def test_subband():
# Get a temporary folder
path, id = testing.get_tmp_folder()
try:
# Create example image
arr = np.random.rand(64, 32, 30).astype(np.float32)
meta = dict(metadata="test")
band_info = BandInfo.from_equidistant(30, 400, 700)
tmp = SpectralImage(arr,
band_info=band_info,
meta=meta,
dtype=np.float32,
copy=False)
tmp_subband = tmp.get_subband(5, 11)
assert isinstance(tmp_subband, SpectralImage)
assert tmp_subband.num_channels == 6
assert meta == tmp_subband.meta
assert np.array_equal(band_info.centers[5:11],
tmp_subband.band_info.centers)
finally:
# Cleanup temporary folder
testing.remove_tmp_folder(id)
return
def test_resample():
# Get a temporary folder
path, id = testing.get_tmp_folder()
try:
# Create example image
arr = np.random.rand(64, 32, 30).astype(np.float32)
meta = dict(metadata="test")
band_info = BandInfo.from_equidistant(30, 400, 700)
tmp = SpectralImage(arr,
band_info=band_info,
meta=meta,
dtype=np.float32,
copy=False)
tmp_subband = tmp.get_resampled_spectrum(17)
band_info_down = BandInfo.from_equidistant(17, 400, 700)
assert isinstance(tmp_subband, SpectralImage)
assert tmp_subband.num_channels == 17
assert meta == tmp_subband.meta
assert band_info_down == tmp_subband.band_info
finally:
# Cleanup temporary folder
testing.remove_tmp_folder(id)
return
def test_from_mat():
testing.needs_internet()
# Get a temporary folder
path, id = testing.get_tmp_folder()
try:
path = path / "test_hsi.mat"
hsi_file = testing.get_remote_file(TEST_HSI_FILENAME)
hsi = np.load(hsi_file)
mdict = dict(data=hsi)
scipy.io.savemat(path, mdict)
# Try loading with and without key
for key in [None, 'data']:
tmp = SpectralImage.from_mat_file(path=path, key=key)
assert np.allclose(hsi, tmp)
finally:
# Cleanup temporary folder
testing.remove_tmp_folder(id)
return
def test_spectral_image_new():
"""Test __new__"""
for arr in [hsi_test_rgb, hsi_test_hyper]:
tmp = SpectralImage(arr, dtype=np.float32, copy=True)
assert np.array_equal(arr, tmp)
# Wrong dimension
arr = hsi_test_mono
with raises(DimensionError) as cm:
tmp = SpectralImage(arr, dtype=np.float32, copy=True)
assert f"Expected 3D input. Found 2D." == str(cm.value)
return
def test_from_file():
"""Test from_file() classmethod. Read RGB image."""
testing.needs_internet()
rgb_file = testing.get_remote_file(TEST_RGB_FILENAME)
rgb = (imageio.imread(rgb_file) / 255.0).astype(np.float32)
tmp = SpectralImage.from_file(rgb_file, format='PNG', dtype=np.float32)
assert np.allclose(rgb, tmp)
grey_file = testing.get_remote_file(TEST_GREY_FILENAME)
grey = (imageio.imread(grey_file) / 255.0).astype(np.float32)
assert grey.ndim == 2
tmp = SpectralImage.from_file(grey_file, format='PNG', dtype=np.float32)
assert tmp.ndim == 3
return
def test_save_rgb():
# Get a temporary folder
path, id = testing.get_tmp_folder()
try:
# Create example image
arr = np.random.rand(64, 32, 30).astype(np.float32)
tmp = SpectralImage(arr, dtype=np.float32, copy=False)
tmp_rgb = tmp.get_rgb()
rgb_filename = path / "test_rgb_img.png"
tmp.save_rgb(rgb_filename)
# Load image
loaded = SpectralImage.from_file(path=rgb_filename, dtype=np.float32)
# Caution: lossy dtype conversion
assert np.allclose(tmp_rgb, loaded, atol=0.01)
finally:
# Cleanup temporary folder
testing.remove_tmp_folder(id)
return
def test_resize():
# Create example image
s, t, ch = 32, 64, 3
arr = np.random.rand(s, t, ch).astype(np.float32)
tmp = SpectralImage(arr, dtype=np.float32, copy=False)
output_shape = (16, 16)
tmp_r = tmp.get_resized(output_shape)
assert tmp_r.shape == (16, 16, 3)
output_shape = (32, 64)
tmp_r = tmp.get_resized(output_shape)
assert tmp_r.shape == (32, 64, 3)
output_shape = (16, 32)
tmp_r = tmp.get_resized(output_shape)
assert tmp_r.shape == (16, 32, 3)
# Create example multispectral light field
s, t, ch = 32, 64, 13
arr = np.random.rand(s, t, ch).astype(np.float32)
tmp = SpectralImage(arr, dtype=np.float32, copy=False)
output_shape = (16, 16)
tmp_r = tmp.get_resized(output_shape)
assert tmp_r.shape == (16, 16, 13)
output_shape = (32, 64)
tmp_r = tmp.get_resized(output_shape)
assert tmp_r.shape == (32, 64, 13)
output_shape = (16, 32)
tmp_r = tmp.get_resized(output_shape)
assert tmp_r.shape == (16, 32, 13)
return
def test_from_file_collection():
testing.needs_internet()
for i in range(1, 32):
fname = TEST_FILE_COLLECTION + f"balloons_ms_{i:02d}.png"
testing.get_remote_file(fname)
# Read from file collection
path = testing.appdata_dir(APPNAME) / TEST_FILE_COLLECTION
tmp = SpectralImage.from_file_collection(path=path)
hsi_file = testing.get_remote_file(TEST_HSI_FILENAME)
hsi = np.load(hsi_file)
assert np.allclose(hsi, tmp)
return
def test_rescale():
# Create example image
s, t, ch = 32, 64, 3
arr = np.random.rand(s, t, ch).astype(np.float32)
tmp = SpectralImage(arr, dtype=np.float32, copy=False)
scale = 0.5
tmp_r = tmp.get_rescaled(scale)
assert tmp_r.shape == (16, 32, 3)
scale = 2
tmp_r = tmp.get_rescaled(scale)
assert tmp_r.shape == (64, 128, 3)
scale = (2, 1)
tmp_r = tmp.get_rescaled(scale)
assert tmp_r.shape == (64, 64, 3)
scale = (2, 0.5)
tmp_r = tmp.get_rescaled(scale)
assert tmp_r.shape == (64, 32, 3)
# Create example multispectral light field
s, t, ch = 32, 64, 13
arr = np.random.rand(s, t, ch).astype(np.float32)
tmp = SpectralImage(arr, dtype=np.float32, copy=False)
scale = 0.5
tmp_r = tmp.get_rescaled(scale)
assert tmp_r.shape == (16, 32, 13)
scale = 2
tmp_r = tmp.get_rescaled(scale)
assert tmp_r.shape == (64, 128, 13)
scale = (2, 1)
tmp_r = tmp.get_rescaled(scale)
assert tmp_r.shape == (64, 64, 13)
scale = (2, 0.5)
tmp_r = tmp.get_rescaled(scale)
assert tmp_r.shape == (64, 32, 13)
return
def test_save():
# Get a temporary folder
path, id = testing.get_tmp_folder()
try:
# Create example image
arr = np.random.rand(64, 32, 30).astype(np.float32)
tmp = SpectralImage(arr, dtype=np.float32, copy=False)
file = path / "test_img.png"
for d in [np.uint8, np.uint16]:
tmp.save(path=file, dtype=d)
# Load image
loaded = SpectralImage.from_file_collection(path=file.parent,
dtype=np.float32)
# Caution: lossy dtype conversion
assert np.allclose(arr, loaded, atol=0.01)
# Test error handling
with raises(ValueError) as cm:
file = path / "test_img"
tmp.save(path=file)
assert f"Path needs to include an extension." == str(cm.value)
with raises(FileNotFoundError) as cm:
file = path / "does_not_exist/test_img.png"
tmp.save(path=file)
assert f"Path '{file.parent}' does not exist." == str(cm.value)
tmp.save(path=file, create_dir=True, dtype=np.uint16)
# Load image
loaded = SpectralImage.from_file_collection(path=file.parent,
dtype=np.float32)
# Caution: lossy dtype conversion
assert np.allclose(arr, loaded, atol=0.01)
finally:
# Cleanup temporary folder
testing.remove_tmp_folder(id)
return