karabo.imaging.imager_rascil

Overview

This package summerizes tools and functions to be used with the imager from the RASCIL backend. This backend allows both calculating a dirty images and a cleaned image, respoectivley.

Classes

class RascilDirtyImagerConfig(imaging_npixel: int, imaging_cellsize: float, combine_across_frequencies: bool = True, override_cellsize: bool = False)

Config / parameters of a RascilDirtyImager.

Adds parameters specific to RascilDirtyImager.

Attributes:

imaging_npixel (int): see DirtyImagerConfig imaging_cellsize (float): see DirtyImagerConfig combine_across_frequencies (bool): see DirtyImagerConfig override_cellsize (bool): Override the cellsize if it is

above the critical cellsize. Defaults to False.

__init__(imaging_npixel: int, imaging_cellsize: float, combine_across_frequencies: bool = True, override_cellsize: bool = False) → None
class RascilDirtyImager(config: RascilDirtyImagerConfig)

Dirty imager based on the RASCIL library.

Attributes:
config (RascilDirtyImagerConfig): Config containing parameters for

RASCIL dirty imaging.

__init__(config: RascilDirtyImagerConfig) → None

Initializes the instance with a config.

Args:

config (RascilDirtyImagerConfig): see config attribute

create_dirty_image(visibility: Visibility, /, *, output_fits_path: Path | str | None = None) → Image

Creates a dirty image from a visibility.

Args:
visibility: Visibility object

from which to create the dirty image. Contains the visibilities of an observation.

output_fits_path: Path to write the dirty image to.

Example: /tmp/dirty.fits. If None, will be set to a temporary directory and a default file name. Defaults to None.

Returns:

Image: Dirty image

class RascilImageCleanerConfig(imaging_npixel: int, imaging_cellsize: float, ingest_dd: ~typing.List[int] = <factory>, ingest_vis_nchan: int | None = None, ingest_chan_per_vis: int = 1, imaging_nchan: int = 1, imaging_w_stacking: bool | str = True, imaging_flat_sky: bool | str = False, override_cellsize: bool = False, imaging_uvmax: float | None = None, imaging_uvmin: float = 0, imaging_dft_kernel: ~typing.Literal['cpu_looped', 'gpu_raw'] | None = None, client: ~distributed.client.Client | None = None, use_dask: bool = False, n_threads: int = 1, use_cuda: bool = False, img_context: ~typing.Literal['awprojection', '2d', 'ng', 'wg'] = 'ng', clean_algorithm: ~typing.Literal['hogbom', 'msclean', 'mmclean'] = 'hogbom', clean_beam: ~typing.Dict[~typing.Literal['bmaj', 'bmin', 'bpa'], float] | None = None, clean_scales: ~typing.List[int] = <factory>, clean_nmoment: int = 4, clean_nmajor: int = 5, clean_niter: int = 1000, clean_psf_support: int = 256, clean_gain: float = 0.1, clean_threshold: float = 0.0001, clean_component_threshold: float | None = None, clean_component_method: ~typing.Literal['fit', 'extract'] = 'fit', clean_fractional_threshold: float = 0.3, clean_facets: int = 1, clean_overlap: int = 32, clean_taper: ~typing.Literal['none', 'linear', 'tukey'] = 'tukey', clean_restore_facets: int = 1, clean_restore_overlap: int = 32, clean_restore_taper: ~typing.Literal['none', 'linear', 'tukey'] = 'tukey', clean_restored_output: ~typing.Literal['taylor', 'list', 'integrated'] = 'list')

Config / parameters of a RascilImageCleaner.

Adds parameters specific to RascilImageCleaner.

Parameters:

  • imaging_npixel – see ImageCleanerConfig

  • imaging_cellsize – see ImageCleanerConfig

  • ingest_dd – Data descriptors in MS to read (all must have the same number of channels). Defaults to [0].

  • ingest_vis_nchan – Number of channels in a single data descriptor in the MS. Defaults to None.

  • ingest_chan_per_vis – Number of channels per blockvis (before any average). Defaults to 1.

  • imaging_nchan – Number of channels per image. Defaults to 1.

  • imaging_w_stacking – Use the improved w stacking method in Nifty Gridder. Defaults to True.

  • imaging_flat_sky – If using a primary beam, normalise to flat sky. Defaults to False.

  • override_cellsize – Override the cellsize if it is above the critical cellsize. Defaults to False.

  • imaging_uvmax – TODO. Defaults to None.

  • imaging_uvmin – TODO. Defaults to 0.

  • imaging_dft_kernel – DFT kernel: cpu_looped | gpu_raw. Defaults to None.

  • client – Dask client. Defaults to None.

  • use_dask – Use dask? Defaults to False.

  • n_threads – n_threads per worker. Defaults to 1.

  • use_cuda – Use CUDA for Nifty Gridder? Defaults to False.

  • img_context – Which nifty gridder to use. Defaults to “ng”.

  • clean_algorithm – Deconvolution algorithm (hogbom or msclean or mmclean). Defaults to “hogbom”.

  • clean_beam – major axis, minor axis, position angle (deg). Defaults to None.

  • clean_scales – Scales for multiscale clean (pixels) e.g. [0, 6, 10]. Defaults to [0].

  • clean_nmoment – Number of frequency moments in mmclean (1=constant, 2=linear). Defaults to 4.

  • clean_nmajor – Number of major cycles in cip or ical. Defaults to 5.

  • clean_niter – Number of minor cycles in CLEAN. Defaults to 1000.

  • clean_psf_support – Half-width of psf used in cleaning (pixels). Defaults to 256.

  • clean_gain – Clean loop gain. Defaults to 0.1.

  • clean_threshold – Clean stopping threshold (Jy/beam). Defaults to 1e-4.

  • clean_component_threshold – Sources with absolute flux > this level (Jy) are fit or extracted using skycomponents. Defaults to None.

  • clean_component_method – Method to convert sources in image to skycomponents: “fit” in frequency or “extract” actual values. Defaults to “fit”.

  • clean_fractional_threshold – Fractional stopping threshold for major cycle. Defaults to 0.3.

  • clean_facets – Number of overlapping facets in faceted clean along each axis. Defaults to 1.

  • clean_overlap – Overlap of facets in clean (pixels). Defaults to 32.

  • clean_taper – Type of interpolation between facets in deconvolution: none or linear or tukey. Defaults to “tukey”.

  • clean_restore_facets – Number of overlapping facets in restore step along each axis. Defaults to 1.

  • clean_restore_overlap – Overlap of facets in restore step (pixels). Defaults to 32.

  • clean_restore_taper – Type of interpolation between facets in restore step (none, linear or tukey). Defaults to “tukey”.

  • clean_restored_output – Type of restored image output: taylor, list, or integrated. Defaults to “list”.

__init__(imaging_npixel: int, imaging_cellsize: float, ingest_dd: ~typing.List[int] = <factory>, ingest_vis_nchan: int | None = None, ingest_chan_per_vis: int = 1, imaging_nchan: int = 1, imaging_w_stacking: bool | str = True, imaging_flat_sky: bool | str = False, override_cellsize: bool = False, imaging_uvmax: float | None = None, imaging_uvmin: float = 0, imaging_dft_kernel: ~typing.Literal['cpu_looped', 'gpu_raw'] | None = None, client: ~distributed.client.Client | None = None, use_dask: bool = False, n_threads: int = 1, use_cuda: bool = False, img_context: ~typing.Literal['awprojection', '2d', 'ng', 'wg'] = 'ng', clean_algorithm: ~typing.Literal['hogbom', 'msclean', 'mmclean'] = 'hogbom', clean_beam: ~typing.Dict[~typing.Literal['bmaj', 'bmin', 'bpa'], float] | None = None, clean_scales: ~typing.List[int] = <factory>, clean_nmoment: int = 4, clean_nmajor: int = 5, clean_niter: int = 1000, clean_psf_support: int = 256, clean_gain: float = 0.1, clean_threshold: float = 0.0001, clean_component_threshold: float | None = None, clean_component_method: ~typing.Literal['fit', 'extract'] = 'fit', clean_fractional_threshold: float = 0.3, clean_facets: int = 1, clean_overlap: int = 32, clean_taper: ~typing.Literal['none', 'linear', 'tukey'] = 'tukey', clean_restore_facets: int = 1, clean_restore_overlap: int = 32, clean_restore_taper: ~typing.Literal['none', 'linear', 'tukey'] = 'tukey', clean_restored_output: ~typing.Literal['taylor', 'list', 'integrated'] = 'list') → None
class RascilImageCleaner(config: RascilImageCleanerConfig)

Image cleaner based on the RASCIL library.

Attributes:
config (RascilImageCleanerConfig): Config containing parameters for

RASCIL image cleaning.

__init__(config: RascilImageCleanerConfig) → None

Initializes the instance with a config.

Args:

config (RascilImageCleanerConfig): see config attribute

create_cleaned_image(visibility: Visibility, /, *, dirty_fits_path: Path | str | None = None, output_fits_path: Path | str | None = None) → Image

Creates a clean image from a dirty image or from visibilities.

Arguments:

visibility: Visibility from which a clean image should be created. dirty_fits_path: Path to dirty image FITS file that

should be reused to create a clean image. If None, dirty image will be created first from the visibilities. Defaults to None.

output_fits_path: Path to write the clean image to.

Example: /tmp/restored.fits. If None, will be set to a temporary directory and a default file name. Defaults to None.

Returns:

Image: Clean image

create_cleaned_image_variants(visibility: Visibility, /, deconvolved_fits_path: Path | str | None = None, restored_fits_path: Path | str | None = None, residual_fits_path: Path | str | None = None) → Tuple[Image, Image, Image]

Creates a clean image from visibilities.

Args:
visibility: Visibility from which a clean image should be created.

Please note: only MS visibilities supported.

deconvolved_fits_path: Path to write the deconvolved image to.

Example: /tmp/deconvolved.fits. If None, will be set to a temporary directory and a default file name.

restored_fits_path: Path to write the restored image to.

Example: /tmp/restored.fits. If None, will be set to a temporary directory and a default file name.

residual_fits_path: Path to write the residual image to.

Example: /tmp/residual.fits. If None, will be set to a temporary directory and a default file name.

Returns:

A tuple (deconvolved, restored, residual) of images