), vmax=10)

# clip the intensity to handle the high dynamic range res.pwf.plot.imshow(figsize=(7,9), vmax=10)

Out[20]:

<matplotlib.image.AxesImage at 0x75d66cca19d0>