NRAO Home > CASA > CASA Toolkit Reference Manual
deconvolver.mem - Function

3.1.2 Make the mem image


Description

Makes a mem image using the Cornwell-Evans algroithm, using either maximum entropy (entropy) or maxmimum emptiness (emptiness). The maximum entropy algorithm is the default. You can restart a MEM deconvolution on an existing model image, but the alpha and beta parameters are not yet saved.

Mask images can be used to restrict where the algorithm puts flux. A prior, or bias, image can provide a priori information to the algorithm and effectively limit the support as well as a mask. The prior image can be constructed by smoothing an existing estimate for the brightness distribution and clipping. Any pixel values below 1e-6 will be clipped to this level, so zero or negative pixels will not cause problems.

Currently, only one Stokes parameter may be deconvolved at a time. Stokes I images can be deconvolved with either maximum entropy or maxmimum emptiness. Stokes Q, U, or V should be deconvolved with maxmimum emptiness, which permits negative pixel values. Joint polarization MEM deconvolution is planned for the future.

The mem entropies possible are:

entropy
The smoothness of the image, relative to some prior (also called default or bias) image is maximized. The functional form of the entropy is H = Iln(I∕M), where I is the mem image brightness and M is the prior image. As the prior image is positive definite, the entropy constrains the mem image pixels to be positive, hence only stokes I can be imaged.
emptiness
The number of pixels with absolute value of the flux greater than the noise level is minimized. This treats positive and negative pixel values equally, so it is appropriate for any Stokes image.

This MEM algorithm works in the image plane (ie, is ignorant of visibility data), but performs the convolution by multiplication in the Fourier plane. Not to be confused with this usage of the term ”image plane”, some problems are ”image plane” problems, such as a single dish performing On-The-Fly mapping. Independent noise is added at each integration as the beam sweeps over the object (ie, in the image plane). This can lead to a noise signal at non-physically large spatial frequencies. This non-physical signal can be removed by convolving the residual image with the PSF. Also key to this problem is that the PSF is of finite extent, permitting the deconvolution of nearly the entire dirty image rather than just the inner quarter. These options are accessed by setting imageplane to T.

Arguments





Inputs

entropy

entropy to use

allowed:

string

Default:

emptiness

entropy

niter

Number of Iterations, set to zero for no MEMing

allowed:

int

Default:

20

sigma

Noise level to try to achieve

allowed:

any

Default:

variant 0.001Jy

targetflux

Total image flux to try to achieve

allowed:

any

Default:

variant 1.0Jy

constrainflux

Use targetflux as a constraint? (or starting flux)

allowed:

bool

Default:

false

displayprogress

Display progress

allowed:

bool

Default:

false

model

Name of input/output model image

allowed:

string

Default:

prior

Name of prior (default) image used for mem

allowed:

string

Default:

mask

Mask image restricting emission (all pixels 0 or 1)

allowed:

string

Default:

imageplane

Is this an image plane problem (like single dish)?

allowed:

bool

Default:

false

async

Run asynchronously in the background?

allowed:

bool

Default:

false

Returns
bool

Example

 
 
deco.mem(entropy=’entropy’, niter=30, sigma=0.01, targetflux=10.0,  
model=’3C273XC1.mem.image’, prior=’3C283XC1.prior’)  
 

__________________________________________________________________


More information about CASA may be found at the CASA web page

Copyright 2016 Associated Universities Inc., Washington, D.C.

This code is available under the terms of the GNU General Public Lincense


Home | Contact Us | Directories | Site Map | Help | Privacy Policy | Search