Skip to content

Imaging

Imaging is the process of transforming interferometric measurements of the sky from Fourier space (visibilities) into image space (pixels) in order to produce a visual representation of the sky brightness distribution.

Imaging involves mapping irregularly sampled visibilities onto a regular grid in the uv-plane and forming an image via an inverse Fourier transform, which represents the sky convolved with the point-spread function (PSF) of the interferometer. Deconvolution techniques are typically used to estimate the underlying sky brightness distribution by accounting for the effects of incomplete uv-coverage.

Deconvolution

This step in imaging iteratively builds a sky model by identifying significant structure in the dirty image, subtracting the PSF response of that structure, and repeating until the residuals are consistent with noise.

CLEAN is a class of algorithms (including Högbom, Clark, Cotton-Schwab) that are used in deconvolution: deconvolution is often called "CLEANing"; model components identified during deconvolution are often called CLEAN components; and the final image is often called the CLEAN image.

Inputs

  • Visibilities (VisibilitySet): complex visibilities indexed by baseline, time, frequency and polarization
  • UVW coordinates, weights, and flags (if present)
  • Observation parameters: phase centre, frequency information, and time sampling
  • Imaging configuration:
    • Image parameters (image size, pixel size/cell size, field of view)
    • Gridding configuration (kernel, oversampling/support, w-term handling)
    • Weighting scheme (natural / uniform / robust)
    • Deconvolution settings (algorithm choice, stopping criteria, masks)

Outputs

Imaging produces one or more image products, including:

  • Dirty image: the direct inverse transform of gridded visibilities.
  • PSF (dirty beam): the image-plane response of the sampling pattern and weighting.
  • Model image: a deconvolved estimate of the sky brightness.
  • Residual image: dirty image minus the model convolved with the PSF.
  • Restored image: model image convolved with a restoring beam plus residuals.

These may be produced as 2D images or image cubes (i.e. an image per frequency channel).

Role

  • Visualise the effect of array layout, observation parameters, measurement effects.
  • Producing improved sky models during iterative self-calibration.
  • Experiment with (and optimise) imaging and gridding settings and algorithms.
  • Provide data for quality assurance and diagnostics that reveal sampling limitations and calibration errors.

Invariants

  • Imaging products are defined relative to a specific phase centre and image coordinate system.
  • Image products share a common image grid (pixel size, shape, coordinate definition).
  • The PSF corresponds to the same weighting and gridding choices used to form the dirty image.
  • Residuals are computed by subtracting the model image convolved with the PSF from the dirty image.
  • Imaging does not modify input visibilities.