Restoration (darsia.restoration)#

Smoothness and shape altering restoration algorithms.

Submodules#

darsia.restoration.binaryinpaint module#

Module containing class with inpainting workflow for binary signals.

class BinaryFillHoles(area_threshold=None, key='', **kwargs)[source]#

Bases: object

Wrapper for filing holes in binary images with skimage.

class BinaryLocalConvexCover(cover_patch_size=None, key='', **kwargs)[source]#

Bases: object

Local convex cover using local convex hulls with skimage.

class BinaryRemoveSmallObjects(min_size=None, key='', **kwargs)[source]#

Bases: object

Wrapper for removing small objects in binary images with skimage.

darsia.restoration.heterogeneous_tvd module#

darsia.restoration.median module#

Module wrapping median algoriths from skimage into the realm of DarSIA.

class Median(key='', **kwargs)[source]#

Bases: object

Median through skimage.filter.rank.

darsia.restoration.resize module#

Module containing wrappers to resize routines from skimage and cv2.

Access is given through objects. Also contains utility routine which equalizes voxel size lengths.

class Resize(ref_image=None, shape=None, fx=None, fy=None, interpolation=None, dtype=None, key='', **kwargs)[source]#

Bases: object

Object for resizing 2d arrays.

Example:

import darsia options = {

“example resize x”: 0.2, “example resize y”: 0.5, “example interpolation”: “inter_nearest”,

} resizer = darsia.Resize(key = “example “, **options) img_small = resizer(img_large)

load(path)[source]#

Load parameters from file.

Parameters:

path (str) – path to load the parameters from

save(path)[source]#

Save parameters to npz file.

Parameters:

path (str) – path to save the parameters to

equalize_voxel_size(image, voxel_size=None, **kwargs)[source]#

Resize routine which keeps physical dimensions, but unifies the voxel length.

Parameters:

  • image (darsia.Image) – image to be resized

  • voxel_size (float, optional) – side length, min of the voxel side of the image if None.

  • arguments (keyword) – interpolation (str): interpolation type used for resize

Returns:

resized image

Return type:

darsia.Image

resize(image, ref_image=None, shape=None, fx=None, fy=None, interpolation=None, dtype=None)[source]#

Function wrapper to Resize object.

Parameters:

  • image (darsia.Image) – image to be resized

  • ref_image (Image, optional) – reference image whose shape is desired

  • shape (tuple of int, optional) – desired shape (in matrix indexing)

  • fx (float, optional) – resize factor in x-dimension.

  • fy (float, optional) – resize factor in y-dimension.

  • interpolation (str, optional) – interpolation method, default: None, invoking the default option in cv2.resize.

  • dtype – conversion dtype before resizing; noting happens if None

uniform_refinement(image, levels)[source]#

Uniform refinement.

Parameters:

  • image (darsia.Image) – image

  • levels (int) – refinement levels, if positive, coarsing levels, if negative.

Returns:

resized image

Return type:

darsia.Image

darsia.restoration.tvd module#

Module wrapping TV denoising algoriths from skimage into the realm of DarSIA. These can be directly integrated in darsia.ConcentrationAnalysis, in particular as part of the definition of a restoration object.

class TVD(key='', **kwargs)[source]#

Bases: object

Total variation denoising interface.

Connects to skimage.restoration.denoise_tv_* methods as well as to the DarSIA implementation of the split Bregman method (:mod:’split_bregman_tvd.py’).

tvd(img, method='chambolle', weight=1.0, max_num_iter=200, eps=0.0002, **kwargs)[source]#

Inline application of TVD.

Parameters:

  • img (np.ndarray or Image) – image

  • method (str) – TVD method

  • weight (float or np.ndarray) – weight

  • max_num_iter (int) – maximum number of iterations

  • eps (float) – tolerance

  • **kwargs – additional arguments - omega (array or float): data fidelity weight for heterogeneous bregman - regularization (float): regularization parameter for heterogeneous bregman

Returns:

upscaled image (same type as input)

Return type:

np.ndarray or Image