Algorithms

enum class radler::AlgorithmType

The deconvolution algorithm type.

Values:

enumerator kGenericClean

A “Högbom” CLEAN algorithm, extended with multi frequency/polarization clean. It also extends the basic CLEAN algorithm with features such as auto-masking and spectral fitting (both are described in Offringa & Smirnov, 2017).

enumerator kAdaptiveScalePixel

The adaptive scale pixel algorith described by Bhatnagar & Cornwell (2004), extended with support for multi-frequency deconvolution and sub-image deconvolution. The algorithm is rather slow and generally does not result in better results compared to Radler’s multiscale algorithm. In specific cases with diffuse structures it may be useful.

enumerator kIuwt

An algorithm similar to the MORESANE algorithm (A Dabbech et al., 2014), but reimplemented in C++ and extended for multi frequency/polarization clean.

enumerator kMoreSane

Makes use of the external MORESANE package that implements the algorithm described by A. Dabbech et al. (2014). Requires specification of the location of MORESANE (with Settings::MoreSane::location). This method does not support multi-frequency/polarization cleaning.

enumerator kMultiscale

Implements the algorithm described by Offringa & Smirnov (2017). This algorithms allows deconvolving resolved and/or diffuse emission. It allows cleaning of multiple polarizations or frequencies and integrates auto-masking. This method results in accurate deconvolution and is at present fast enough to deconvolve very large (60K^2 pixels) images. For almost all cases, this should be the preferred algorithm.

enumerator kPython

This option allows implementing a custom algorithm in Python. A location to the Python code should be provided (Settings::Python::filename), and WSClean will call this for performing a major deconvolution iteration. The Python algorithm should then provide its best new estimate for the model image.