aics_tf_registration.core package

Submodules

aics_tf_registration.core.alignment module

aics_tf_registration.core.alignment.align_xy(fixed: numpy.ndarray, moving: numpy.ndarray)[source]

Perform alignment of the images in 2d.

Parameters

  • - fixed (image with larger field of view)

  • - moving (image with smaller field of view)

Returns

  • - fixed_2dAlign_offset_x (rigid offset in x)

  • - fixed_2dAlign_offset_y (rigid offset in y)

aics_tf_registration.core.alignment.align_z(fixed: numpy.ndarray, moving: numpy.ndarray, prealign: bool, denoise: bool)[source]

Align the images in the z-plane.

Parameters

  • - fixed (image with larger field of view)

  • - moving (image with smaller field of view)

  • - prealign (whether to initially estimate alignment by overlapping segmentations)

  • - denoise (“denoise” image through 10th & 90th percentile clipping)

Returns

  • - fixed_addition_offset_z (rigid offset in z)

  • - fixed_addition_offset_x (rigid offset in x)

  • - fixed_addition_offset_y (rigid offset in y)

aics_tf_registration.core.alignment.final_refinement(lr: numpy.ndarray, hr: numpy.ndarray, scale_factor_xy: float, scale_factor_z: float, min_subcrop_xy: int, min_subcrop_z: int, error_thresh: float = 0.01)[source]

Adjust the final 3d alignment by repeating alignment in high resolution image scale.

Parameters

  • - lr (low-res (source) image after initial alignment)

  • - hr (high-res (target) image after initial alignment)

  • - scale_factor_xy (upsample/downsample rate for x and y)

  • - scale_factor_z (upsample/downsample rate for z)

  • - min_subcrop_xy (minimum number of pixels to crop to match scale factor)

  • - min_subcrop_z (minimum number of pixels to crop to match scale factor)

  • - error_thresh (maximum error in scale factors tolerated for final image)

Returns

  • - fixed_final (aligned and cropped fixed image)

  • - moving_final (aligned and cropped moving image)

aics_tf_registration.core.alignment.finalize_alignment(fixed: numpy.ndarray, moving_orig: numpy.ndarray, moving_scaled_shape: Tuple[int], moving_adjust_z_shape: Tuple[int], init_z_padding: int, fixed_addition_offset_z: int, fixed_addition_offset_y: int, fixed_addition_offset_x: int, fixed_2dAlign_offset_y: int, fixed_2dAlign_offset_x: int, scale_factor_xy: float, scale_factor_z: float)[source]

Use refineuse_refinement-force search adjust the final 3d alignment.

Parameters

  • - fixed (image with larger field of view)

  • - moving_orig (image with smaller field of view)

  • - moving_scaled_shape (image dimensions of rescaled smaller FOV)

  • - moving_adjust_z_shape (image dimensions after adjustments to stack size)

  • - init_z_padding (number of stacks padded or clipped from rescaled image)

  • - fixed_addition_offset_z (rigid offset in z from itk alignment)

  • - fixed_addition_offset_y (rigid offset in y from itk alignment)

  • - fixed_addition_offset_x (rigid offest in x from itk alignment)

  • - fixed_2dAlign_offset_y (rigid offset in y from 2d alignment)

  • - fixed_2dAlign_offset_x (rigid offest in x from 2d alignment)

  • - scale_factor_xy (upsample/downsample rate for x and y)

  • - scale_factor_z (upsample/downsample rate for z)

Returns

  • - fixed_final (aligned and cropped fixed image)

  • - moving_final (aligned and cropped moving image)

aics_tf_registration.core.alignment.perform_alignment(source: numpy.ndarray, target: numpy.ndarray, smaller_fov_modality: str, scale_factor_xy: float, scale_factor_z: float, source_alignment_channel: int, target_alignment_channel: int, source_output_channel: list, target_output_channel: list, prealign_z: bool, denoise_z: bool, use_refinement: bool, save_composite: bool)[source]

Wrapper function for all of the steps necessary to calculate alignment.

Parameters

  • - source (low-res modality image)

  • - target (high-res modality image)

  • - smaller_fov_modality (which modality has the smaller field of view)

  • - scale_factor_xy (upsample/downsample rate to match image scales in x and y)

  • - scale_factor_z (upsample/downsample rate to match image scales in z)

  • - source_alignment_channel (source image channel used for calculating alignment)

  • - target_alignment_channel (target image channel used for calculating alignment)

  • - source_output_channel (source image channel to apply alignment on)

  • - target_output_channel (target image channel to apply alignment on)

  • - prealign (whether to calculate intitial estimate of z-alignment)

  • - denoise_z (denoise z-stacks prior to z-alignment)

  • - use_refinement (refine alignment by repeating in the target image resolution)

  • - save_composite (save composite image of final alignment)

Returns

  • - source_aligned (aligned source image)

  • - target_aligned (aligned target image)

aics_tf_registration.core.image_aligner module

class aics_tf_registration.core.image_aligner.Image_Aligner(config_path: str)[source]

Bases: object

Load settings and file paths from configuration file

Parameters

- config_path (path to yaml file with settings and file paths)

align_images()[source]

Run alignment algorithm on image pairs in dataset

get_image_list(path)[source]

aics_tf_registration.core.preprocessing module

aics_tf_registration.core.preprocessing.apply_transforms(image: numpy.ndarray, transform_list: List[str])[source]

Apply the transformations from list provided to the moving image

Parameters

  • - image (3D or 4D image)

  • - transform list (list of transforms to apply in sequence to image)

Returns

- image

Return type

transformed image

aics_tf_registration.core.preprocessing.precrop_image(image: numpy.ndarray, scales: numpy.ndarray)[source]

Applies precropping to the image

Parameters

  • - image (3D or 4D image)

  • - scales (array containing percent of each axis to crop)

Returns

- image_cropped

Return type

cropped image

aics_tf_registration.core.preprocessing.rearrange_multichannel_image(image: numpy.ndarray)[source]

Rearrange the order of dimensions so that the channel dimension is first

Parameters

- image (4d image)

Returns

-image

Return type

image with channels rearranged

aics_tf_registration.core.preprocessing.rescale_image(image: numpy.ndarray, scale_factor_xy: float, scale_factor_z: float)[source]

Upsample/Downsample the image to match voxel dimensions of the other image.

Parameters

  • - image (3D or 4D image to rescale)

  • - scale_factor_xy (Upsample/downsample rate in x and y)

  • - scale_factor_z (Upsample/downsample rate in z)

Returns

- image

Return type

rescaled image

Module contents