Shape corrections (darsia.corrections.shape)

Shape corrections target geometrical transformations as translations, rotations, but also locally varying bulge and stretch as well as general deformation resulting from direct comparison with a reference image.

Curvature correction (darsia.corrections.shape.curvature)

Module containing polynomical Curvature correction.

class CurvatureCorrection(config=None, **kwargs)

Bases: BaseCorrection

Polynomial curvature correction.

Contains routines for setting up the curvature correction, as well as applying it to images.

config

config dictionary for curvture correction.

Type:

dict

Circumstantial attributes

reference_image (np.ndarray): image matrix of the reference image. current_image (np.ndarray): image matrix of the updated reference image. width (float): physical width of reference image. height (float): physical height of reference image. in_meters (bool): True if width/height is in meters. Ny (int): number of pixels in vertical direction in reference image. Nx (int): number of pixels in horizontal direction in reference image.

bulge_correction(left=0, right=0, top=0, bottom=0)

Bulge correction

Corrects bulging of image, depending on the amount of pixels that the image is bulged inwards on each side.

Parameters:

• left (int) – amount of bulged pixels on the left side of the image.

• right (int) – amount of bulged pixels on the right side of the image.

• top (int) – amount of bulged pixels on the top of the image.

• bottom (int) – amount of bulged pixels on the bottom of the image.

compute_bulge(img=None, **kwargs)

Compute the bulge parameters depending on the maximum number of pixels that the image has been displaced on each side.

Parameters:

• img (np.ndarray, optional) – image array, basis for the computation.

• kwargs (optional keyword arguments) –

  “left” (int): the maximum number of pixels that the image

  has been displaced on the left side

  ”right” (int): the maximum number of pixels that the image

  has been displaced on the right side

  ”top” (int): the maximum number of pixels that the image

  has been displaced on the top

  ”bottom” (int): the maximum number of pixels that the image

  has been displaced on the bottom

compute_stretch(img=None, **kwargs)

Compute the stretch parameters depending on the stretch center, and a known translation.

Parameters:

• img (np.ndarray, optional) – image array, basis for the computation.

• kwargs (optional keyword arguments) – “point_source” (list): point that has been translated. “point_destination” (list): the ought to be position. “stretch_center” (list): the stretch center.

correct_array(img)

Call method of the curvature correction.

Applies the curvature correction to a provided image, and returns the corrected image as an array. If set in the constructor, the image will be resized in the first step.

Parameters:

img (np.ndarray) – image array

Returns:

curvature corrected image.

Return type:

np.ndarray

correct_metadata(metadata={})

Extract metadata from the config file.

Parameters:

metadata (dict, optional) – metadata dictionary to be updated. Defaults to {}.

Returns:

metadata

Return type:

dict

crop(corner_points)

Crop the image along the corners of the image.

The four corner points of the image should be provided, and this method will update the config file and modify the current image.

Parameters:

corner_points (VoxelArray) – list of the corner points. Preferably the list should be ordered starting from the upper left corner and going counter clockwise.

load(path)

Load the curvature correction from a file.

Parameters:

path (Path) – path to the file

pre_bulge_correction(**kwargs)

Initialize the curvature correction by forcing all stright lines to curve inwards and not outwards.

Parameters:

kwargs (optional keyword arguments) –

“horizontal_bulge” (float): parameter for the curvature correction related to

the horizontal bulge of the image.

”horizontal_center_offset” (int): offset in terms of pixel of the image center

in x-direction, as compared to the numerical center

vertical_bulge (float): parameter for the curvature correction related to the

vertical bulge of the image.

”vertical_center_offset” (int): offset in terms of pixel of the image center in

y-direction, as compared to the numerical center

read_config_from_file(path)

Reads a json-file to the config disctionary.

Parameters:

path (Path) – path to the json-file.

return_image()

Returns the current image as a darsia image width provided width and height.

save(path)

Save the curvature correction to a file.

Parameters:

path (Path) – path to the file

setup_config(config=None)

show_image()

Shows the current image using matplotlib.pyplot

simple_curvature_correction(img, **kwargs)

General routine for applying stretch and bulge transforms. This routine in contrast to __call__ does always use the keyword arguments and constructs the transformation instead of using cached values.

Parameters:

• img (np.ndarray) – image array

• kwargs (optional keyword arguments) – see _transform_coordinates for more details.

Returns:

corrected image

Return type:

np.ndarray

stretch_correction(point_source, point_destination, stretch_center)

Stretch correction.

Stretches the image depending on the displacement of a single point (point source <–> point_destination) and an undisplaced point (stretch_center)

Parameters:

• "point_source" (list) – point that has been translated.

• "point_destination" (list) – the ought to be position.

• "stretch_center" (list) – the stretch center.

write_config_to_file(path)

Writes the config dictionary to a json-file.

Parameters:

path (Path) – path to the json file

property temporary_image

load_curvature_correction_config_from_dict(sec)

Load curvature correction config from a dictionary.

Parameters:

path (Path) – path to the toml file.

Returns:

config dictionary for curvature correction.

Return type:

config (dict)

load_curvature_correction_config_from_toml(path)

Load curvature correction config from a toml file.

Parameters:

path (Path) – path to the toml file.

Returns:

config dictionary for curvature correction.

Return type:

config (dict)

Local deformation correction (darsia.corrections.shape.deformation)

Module containing a local deformation correction.

This is essentially a correction object through wrappping darsia.DiffeomorphicImageRegistration.

class DeformationCorrection(base, config)

Bases: BaseCorrection

Patch-wise defined deformation correction defined through image registration.

base

reference (baseline) image.

Type:

darsia.Image

correct_array(img)

Main routine for aligning image with baseline image.

Parameters:

img (np.ndarray) – input image, to be aligned.

Returns:

aligned image array.

Return type:

np.ndarray

correct_metadata(metadata={})

Correction routine on metadata level.

Parameters:

metadata (dict) – metadata dictionary.

Returns:

corrected metadata dictionary.

Return type:

dict

abstractmethod load(path)

Load the correction from a file.

The method should load a npz file, containing the class name and required data for loading the correction from file.

Parameters:

path (str) – path to the file

abstractmethod save(path)

Save the correction to a file.

The method should store a npz file, continaing the class name and required data for loading the correction from file.

Parameters:

path (str) – path to the file

Drift correction (darsia.corrections.shape.drift)

Module containing objects to align images with respect to a baseline image.

The alignment is based on feature detection. It is possible to use segments of the image. Finally, corrections are based on single, global translation.

class DriftCorrection(base=None, config=None)

Bases: BaseCorrection

Class for drift correction of images wrt. a baseline image.

correct_array(img, roi=None)

Main routine for aligning image with baseline image.

Parameters:

• img (np.ndarray) – input image, to be aligned.

• roi (2-tuple of slices, optional) – ROI to be applied to img; if None the cached roi is used.

Returns:

aligned image array.

Return type:

np.ndarray

correct_metadata(metadata={})

Correction routine on metadata level.

Parameters:

metadata (dict) – metadata dictionary.

Returns:

corrected metadata dictionary.

Return type:

dict

load(path)

Load the drift correction from a file.

return_config()

Return config file for the drift correction.

save(path)

Save the drift correction to a file.

base

Base image array.

translation_estimator

Detection of effective translation based on feature detection.

Piecewise perspective transformation (darsia.corrections.shape.piecewiseperspective)

Module defining how to determine and apply patch-wise defined perspective transform.

class PiecewisePerspectiveTransform(**kwargs)

Bases: object

Globally constant, patch-wise defined perspective transform.

find_and_warp(patches, displacement, reverse=False)

Continuously transform entire image via patchwise perspective transform.

Perspective transforms determined by the function evaluation in all corners of each patch results in continuous transformations. Hence, stiching together the transformed images results in an image without overlap and gaps.

Parameters:

• patches (darsia.Patches) – patched image

• displacement (Callable) – relative deformation map; assumed to be continuous

• reverse (bool) – flag whether displacement is applied with negative weight

Returns:

transformed image

Return type:

darsia.Image

Rotation correction (darsia.corrections.shape.rotation)

Module containing objects useful for correcting images wrt rotations

class RotationCorrection(anchor, **kwargs)

Bases: BaseCorrection

Rotation correction.

Rotations are defined as combination of multiple basic rotations. In 2d, a single basic rotation is sufficient. In 3d, although three are available, two are sufficient.

dim

ambient dimension

Type:

int

anchor

voxel coordinates of anchor

Type:

array or list

rotation

rotation matrix

Type:

array

rotation_inv

inverted rotation matrix

Type:

array

correct_array(img)

Main routine: Application of inherent rotation to provided image.

Parameters:

img (array) – image

Returns:

rotated image

Return type:

array

correct_metadata(metadata={})

Correction routine on metadata level.

Parameters:

metadata (dict) – metadata dictionary.

Returns:

corrected metadata dictionary.

Return type:

dict

load(path)

save(path)

Translation correction (darsia.corrections.shape.translation)

Module defining estimator for translation between ROIs and corresponding correction.

class TranslationCorrection(translation=None)

Bases: BaseCorrection

Translation correction based on user-input global translation.

Translation is provided as array or path to numpy array.

correct_array(img)

Main routine. Perform translation.

Parameters:

img (np.ndarray) – image to be corrected.

Returns:

Corrected image

Return type:

array

correct_metadata(metadata={})

Correction routine on metadata level.

Parameters:

metadata (dict) – metadata dictionary.

Returns:

corrected metadata dictionary.

Return type:

dict

load(path)

save(path)

active

Flag controlling whether the correction is active.

translation

Translation map/array.

class TranslationEstimator(max_features=200, tol=0.05, keep_percent=0.1)

Bases: object

Estimator of translation to align images based on feature detection.

Estimation is performed by restriction to ROIs.

find_effective_translation(img_src, img_dst, roi_src=None, roi_dst=None, mask_src=None, mask_dst=None, plot_matches=False)

Find translation to align two images through feature matching.

All lengths are measured in number of pixels.

Parameters:

• img_src (np.ndarray) – source image

• img_dst (np.ndarray) – destination image

• roi_src (tuple of slices) – region of interested associated to the source image

• roi_dst (tuple of slices) – region of interested associated to the destination image

• mask_src (np.ndarray, optional) – boolean mask detecting considered pixels in the analysis; if None, all pixels are considered.

• mask_dst (np.ndarray, optional) – boolean mask detecting considered pixels in the analysis; if None, all pixels are considered.

• plot_matches (bool) – flag controlling whether the matching features are plotted; useful for debugging; default value is False

Returns:

transformation matrix operating on pixel coordinates using reverse

matrix indexing

bool: flag indicating whether the procedure was successful

Return type:

np.ndarray

match_roi(img_src, img_dst, roi_src=None, roi_dst=None, plot_matches=False)

Align two images through feature matching in provided ROIs.

Parameters:

• img_src (np.ndarray or darsia.Image) – source image

• img_dst (np.ndarray or darsia.Image) – destination image

• roi_src (tuple of slices) – region of interested associated to the source image

• roi_dst (tuple of slices) – region of interested associated to the destination image translation

• plot_matches (bool) – flag controlling whether the matching features are plotted; useful for debugging; default value is False

Returns:

aligned source image, if input of type array

Return type:

np.ndarray, optional