def test_ignore_time_dimension():
# Each timestep should be treated completely
# independent from each other if r['time'] == 0.
ds_nlm = NLMeansFilter(dims=('y', 'x', 'time'), r=(1, 1, 0), sigma=2,
h=2).apply(ds)
t0 = ds.isel(time=0)
t0_nlm = NLMeansFilter(dims=('y', 'x'), r=1, sigma=2, h=2).apply(t0)
assert (np.abs((ds_nlm.isel(time=0) - t0_nlm).to_array()) < 1e-8).all()
def test_nlmeans():
ds_nlm = NLMeansFilter(dims=('y', 'x'), r=0, f=1, sigma=2, h=2).apply(ds)
# Check that the difference between the original data and the filtered
# data is not too large.
assert (np.abs(ds.mean() - ds_nlm.mean()).to_array() < 1e-3).all()
def test_nlmeans_empty_dim():
ds_nlm = NLMeansFilter(dims=(), r=1, f=1, sigma=1, h=1).apply(ds)
assert ds.equals(ds_nlm)
def test_reduce_std():
# Check that applying nlmeans reduces the standard deviation.
ds_nlm = NLMeansFilter(dims=('y', 'x', 'time'), r=(1, 1, 0), sigma=2,
h=2).apply(ds)
assert_all_true(ds_nlm.std() < ds.std())
def test_nlmeans_zero_radius():
# Check that passing r=0 returns the input Dataset unaltered.
ds_nlm = NLMeansFilter(dims=('y', 'x'), r=0, f=1, sigma=1, h=1).apply(ds)
assert ds.equals(ds_nlm)