eolearn.coregistration.coregistration

This module implements the co-registration transformers.

class eolearn.coregistration.coregistration.ECCRegistrationTask(registration_feature, reference_feature, channel, valid_mask_feature=None, apply_to_features=Ellipsis, interpolation_mode=1, warp_mode=0, max_iter=100, gauss_kernel_size=1, border_mode=1, border_value=0, num_threads=1, max_translation=5.0)

Bases: EOTask

Multi-temporal image co-registration using OpenCV Enhanced Cross-Correlation method

The task uses a temporal stack of images of the same location (i.e. a temporal-spatial feature in EOPatch) and a reference timeless feature to estimate a transformation that aligns each frame of the temporal stack to the reference feature.

Each transformation is calculated using only a single channel of the images. If feature which contains masks of valid pixels is specified it is used during the estimation of the transformation. The estimated transformations are applied to each of the specified features.

Parameters:

• registration_feature (Feature) – Feature in EOPatch holding the multi-temporal stack to register to the reference. Needs to be of FeatureType.DATA.

• reference_feature (Feature) – Feature in EOPatch used as reference frame for registration.

• channel (int) – Defines the index of the stack and reference feature to use during registration.

• valid_mask_feature (Feature | None) – Optional feature in EOPatch that defines which pixels should be used for registration.

• apply_to_features (FeaturesSpecification) – List of temporal features in EOPatch to which applied the estimated transformation.

• interpolation_mode (int) – Interpolation type used when transforming the stack of images.

• warp_mode (int) – Defines the transformation model used to match the stack and the reference. Examples include TRANSLATION, RIGID_MOTION, AFFINE.

• max_iter (int) – Maximum number of iterations used during optimization of algorithm.

• gauss_kernel_size (int) – Size of Gaussian kernel used to smooth images prior to registration.

• border_mode (int) – Defines the padding strategy when transforming the images with estimated transformation.

• border_value (float) – Value used for padding when border mode is set to CONSTANT.

• num_threads (int) – Number of threads used for optimization of the algorithm.

• max_translation (float) – Estimated transformations are considered incorrect when the norm of the translation component is larger than this parameter.

register(src, trg, valid_mask=None, warp_mode=0)

Method that estimates the transformation between source and target image

Parameters:

• src (ndarray) –

• trg (ndarray) –

• valid_mask (ndarray | None) –

• warp_mode (int) –

Return type:

ndarray

execute(eopatch)

Method that estimates registrations and warps EOPatch objects

Parameters:

eopatch (EOPatch) –

Return type:

EOPatch

warp(img, warp_matrix, shape, flags)

Transform the target image with the estimated transformation matrix

Parameters:

• img (ndarray) –

• warp_matrix (ndarray) –

• shape (tuple[int, int]) –

• flags (int) –

Return type:

ndarray

warp_feature(img, warp_matrix)

Function to warp input image given an estimated 2D linear transformation

Parameters:

• img (ndarray) –

• warp_matrix (ndarray) –

Return type:

ndarray

is_translation_large(warp_matrix)

Method that checks if estimated linear translation could be implausible.

This function checks whether the norm of the estimated translation in pixels exceeds a predefined value.

Parameters:

warp_matrix (ndarray) –

Return type:

bool

eolearn.coregistration.coregistration.get_gradient(src)

Method which calculates and returns the gradients for the input image, which are better suited for co-registration

Parameters:

src (ndarray) –

Return type:

ndarray