Filter API

Filter API#

class serket.image.AvgBlur2D(kernel_size)[source]#

Average blur 2D layer.

Parameters:

kernel_size (int | tuple[int, int]) – size of the convolving kernel.
dtype – data type of the layer. float32

Example

>>> import serket as sk
>>> import jax.numpy as jnp
>>> layer = sk.image.AvgBlur2D(kernel_size=3)
>>> print(layer(jnp.ones((1, 5, 5))))  
[[[0.44444448 0.6666667  0.6666667  0.6666667  0.44444448]
  [0.6666667  1.         1.         1.         0.6666667 ]
  [0.6666667  1.         1.         1.         0.6666667 ]
  [0.6666667  1.         1.         1.         0.6666667 ]
  [0.44444448 0.6666667  0.6666667  0.6666667  0.44444448]]]

__call__(image)#

Call self as a function.

Parameters:: image (Array)
Return type:: Array

class serket.image.BilateralBlur2D(kernel_size, *, sigma_space, sigma_color)[source]#

Apply bilateral blur to a channel-first image.

Parameters:

kernel_size (int | tuple[int, int]) – kernel size. accepts int or tuple of two ints.
sigma_space (float | tuple[float, float]) – sigma in the coordinate space. accepts float or tuple of two floats.
sigma_color (float) – sigma in the color space. accepts float.

Example

>>> import serket as sk
>>> import jax.numpy as jnp
>>> x = jnp.ones([1, 5, 5])
>>> layer = sk.image.BilateralBlur2D((3, 5), sigma_space=(1.2, 1.3), sigma_color=1.5)
>>> print(layer(x))  
[[[0.5231399  0.6869784  0.75100434 0.6869784  0.5231399 ]
  [0.70914114 0.9193193  1.         0.9193192  0.70914114]
  [0.70914114 0.9193193  1.         0.9193192  0.70914114]
  [0.70914114 0.9193193  1.         0.9193192  0.70914114]
  [0.5231399  0.6869784  0.75100434 0.6869784  0.5231399 ]]]

__call__(image)[source]#

Call self as a function.

Parameters:: image (Array)
Return type:: Array

class serket.image.BlurPool2D(kernel_size, strides)[source]#

Blur and downsample a channel-first image.

Parameters:

kernel_size (int | tuple[int, int]) – kernel size. accepts int or tuple of two ints.
strides (int | tuple[int, int]) – strides. accepts int or tuple of two ints.

Example

>>> import serket as sk
>>> import jax.numpy as jnp
>>> x = jnp.arange(1, 26).reshape(1, 5, 5)
>>> layer = sk.image.BlurPool2D(kernel_size=3, strides=2)
>>> print(layer(x))  
[[[ 1.6875  3.5     3.5625]
  [ 8.5    13.     11.    ]
  [11.0625 16.     12.9375]]]

__call__(image)#

Call self as a function.

Parameters:: image (Array)
Return type:: Array

class serket.image.BoxBlur2D(kernel_size)[source]#

Box blur 2D layer.

Parameters:

kernel_size (int | tuple[int, int]) – size of the convolving kernel. Accepts int or tuple of two ints.
dtype – data type of the layer. float32

Example

>>> import serket as sk
>>> import jax.numpy as jnp
>>> layer = sk.image.BoxBlur2D((3, 5))
>>> print(layer(jnp.ones((1, 5, 5))))  
[[[0.40000004 0.53333336 0.6666667  0.53333336 0.40000004]
  [0.6        0.8        1.         0.8        0.6       ]
  [0.6        0.8        1.         0.8        0.6       ]
  [0.6        0.8        1.         0.8        0.6       ]
  [0.40000004 0.53333336 0.6666667  0.53333336 0.40000004]]]

__call__(image)#

Call self as a function.

Parameters:: image (Array)
Return type:: Array

class serket.image.ElasticTransform2D(kernel_size, sigma, alpha)[source]#

Apply an elastic transform to an image.

Parameters:

kernel_size (int | tuple[int, int]) – The size of the Gaussian kernel. either a single integer or a tuple of two integers.
sigma (float | tuple[float, float]) – The standard deviation of the Gaussian in the y and x directions,
alpha (float | tuple[float, float]) – The scaling factor that controls the intensity of the deformation in the y and x directions, respectively.

Example

>>> import serket as sk
>>> import jax.random as jr
>>> import jax.numpy as jnp
>>> layer = sk.image.ElasticTransform2D(kernel_size=3, sigma=1.0, alpha=1.0)
>>> key = jr.key(0)
>>> image = jnp.arange(1, 26).reshape(1, 5, 5).astype(jnp.float32)
>>> print(layer(image, key=key))  
[[[ 1.0669159  2.2596366  3.210071   3.9703817  4.9207525]
  [ 5.70821    7.483665   8.857002   8.663773   8.794132 ]
  [13.809857  15.865877  15.109764  12.897442  13.0018215]
  [18.35189   18.817993  17.2193    15.731948  17.026705 ]
  [21.        21.659977  21.43855   21.138866  22.583244 ]]]

__call__(image, *, key)#

Call self as a function.

Parameters:

image (Array)
key (Array)

Return type:

Array

class serket.image.FFTAvgBlur2D(kernel_size)[source]#

Average blur 2D layer using FFT.

Parameters:

kernel_size (int | tuple[int, int]) – size of the convolving kernel.
dtype – data type of the layer. float32

Example

>>> import serket as sk
>>> import jax.numpy as jnp
>>> layer = sk.image.FFTAvgBlur2D(kernel_size=3)
>>> print(layer(jnp.ones((1, 5, 5))))  
[[[0.44444448 0.6666667  0.6666667  0.6666667  0.44444448]
  [0.6666667  1.         1.         1.         0.6666667 ]
  [0.6666667  1.         1.         1.         0.6666667 ]
  [0.6666667  1.         1.         1.         0.6666667 ]
  [0.44444448 0.6666667  0.6666667  0.6666667  0.44444448]]]

__call__(image)#

Call self as a function.

Parameters:: image (Array)
Return type:: Array

class serket.image.FFTBlurPool2D(kernel_size, strides)[source]#

Blur and downsample a channel-first image using FFT.

Parameters:

kernel_size (int | tuple[int, int]) – kernel size. accepts int or tuple of two ints.
strides (int | tuple[int, int]) – stride. accepts int or tuple of two ints.

Example

>>> import serket as sk
>>> import jax.numpy as jnp
>>> x = jnp.arange(1, 26).reshape(1, 5, 5)
>>> layer = sk.image.FFTBlurPool2D(kernel_size=3, strides=2)
>>> print(layer(x))  
[[[ 1.6875  3.5     3.5625]
  [ 8.5    13.     11.    ]
  [11.0625 16.     12.9375]]]

__call__(image)#

Call self as a function.

Parameters:: image (Array)
Return type:: Array

class serket.image.FFTBoxBlur2D(kernel_size)[source]#

Box blur 2D layer using FFT.

Parameters:

kernel_size (int | tuple[int, int]) – size of the convolving kernel. Accepts int or tuple of two ints.
dtype – data type of the layer. float32

Example

>>> import serket as sk
>>> import jax.numpy as jnp
>>> layer = sk.image.BoxBlur2D((3, 5))
>>> print(layer(jnp.ones((1, 5, 5))))  
[[[0.40000004 0.53333336 0.6666667  0.53333336 0.40000004]
  [0.6        0.8        1.         0.8        0.6       ]
  [0.6        0.8        1.         0.8        0.6       ]
  [0.6        0.8        1.         0.8        0.6       ]
  [0.40000004 0.53333336 0.6666667  0.53333336 0.40000004]]]

__call__(image)#

Call self as a function.

Parameters:: image (Array)
Return type:: Array

class serket.image.FFTElasticTransform2D(kernel_size, sigma, alpha)[source]#

Apply an elastic transform to an image using FFT.

Parameters:

kernel_size (int | tuple[int, int]) – The size of the Gaussian kernel. either a single integer or a tuple of two integers.
sigma (float | tuple[float, float]) – The standard deviation of the Gaussian in the y and x directions,
alpha (float | tuple[float, float]) – The scaling factor that controls the intensity of the deformation in the y and x directions, respectively.

Example

>>> import serket as sk
>>> import jax.random as jr
>>> import jax.numpy as jnp
>>> layer = sk.image.FFTElasticTransform2D(kernel_size=3, sigma=1.0, alpha=1.0)
>>> key = jr.key(0)
>>> image = jnp.arange(1, 26).reshape(1, 5, 5).astype(jnp.float32)
>>> print(layer(image, key=key))  
[[[ 1.0669159  2.2596366  3.210071   3.9703817  4.9207525]
  [ 5.70821    7.483665   8.857002   8.663773   8.794132 ]
  [13.809857  15.865877  15.109764  12.897442  13.0018215]
  [18.35189   18.817993  17.2193    15.731948  17.026705 ]
  [21.        21.659977  21.43855   21.138866  22.583244 ]]]

__call__(image, *, key)#

Call self as a function.

Parameters:

image (Array)
key (Array)

Return type:

Array

class serket.image.FFTGaussianBlur2D(kernel_size, *, sigma=1.0)[source]#

Apply Gaussian blur to a channel-first image using FFT.

Parameters:

kernel_size (int | tuple[int, int]) – kernel size. accepts int or tuple of two ints.
sigma (float | tuple[float, float]) – sigma. Defaults to 1. accepts float or tuple of two floats.
dtype – data type of the layer. float32

Example

>>> import serket as sk
>>> import jax.numpy as jnp
>>> layer = sk.image.FFTGaussianBlur2D(kernel_size=3)
>>> print(layer(jnp.ones((1, 5, 5))))  
[[[0.5269764 0.7259314 0.7259314 0.7259314 0.5269764]
  [0.7259314 1.        1.        1.        0.7259314]
  [0.7259314 1.        1.        1.        0.7259314]
  [0.7259314 1.        1.        1.        0.7259314]
  [0.5269764 0.7259314 0.7259314 0.7259314 0.5269764]]]

__call__(image)#

Call self as a function.

Parameters:: image (Array)
Return type:: Array

class serket.image.FFTLaplacian2D(kernel_size)[source]#

Apply Laplacian filter to a channel-first image using FFT.

Parameters:

kernel_size (int | tuple[int, int]) – size of the convolving kernel. Accepts int or tuple of two ints.
dtype – data type of the layer. float32

Example

>>> import serket as sk
>>> import jax.numpy as jnp
>>> layer = sk.image.FFTLaplacian2D(kernel_size=(3, 5))
>>> print(layer(jnp.ones((1, 5, 5))))  
[[[-9. -7. -5. -7. -9.]
  [-6. -3.  0. -3. -6.]
  [-6. -3.  0. -3. -6.]
  [-6. -3.  0. -3. -6.]
  [-9. -7. -5. -7. -9.]]]

Note

The laplacian considers all the neighbors of a pixel.

__call__(image)#

Call self as a function.

Parameters:: image (Array)
Return type:: Array

class serket.image.FFTMotionBlur2D(kernel_size, *, angle=0.0, direction=0.0)[source]#

Apply motion blur to a channel-first image using FFT.

Parameters:

kernel_size (int) – motion kernel width and height. It should be odd and positive.
angle (float) – angle of the motion blur in degrees (anti-clockwise rotation).
direction (float) – direction of the motion blur.
dtype – data type of the layer. float32

Example

>>> import serket as sk
>>> import jax.numpy as jnp
>>> x = jnp.arange(1, 17).reshape(1, 4, 4) + 0.0
>>> print(sk.image.MotionBlur2D(3, angle=30, direction=0.5)(x))  
[[[ 0.7827108  2.4696379  3.3715053  3.8119273]
  [ 2.8356633  6.3387947  7.3387947  7.1810846]
  [ 5.117592  10.338796  11.338796  10.550241 ]
  [ 6.472714  10.020969  10.770187   9.100007 ]]]

__call__(image)#

Call self as a function.

Parameters:: image (Array)
Return type:: Array

class serket.image.FFTSobel2D(*a, **k)[source]#

Apply Sobel filter to a channel-first image using FFT.

Example

>>> import serket as sk
>>> import jax.numpy as jnp
>>> x = jnp.arange(1,26).reshape(1, 5,5).astype(jnp.float32)
>>> layer = sk.image.FFTSobel2D()
>>> layer(x)  
[[[21.954498, 28.635643, 32.55764 , 36.496574, 33.61547 ],
  [41.036568, 40.792156, 40.792156, 40.792156, 46.8615  ],
  [56.603886, 40.792156, 40.792156, 40.792156, 63.529522],
  [74.323616, 40.792156, 40.792156, 40.792156, 81.706795],
  [78.24321 , 68.26419 , 72.249565, 76.23647 , 89.27486 ]]]

__call__(image)#

Call self as a function.

Parameters:: image (Array)
Return type:: Array

class serket.image.FFTUnsharpMask2D(kernel_size, *, sigma=1.0)[source]#

Apply unsharp mask to a channel-first image using FFT.

Parameters:

kernel_size (int | tuple[int, int]) – kernel size. accepts int or tuple of two ints.
sigma (float | tuple[float, float]) – sigma. Defaults to 1. accepts float or tuple of two floats.
dtype – data type of the layer. float32

Example

>>> import serket as sk
>>> import jax.numpy as jnp
>>> layer = sk.image.FFTUnsharpMask2D(kernel_size=3)
>>> print(layer(jnp.ones((1, 5, 5))))  
[[[1.4730237 1.2740686 1.2740686 1.2740686 1.4730237]
  [1.2740686 1.        1.        1.        1.2740686]
  [1.2740686 1.        1.        1.        1.2740686]
  [1.2740686 1.        1.        1.        1.2740686]
  [1.4730237 1.2740686 1.2740686 1.2740686 1.4730237]]]

__call__(image)#

Call self as a function.

Parameters:: image (Array)
Return type:: Array

class serket.image.GaussianBlur2D(kernel_size, *, sigma=1.0)[source]#

Apply Gaussian blur to a channel-first image.

Parameters:

kernel_size (int | tuple[int, int]) – kernel size. accepts int or tuple of two ints.
sigma (float | tuple[float, float]) – sigma. Defaults to 1. accepts float or tuple of two floats.
dtype – data type of the layer. float32

Example

>>> import serket as sk
>>> import jax.numpy as jnp
>>> layer = sk.image.GaussianBlur2D(kernel_size=3)
>>> print(layer(jnp.ones((1, 5, 5))))  
[[[0.5269764 0.7259314 0.7259314 0.7259314 0.5269764]
  [0.7259314 1.        1.        1.        0.7259314]
  [0.7259314 1.        1.        1.        0.7259314]
  [0.7259314 1.        1.        1.        0.7259314]
  [0.5269764 0.7259314 0.7259314 0.7259314 0.5269764]]]

__call__(image)#

Call self as a function.

Parameters:: image (Array)
Return type:: Array

class serket.image.JointBilateralBlur2D(kernel_size, *, sigma_space, sigma_color)[source]#

Apply joint bilateral blur to a channel-first image.

Parameters:

kernel_size (int | tuple[int, int]) – kernel size. accepts int or tuple of two ints.
sigma_space (float | tuple[float, float]) – sigma in the coordinate space. accepts float or tuple of two floats.
sigma_color (float) – sigma in the color space. accepts float.

Example

>>> import serket as sk
>>> import jax.numpy as jnp
>>> x = jnp.ones([1, 5, 5])
>>> guide = jnp.ones([1, 5, 5])
>>> layer = sk.image.JointBilateralBlur2D((3, 5), sigma_space=(1.2, 1.3), sigma_color=1.5)
>>> print(layer(x, guide))  
[[[0.5231399  0.6869784  0.75100434 0.6869784  0.5231399 ]
  [0.70914114 0.9193193  1.         0.9193192  0.70914114]
  [0.70914114 0.9193193  1.         0.9193192  0.70914114]
  [0.70914114 0.9193193  1.         0.9193192  0.70914114]
  [0.5231399  0.6869784  0.75100434 0.6869784  0.5231399 ]]]

__call__(image, guide)[source]#

Apply joint bilateral blur to a channel-first image.

Parameters:

image (Array) – input image.
guide (Array) – guide image used for computing the gaussian for color space.

Return type:

Array

class serket.image.Laplacian2D(kernel_size)[source]#

Apply Laplacian filter to a channel-first image.

Parameters:

kernel_size (int | tuple[int, int]) – size of the convolving kernel. Accepts int or tuple of two ints.
dtype – data type of the layer. float32

Example

>>> import serket as sk
>>> import jax.numpy as jnp
>>> layer = sk.image.Laplacian2D(kernel_size=(3, 5))
>>> print(layer(jnp.ones((1, 5, 5))))  
[[[-9. -7. -5. -7. -9.]
  [-6. -3.  0. -3. -6.]
  [-6. -3.  0. -3. -6.]
  [-6. -3.  0. -3. -6.]
  [-9. -7. -5. -7. -9.]]]

Note

The laplacian considers all the neighbors of a pixel.

__call__(image)#

Call self as a function.

Parameters:: image (Array)
Return type:: Array

class serket.image.MedianBlur2D(kernel_size)[source]#

Apply median filter to a channel-first image.

Parameters:: kernel_size (int | tuple[int, int]) – size of the convolving kernel. Accepts int or tuple of two ints.

Example

>>> import serket as sk
>>> import jax.numpy as jnp
>>> x = jnp.arange(1, 26).reshape(1, 5, 5) + 0.0
>>> print(x)
[[[ 1.  2.  3.  4.  5.]
  [ 6.  7.  8.  9. 10.]
  [11. 12. 13. 14. 15.]
  [16. 17. 18. 19. 20.]
  [21. 22. 23. 24. 25.]]]
>>> print(sk.image.MedianBlur2D(3)(x))
[[[ 0.  2.  3.  4.  0.]
  [ 2.  7.  8.  9.  5.]
  [ 7. 12. 13. 14. 10.]
  [12. 17. 18. 19. 15.]
  [ 0. 17. 18. 19.  0.]]]

__call__(image)[source]#

Call self as a function.

Parameters:: image (Array)
Return type:: Array

class serket.image.MotionBlur2D(kernel_size, *, angle=0.0, direction=0.0)[source]#

Apply motion blur to a channel-first image.

Parameters:

kernel_size (int) – motion kernel width and height. It should be odd and positive.
angle (float) – angle of the motion blur in degrees (anti-clockwise rotation).
direction (float) – direction of the motion blur.
dtype – data type of the layer. float32

Example

>>> import serket as sk
>>> import jax.numpy as jnp
>>> x = jnp.arange(1, 17).reshape(1, 4, 4) + 0.0
>>> print(sk.image.MotionBlur2D(3, angle=30, direction=0.5)(x))  
[[[ 0.7827108  2.4696379  3.3715053  3.8119273]
  [ 2.8356633  6.3387947  7.3387947  7.1810846]
  [ 5.117592  10.338796  11.338796  10.550241 ]
  [ 6.472714  10.020969  10.770187   9.100007 ]]]

__call__(image)#

Call self as a function.

Parameters:: image (Array)
Return type:: Array

class serket.image.Sobel2D(*a, **k)[source]#

Apply Sobel filter to a channel-first image.

Parameters:: dtype – data type of the layer. Defaults to float32.

Example

>>> import serket as sk
>>> import jax.numpy as jnp
>>> x = jnp.arange(1,26).reshape(1, 5,5).astype(jnp.float32)
>>> layer = sk.image.Sobel2D()
>>> layer(x)  
[[[21.954498, 28.635643, 32.55764 , 36.496574, 33.61547 ],
  [41.036568, 40.792156, 40.792156, 40.792156, 46.8615  ],
  [56.603886, 40.792156, 40.792156, 40.792156, 63.529522],
  [74.323616, 40.792156, 40.792156, 40.792156, 81.706795],
  [78.24321 , 68.26419 , 72.249565, 76.23647 , 89.27486 ]]]

__call__(image)#

Call self as a function.

Parameters:: image (Array)
Return type:: Array

class serket.image.UnsharpMask2D(kernel_size, *, sigma=1.0)[source]#

Apply unsharp mask to a channel-first image.

Parameters:

kernel_size (int | tuple[int, int]) – kernel size. accepts int or tuple of two ints.
sigma (float | tuple[float, float]) – sigma. Defaults to 1. accepts float or tuple of two floats.
dtype – data type of the layer. float32

Example

>>> import serket as sk
>>> import jax.numpy as jnp
>>> layer = sk.image.UnsharpMask2D(kernel_size=3)
>>> print(layer(jnp.ones((1, 5, 5))))  
[[[1.4730237 1.2740686 1.2740686 1.2740686 1.4730237]
  [1.2740686 1.        1.        1.        1.2740686]
  [1.2740686 1.        1.        1.        1.2740686]
  [1.2740686 1.        1.        1.        1.2740686]
  [1.4730237 1.2740686 1.2740686 1.2740686 1.4730237]]]

__call__(image)#

Call self as a function.

Parameters:: image (Array)
Return type:: Array

serket.image.avg_blur_2d(image, kernel_size, dtype=None)[source]#

Average blur.

Parameters:

image (Array) – 2D input array. shape is (height, width).
kernel_size (tuple[int, int]) – size of the convolving kernel. Accepts tuple of two ints.
dtype (Union[dtype, str, Any, None]) – data type of the kernel. Defaults to None to use the same data type as array.

Return type:

Array

Returns:

Average blurred array. shape is (height, width).

serket.image.bilateral_blur_2d(image, kernel_size, sigma_space, sigma_color, dtype=None)[source]#

Bilateral blur.

Parameters:

image (Array) – 2D input array. shape is (height, width).
kernel_size (tuple[int, int]) – size of the convolving kernel. Accepts tuple of two ints.
sigma_space (tuple[float, float]) – sigma of gaussian kernel.
sigma_color (float) – sigma of gaussian kernel.
dtype (Union[dtype, str, Any, None]) – data type of the kernel. Defaults to None to use the same data type as array.

serket.image.blur_pool_2d(image, kernel_size, strides)[source]#

Blur pooling see https://arxiv.org/abs/1904.11486

Parameters:

image (Array) – 2D input array. shape is (height, width).
kernel_size (tuple[int, int]) – size of the convolving kernel. Accepts tuple of two ints.
strides (tuple[int, int]) – stride of the convolution. Accepts tuple of two ints.

Return type:

Array

serket.image.box_blur_2d(image, kernel_size, dtype=None)[source]#

Box blur.

Parameters:

image (Array) – 2D input array. shape is (height, width).
kernel_size (tuple[int, int]) – size of the convolving kernel. Accepts tuple of two ints.
dtype (Union[dtype, str, Any, None]) – data type of the kernel. Defaults to None to use the same data type as array.

Return type:

Array

Returns:

Box blurred array. shape is (height, width).

serket.image.elastic_transform_2d(key, image, kernel_size, sigma, alpha, dtype=None)[source]#

Elastic transform.

Parameters:

key (Array) – jax random key.
image (Array) – 2D input array. shape is (height, width).
kernel_size (tuple[int, int]) – size of the convolving kernel. Accepts tuple of two ints.
sigma (tuple[float, float]) – sigma of gaussian kernel.
alpha (tuple[float, float]) – alpha of gaussian kernel.
dtype (Union[dtype, str, Any, None]) – data type of the kernel. Defaults to None to use the same data type as array.

serket.image.fft_avg_blur_2d(image, kernel_size, dtype=None)[source]#

Average blur using FFT.

Parameters:

image (Array) – 2D input array. shape is (height, width).
kernel_size (tuple[int, int]) – size of the convolving kernel. Accepts tuple of two ints.
dtype (Union[dtype, str, Any, None]) – data type of the kernel. Defaults to None to use the same data type as array.

Return type:

Array

Returns:

Average blurred array. shape is (height, width).

serket.image.fft_blur_pool_2d(image, kernel_size, strides)[source]#

Blur pooling see https://arxiv.org/abs/1904.11486

Parameters:

image (Array) – 2D input array. shape is (height, width).
kernel_size (tuple[int, int]) – size of the convolving kernel. Accepts tuple of two ints.
strides (tuple[int, int]) – stride of the convolution. Accepts tuple of two ints.

Return type:

Array

serket.image.fft_box_blur_2d(image, kernel_size, dtype=None)[source]#

Box blur using FFT.

Parameters:

image (Array) – 2D input array. shape is (height, width).
kernel_size (tuple[int, int]) – size of the convolving kernel. Accepts tuple of two ints.
dtype (Union[dtype, str, Any, None]) – data type of the kernel. Defaults to None to use the same data type as array.

Return type:

Array

Returns:

Box blurred array. shape is (height, width).

serket.image.fft_elastic_transform_2d(key, image, kernel_size, sigma, alpha, dtype=None)[source]#

Elastic transform using FFT.

Parameters:

key (Array) – jax random key.
image (Array) – 2D input array. shape is (height, width).
kernel_size (tuple[int, int]) – size of the convolving kernel. Accepts tuple of two ints.
sigma (tuple[float, float]) – sigma of gaussian kernel.
alpha (tuple[float, float]) – alpha of gaussian kernel.
dtype (Union[dtype, str, Any, None]) – data type of the kernel. Defaults to None to use the same data type as array.

serket.image.fft_filter_2d(image, weight, strides=(1, 1))[source]#

Filtering wrapping serket fft_conv_general_dilated

Parameters:

image (Array) – 2D input array. shape is (height, width).
weight (Array) – convolutional kernel. shape is (height, width).
strides (tuple[int, int]) – stride of the convolution. Accepts tuple of two ints.

Return type:

Array

serket.image.fft_gaussian_blur_2d(image, kernel_size, sigma, dtype=None)[source]#

Gaussian blur using FFT.

Parameters:

image (Array) – 2D input array. shape is (height, width).
kernel_size (tuple[int, int]) – size of the convolving kernel. Accepts tuple of two ints.
sigma (tuple[float, float]) – sigma of gaussian kernel.
dtype (Union[dtype, str, Any, None]) – data type of the kernel. Defaults to None to use the same data type as array.

Return type:

Array

serket.image.fft_laplacian_2d(image, kernel_size, dtype=None)[source]#

Laplacian using FFT.

Parameters:

image (Array) – 2D input array. shape is (height, width).
kernel_size (tuple[int, int]) – size of the convolving kernel. Accepts tuple of two ints.
dtype (Union[dtype, str, Any, None]) – data type of the kernel. Defaults to None to use the same data type as array.

Return type:

Array

Returns:

Laplacian array. shape is (height, width).

serket.image.fft_motion_blur_2d(image, kernel_size, angle, direction, dtype=None)[source]#

Motion blur using FFT.

Parameters:

image (Array) – 2D input array. shape is (height, width).
kernel_size (tuple[int, int]) – motion kernel width and height. It should be odd and positive.
angle (float) – angle of the motion blur in degrees (anti-clockwise rotation).
direction (int | float) – direction of the motion blur in degrees (anti-clockwise rotation).
dtype (Union[dtype, str, Any, None]) – data type of the kernel. Defaults to None to use the same data type as array.

Return type:

Array

serket.image.fft_sobel_2d(image, dtype=None)[source]#

Sobel filter using FFT.

Parameters:

image (Array) – 2D input array. shape is (height, width).
dtype (Union[dtype, str, Any, None]) – data type of the kernel. Defaults to None to use the same data type as array.

Return type:

Array

serket.image.fft_unsharp_mask_2d(image, kernel_size, sigma, dtype=None)[source]#

Unsharp mask using FFT.

Parameters:

image (Array) – 2D input array. shape is (height, width).
kernel_size (tuple[int, int]) – size of the convolving kernel. Accepts tuple of two ints.
sigma (tuple[float, float]) – sigma of gaussian kernel.
dtype (Union[dtype, str, Any, None]) – data type of the kernel. Defaults to None to use the same data type as array.

Return type:

Array

Returns:

Unsharp masked array. shape is (height, width).

serket.image.filter_2d(image, weight, strides=(1, 1))[source]#

Filtering wrapping jax.lax.conv_general_dilated.

Parameters:

image (Array) – 2D input array. shape is (height, width).
weight (Array) – convolutional kernel. shape is (height, width).
strides (tuple[int, int]) – stride of the convolution. Accepts tuple of two ints.

Return type:

Array

serket.image.gaussian_blur_2d(image, kernel_size, sigma, dtype=None)[source]#

Gaussian blur.

Parameters:

image (Array) – 2D input array. shape is (height, width).
kernel_size (tuple[int, int]) – size of the convolving kernel. Accepts tuple of two ints.
sigma (tuple[float, float]) – sigma of gaussian kernel.
dtype (Union[dtype, str, Any, None]) – data type of the kernel. Defaults to None to use the same data type as array.

Return type:

Array

serket.image.joint_bilateral_blur_2d(image, guidance, kernel_size, sigma_space, sigma_color, dtype=None)[source]#

Joint bilateral blur.

Parameters:

image (Array) – 2D input array. shape is (height, width).
guidance (Array) – 2D guidance array. shape is (height, width).
kernel_size (tuple[int, int]) – size of the convolving kernel. Accepts tuple of two ints.
sigma_space (tuple[float, float]) – sigma of gaussian kernel.
sigma_color (float) – sigma of gaussian kernel.
dtype (Union[dtype, str, Any, None]) – data type of the kernel. Defaults to None to use the same data type as array.

serket.image.laplacian_2d(image, kernel_size, dtype=None)[source]#

Laplacian.

Parameters:

image (Array) – 2D input array. shape is (height, width).
kernel_size (tuple[int, int]) – size of the convolving kernel. Accepts tuple of two ints.
dtype (Union[dtype, str, Any, None]) – data type of the kernel. Defaults to None to use the same data type as array.

Return type:

Array

Returns:

Laplacian array. shape is (height, width).

serket.image.median_blur_2d(image, kernel_size)[source]#

Median blur

Parameters:

image (Array) – 2D input array. shape is (height, width).
kernel_size (tuple[int, int]) – size of the convolving kernel. Accepts tuple of two ints.

Return type:

Array

serket.image.motion_blur_2d(image, kernel_size, angle, direction, dtype=None)[source]#

Motion blur.

Parameters:

image (Array) – 2D input array. shape is (height, width).
kernel_size (tuple[int, int]) – motion kernel width and height. It should be odd and positive.
angle (float) – angle of the motion blur in degrees (anti-clockwise rotation).
direction (int | float) – direction of the motion blur in degrees (anti-clockwise rotation).
dtype (Union[dtype, str, Any, None]) – data type of the kernel. Defaults to None to use the same data type as array.

Return type:

Array

serket.image.sobel_2d(image, dtype=None)[source]#

Sobel filter.

Parameters:

image (Array) – 2D input array. shape is (height, width).
dtype (Union[dtype, str, Any, None]) – data type of the kernel.

Return type:

Array

serket.image.unsharp_mask_2d(image, kernel_size, sigma, dtype=None)[source]#

Unsharp mask.

Parameters:

image (Array) – 2D input array. shape is (height, width).
kernel_size (tuple[int, int]) – size of the convolving kernel. Accepts tuple of two ints.
sigma (tuple[float, float]) – sigma of gaussian kernel.
dtype (Union[dtype, str, Any, None]) – data type of the kernel. Defaults to None to use the same data type as array.

Return type:

Array

Returns:

Unsharp masked array. shape is (height, width).

Filter API

Contents

Filter API#