def test_default_sigma():
a = np.zeros((3, 3))
a[1, 1] = 1.
assert_array_equal(
gaussian(a, preserve_range=True),
gaussian(a, preserve_range=True, sigma=1)
)
def test_multichannel(channel_axis):
a = np.zeros((5, 5, 3))
a[1, 1] = np.arange(1, 4)
a = np.moveaxis(a, -1, channel_axis)
gaussian_rgb_a = gaussian(a, sigma=1, mode='reflect', preserve_range=True,
channel_axis=channel_axis)
# Check that the mean value is conserved in each channel
# (color channels are not mixed together)
spatial_axes = tuple(
[ax for ax in range(a.ndim) if ax != channel_axis % a.ndim]
)
assert np.allclose(a.mean(axis=spatial_axes),
gaussian_rgb_a.mean(axis=spatial_axes))
if channel_axis % a.ndim == 2:
# Test legacy behavior equivalent to old (multichannel = None)
with expected_warnings(['multichannel']):
gaussian_rgb_a = gaussian(a, sigma=1, mode='reflect',
preserve_range=True)
# Check that the mean value is conserved in each channel
# (color channels are not mixed together)
assert np.allclose(a.mean(axis=spatial_axes),
gaussian_rgb_a.mean(axis=spatial_axes))
# Iterable sigma
gaussian_rgb_a = gaussian(a, sigma=[1, 2], mode='reflect',
channel_axis=channel_axis,
preserve_range=True)
assert np.allclose(a.mean(axis=spatial_axes),
gaussian_rgb_a.mean(axis=spatial_axes))
def test_difference_of_gaussians(s, s2):
image = np.random.rand(10, 10)
im1 = gaussian(image, s, preserve_range=True)
im2 = gaussian(image, s2, preserve_range=True)
dog = im1 - im2
dog2 = difference_of_gaussians(image, s, s2)
assert np.allclose(dog, dog2)
def test_auto_sigma2(s):
image = np.random.rand(10, 10)
im1 = gaussian(image, s, preserve_range=True)
s2 = 1.6 * np.array(s)
im2 = gaussian(image, s2, preserve_range=True)
dog = im1 - im2
dog2 = difference_of_gaussians(image, s, s2)
assert np.allclose(dog, dog2)
def test_preserve_range():
"""Test preserve_range parameter."""
ones = np.ones((2, 2), dtype=np.int64)
filtered_ones = gaussian(ones, preserve_range=False)
assert np.all(filtered_ones == filtered_ones[0, 0])
assert filtered_ones[0, 0] < 1e-10
filtered_preserved = gaussian(ones, preserve_range=True)
assert np.all(filtered_preserved == 1.)
img = np.array([[10.0, -10.0], [-4, 3]], dtype=np.float32)
gaussian(img, 1)
def test_deprecated_multichannel():
a = np.zeros((5, 5, 3))
a[1, 1] = np.arange(1, 4)
with expected_warnings(["`multichannel` is a deprecated argument"]):
gaussian_rgb_a = gaussian(a, sigma=1, mode='reflect',
multichannel=True)
# Check that the mean value is conserved in each channel
# (color channels are not mixed together)
assert np.allclose(a.mean(axis=(0, 1)), gaussian_rgb_a.mean(axis=(0, 1)))
# check positional multichannel argument warning
with expected_warnings(["Providing the `multichannel` argument"]):
gaussian_rgb_a = gaussian(a, 1, None, 'reflect', 0, True)
def test_1d_ok():
"""Testing Gaussian Filter for 1D array.
With any array consisting of positive integers and only one zero - it
should filter all values to be greater than 0.1
"""
nums = np.arange(7)
filtered = gaussian(nums, preserve_range=True)
assert np.all(filtered > 0.1)
def test_preserve_output(dtype):
image = np.arange(9, dtype=dtype).reshape((3, 3))
output = np.zeros_like(image, dtype=dtype)
gaussian_image = gaussian(image,
sigma=1,
output=output,
preserve_range=True)
assert gaussian_image is output
def test_multichannel():
a = np.zeros((5, 5, 3))
a[1, 1] = np.arange(1, 4)
gaussian_rgb_a = gaussian(a, sigma=1, mode='reflect',
multichannel=True)
# Check that the mean value is conserved in each channel
# (color channels are not mixed together)
assert np.allclose([a[..., i].mean() for i in range(3)],
[gaussian_rgb_a[..., i].mean() for i in range(3)])
# Test multichannel = None
with expected_warnings(['multichannel']):
gaussian_rgb_a = gaussian(a, sigma=1, mode='reflect')
# Check that the mean value is conserved in each channel
# (color channels are not mixed together)
assert np.allclose([a[..., i].mean() for i in range(3)],
[gaussian_rgb_a[..., i].mean() for i in range(3)])
# Iterable sigma
gaussian_rgb_a = gaussian(a, sigma=[1, 2], mode='reflect',
multichannel=True)
assert np.allclose([a[..., i].mean() for i in range(3)],
[gaussian_rgb_a[..., i].mean() for i in range(3)])
def test_multichannel():
a = np.zeros((5, 5, 3))
a[1, 1] = np.arange(1, 4)
gaussian_rgb_a = gaussian(a, sigma=1, mode='reflect',
multichannel=True)
# Check that the mean value is conserved in each channel
# (color channels are not mixed together)
assert np.allclose([a[..., i].mean() for i in range(3)],
[gaussian_rgb_a[..., i].mean() for i in range(3)])
# Test multichannel = None
with expected_warnings(['multichannel']):
gaussian_rgb_a = gaussian(a, sigma=1, mode='reflect')
# Check that the mean value is conserved in each channel
# (color channels are not mixed together)
assert np.allclose([a[..., i].mean() for i in range(3)],
[gaussian_rgb_a[..., i].mean() for i in range(3)])
# Iterable sigma
gaussian_rgb_a = gaussian(a, sigma=[1, 2], mode='reflect',
multichannel=True)
assert np.allclose([a[..., i].mean() for i in range(3)],
[gaussian_rgb_a[..., i].mean() for i in range(3)])
def test_negative_sigma():
a = np.zeros((3, 3))
a[1, 1] = 1.
with pytest.raises(ValueError):
gaussian(a, sigma=-1.0)
with pytest.raises(ValueError):
gaussian(a, sigma=[-1.0, 1.0])
with pytest.raises(ValueError):
gaussian(a, sigma=np.asarray([-1.0, 1.0]))
def test_negative_sigma():
a = np.zeros((3, 3))
a[1, 1] = 1.
with testing.raises(ValueError):
gaussian(a, sigma=-1.0)
with testing.raises(ValueError):
gaussian(a, sigma=[-1.0, 1.0])
with testing.raises(ValueError):
gaussian(a,
sigma=np.asarray([-1.0, 1.0]))
def ConvertSkech(input_path, output_path):
extensions = ['jpg']
file_list = []
for extension in extensions:
file_glob = os.path.join(input_path, '*.' + extension)
# get all the images of one folder of the input_path
file_list.extend(glob.glob(file_glob))
for image in file_list:
img = io.imread(image)
img = img_as_float(img)
I_out = (img[:, :, 0] + img[:, :, 1] + img[:, :, 2]) / 3.0
I_invert = 1.0 - I_out
I_gauss = gaussian(I_invert, sigma=5)
delta = 0.0001
I_dodge = (I_out + delta) / (1 + delta - I_gauss)
th = 0.95
mask = I_dodge > th
max_val = I_dodge.max()
I_dodge = I_dodge * (1 - mask) + (th + I_dodge / max_val * (1 - th)) * mask
img_out = I_dodge.copy()
image_name = os.path.basename(os.path.splitext(image)[0])
image_suffix = os.path.splitext(image)[1]
scipy.misc.toimage(img_out).save(output_path + image_name + image_suffix)
def test_energy_decrease():
a = np.zeros((3, 3))
a[1, 1] = 1.
gaussian_a = gaussian(a, sigma=1, mode='reflect')
assert gaussian_a.std() < a.std()
def test_default_sigma():
a = np.zeros((3, 3))
a[1, 1] = 1.
assert np.all(gaussian(a) == gaussian(a, sigma=1))
def test_null_sigma():
a = np.zeros((3, 3))
a[1, 1] = 1.
assert np.all(gaussian(a, 0) == a)
def test_preserve_range():
img = np.array([[10.0, -10.0], [-4, 3]], dtype=np.float32)
gaussian(img, 1, preserve_range=True)
def test_4d_ok():
img = np.zeros((5,) * 4)
img[2, 2, 2, 2] = 1
res = gaussian(img, 1, mode='reflect')
assert np.allclose(res.sum(), 1)
def blur(image):
return gaussian(image.astype(float), sigma=2)
def test_null_sigma():
a = np.zeros((3, 3))
a[1, 1] = 1.
assert np.all(gaussian(a, 0, preserve_range=True) == a)
def test_null_sigma():
a = np.zeros((3, 3))
a[1, 1] = 1.
assert np.all(gaussian(a, 0) == a)
def test_4d_ok():
img = np.zeros((5,) * 4)
img[2, 2, 2, 2] = 1
res = gaussian(img, 1, mode='reflect')
assert np.allclose(res.sum(), 1)
def test_default_sigma():
a = np.zeros((3, 3))
a[1, 1] = 1.
assert np.all(gaussian(a) == gaussian(a, sigma=1))
def test_energy_decrease():
a = np.zeros((3, 3))
a[1, 1] = 1.
gaussian_a = gaussian(a, sigma=1, mode='reflect')
assert gaussian_a.std() < a.std()
def test_preserve_range():
img = np.array([[10.0, -10.0], [-4, 3]], dtype=np.float32)
gaussian(img, 1, preserve_range=True)
def test_output_error():
image = np.arange(9, dtype=np.float32).reshape((3, 3))
output = np.zeros_like(image, dtype=np.uint8)
with pytest.raises(ValueError):
gaussian(image, sigma=1, output=output,
preserve_range=True)
from skimage import img_as_float
import matplotlib.pyplot as plt
from skimage import io
from skimage.filters._gaussian import gaussian
file_name='D:/Visual Effects/PS Algorithm/4.jpg'
img=io.imread(file_name)
img = img_as_float(img)
I_out = (img[:, :, 0] + img[:, :, 1] + img[:, :, 2]) / 3.0
I_invert = 1.0 - I_out
I_gauss = gaussian(I_invert, sigma=5)
delta = 0.0001
I_dodge = (I_out + delta) / (1 + delta - I_gauss)
th = 0.95
mask = I_dodge > th
max_val = I_dodge.max()
I_dodge = I_dodge * (1-mask) + (th + I_dodge / max_val * (1 - th)) * mask
img_out = I_dodge.copy()
plt.figure(1)
plt.imshow(img)
plt.axis('off')
plt.figure(2)
from skimage import img_as_float
import matplotlib.pyplot as plt
from skimage import io
from skimage.filters._gaussian import gaussian
file_name = 'D:/Visual Effects/PS Algorithm/4.jpg'
img = io.imread(file_name)
img = img_as_float(img)
img_out = gaussian(img, sigma=3, multichannel=True)
plt.figure(1)
plt.imshow(img)
plt.axis('off')
plt.figure(2)
plt.imshow(img_out)
plt.axis('off')
