SDP Imaging
Select SDP with get_imager(ImagingBackend.SDP, config=SdpImagerConfig(...)).
Import the configuration from karabo.imaging.imager_factory; users do not
need to instantiate the implementation class below.
The adapter reads a Measurement Set, converts its visibilities to Stokes I,
flags autocorrelations, and performs separate dirty and PSF inversions.
Restoration uses Hogbom CLEAN followed by convolution with the fitted clean
beam and addition of the residual. Other CLEAN algorithms currently raise
NotImplementedError. After restoration, last_model_image and
last_residual_image expose the exported model and residual on the SDP imager.
See Imaging Backend Selection for pixel-scale and channel-combination behavior, and Migrating from RASCIL for upgrading older imaging scripts.
Configuration
- class SdpImagerConfig(combine_across_frequencies: 'bool' = True, weighting: 'str' = 'natural', context: 'str' = '2d', override_cellsize: 'bool' = False, clean_algorithm: 'str' = 'hogbom', clean_niter: 'int' = 100, clean_gain: 'float' = 0.1, clean_threshold: 'float' = 0.0, clean_preserve_facets: 'bool' = False, save_model_and_residual: 'bool' = True, clean_fractional_threshold: 'float | None' = None, clean_scales: 'tuple[int, ...] | None' = None)
- clean_algorithm: str = 'hogbom'
- clean_fractional_threshold: float | None = None
- clean_gain: float = 0.1
- clean_niter: int = 100
- clean_preserve_facets: bool = False
- clean_scales: tuple[int, ...] | None = None
- clean_threshold: float = 0.0
- combine_across_frequencies: bool = True
- context: str = '2d'
- override_cellsize: bool = False
- save_model_and_residual: bool = True
- weighting: str = 'natural'
Implementation reference
- class SdpImager(config: SdpImagerConfig | None = None)
- invert(vis: Visibility, image_spec: ImageSpec) tuple[Image, Image]
Grid visibilities into dirty image and PSF images.