clean-algorithms

CASA Clean

• clean involves finding the peaks in residual image, subtraction the effect of PSFs for the peaks.
• operation models of clean algorithms
  • delta function sky model
  • multi-scale sky model
• clean algorithms
  • hogbom (hogbom)
    • uses delta function sky model
    • it uses full PSF and thus it is high compute intensive
    • efficient as delta functions are optimal for fields of compact sources
    • suspectible to errors for inappropriate weights especially when PSF has high sidelobs.
  • clark (clark)
    • it uses only small patch of psf to update the residual image
    • minor cycle is faster because of above point.
    • However, it have errors when sources are spread over wide field of view where the flux estimate at a given location is affected the sidelobs of farout sources.
    • To compensate this issue, the iteration is stopped when the brightest peak in the residual image is below the sidelobe level of the brightest source in the initial residual image.
    • for polarization imaging, it searches peak in combined stokes
      • $I^2+Q^2+U^2+V^2$
  • clarkstokes (clarkstokes)
    • uses clark psf
    • for polarization imaging, stokes planes are cleaned sequentially.
  • multi-scale (cornwell-holdaway Multi-scale clean) multiscale
    • scale-sensitive deconvolution algorithm for images with complicated spatial structure.
    • matched-filtering technique is used locate dominant flux components.
    • it requires scale sizes in pixel for algorithm.
  • Multi-term (Multi Scale) Multi-Frequency Synthesis (mtmfs)
    • a wide-band imaging algorithm
  • Maximum Entropy Method (mem)
  •  Adaptive Scale Pixel algorithm (asp)
    • experimental version in tclean

This algorithm can be given to tclean task using parameter deconvolver (string='hogbom')

Astron Radler

• radler (deconvolver) at astron
  • not able to install radler using source.

open questions

1. chi-square minimization