deconvolve(imagename, model='', psf=[''], alg='clark', niter=10, gain=0.1, threshold=0.0, mask='', scales=[0, 3, 10], sigma=0.0, targetflux=1.0, prior='')[source]

Image based deconvolver

[Description] [Examples] [Development] [Details]


deconvolve performs minor cycle deconvolution directly on a dirty image. No MS is required.


ALERT: deconvolve is defunct. The task deconvolve is currently being refactored and will be available again in an upcoming release. For the moment use tclean (although tclean requires an MS at this stage), or please use deconvolve in an older version of CASA (v4.7.2 or earlier).

Parameter descriptions


The input dirty image that will be deconvolved.


The output cleaned image containing a deconvolved point model.


Can be either be an image (e.g., a psf that tclean has calculated), or a list of values that define a Gaussian, e.g., psf=[‘3arcsec’, ‘2.5arcsec’, ‘10deg’] defines a Gaussian with ‘3arcsec’ as the major axis, ‘2.5arcsec’ as the minor axis, and a position angle of 10 degrees.


The clean algorithm to use by deconvolve. Several algorithms are available to deconvolve an image with a known psf (dirty beam), or a Gaussian beam. The algorithms are ‘clark’ (default) and ‘hogbom’ clean, ‘multiscale’ clean and a Maximum Entropy ‘mem’ clean. Details on the algorithms are given in the Synthesis Imaging chapter in the page describing Deconvolution Algorithms.

alg=’multiscale’ expandable parameters


A list of scales in units of pixels (see the deconvolve examples tab).

alg=’mem’ expandable parameters


This parameter allows the user to input an expected noise value, which will allow for a faster convergence.


The estimated total flux in the image.


A starting model to be used by deconvolve. If no prior is provided, a flat image will be used.


The number of iterations to perform on the image. Default: 10


Sets the CLEAN gain parameter. Default: 0.1


Sets the lower level below which sources will not be deconvolved.


The image mask to limit the region of deconvolution.


Deconvolve the dirty image ‘mydirtyimage.image’ with a dirty beam (psf) ‘mydirtyimage.psf’. No MS is required as only minor cycles are performed. We are using the ‘multiscale’ clean algorithm with scales of 0, 3, and 10 pixels. The stopping criteria are either 10000 iterations, or an RMS threshold of 10mJy:

model='mycleanimage.image', psf='mydirtyimage.psf',
alg='multiscale', scales=[0,3,10], niter=10000, gain=0.1,

No additional development details

Parameter Details

Detailed descriptions of each function parameter

imagename (string) - Input image to deconvolve
model (string='') - Output image containing deconvolved point model
psf (stringArray=['']) - Point spread function (dirty beam)
alg (string='clark') - Algorithm to use (clark, hogbom, multiscale, mem)
niter (int=10) - number of iteration in deconvolution process
gain (double=0.1) - CLEAN gain parameter
threshold (double=0.0) - level below which sources will not be deconvolved
mask (string='') - image mask to limit region of deconvolution
scales (intArray=[0, 3, 10]) - scale sizes (pixels) to deconvolve
sigma (double=0.0) - mem parameter: Expected noise in image
targetflux (double=1.0) - mem parameter: Estimated total flux in image
prior (string='') - mem parameter: prior image for mem search