[ Basic Info | References | User Guide ]
Basic Information on mosmem
Task: mosmem
Purpose: Maximum Entropy deconvolution for a mosaiced image
Categories: deconvolution
MOSMEM is a MIRIAD task which performs a maximum entropy
deconvolution of a mosaiced image. Optionally it can also perform
a joint deconvolution of a mosaic and single dish image.
MOSMEM will also work correctly on a single-pointing observation
interferometric observation. In this case, it will be less efficient
than MAXEN, but it could be used when combining single dish data
with a single pointing.
MOSMEM spits out some information as it goes:
RMSFAC is the ratio (actual rms)/(theoretical rms).
It measures the residuals (i.e. the
difference between the dirty image and the model modified by the point
spread function). RMSFAC should converge to 1.
NormGrd is normalised gradient in the maximisation process.
Convergence requires this to be less than 0.05
Flux is the sum of all the pixel values in the model.
Key: map
One or perhaps two input dirty images (or cubes). These should have
units of Jy/beam. The first should be produced by INVERTs mosaic mode.
The optional second dirty map can be a single-dish image. It must
be on exactly the same pixel grid as the first image. If necessary,
use REGRID to make this so. If two inputs are given, then a joint
deconvolution of the two is performed.
Key: beam
One or perhaps two input dirty beams. The first, corresponding to the
first input dirty map, will be produced by INVERTs mosaic mode. There
is no default. The second dirty beam (which must be given if there
are two dirty map inputs) gives the point-spread function of the
single dish dirty map. This second dirty beam need not be the same
image size as the input dirty maps, and may be appreciably smaller.
This single-dish beam is assumed to be position-independent, but it
need not be symmetric.
Key: model
An initial estimate of the deconvolved image. For point sources,
giving a good initial model may help convergence. In principle,
this only helps convergence, but should not affect the final
solution. The model could be the output from a previous run of
MOSMEM or any other deconvolution task. It must have flux units of
Jy/pixel. The default is a flat estimate, with the correct flux.
Key: default
The default image. This is the image that the final solution will
tend towards. The final result will be influenced by this default
if the constrains that the data put on the solution are weak.
The default is a flat estimate, with the correct flux.
Key: out
The name of the output map. The units of the output will be Jy/pixel.
It can be input to RESTOR to produce a restored image, or alternatively
to MOSMEM, as a model, to continue the deconvolution process.
Key: niters
The maximum number of iterations. The default is 30.
Key: region
This specifies the region to be deconvolved. See the Users Manual
for instructions on how to specify this. The default is the entire
image.
Key: measure
The entropy measure to be used, either "gull" (-p*log(p/e)) or
"cornwell" (-log(cosh(p)) -- also called the maximum emptiness
criteria). Using the maximum emptiness criteria is not recommended.
Key: tol
Tolerance of solution. There is no need to change this from the
default of 0.01.
Key: q
One or two values (corresponding to the mosaic and single dish
observations). These give estimates of the number of points per beam.
MOSMEM can usually come up with a good, image-dependent estimate.
Key: rmsfac
MOSMEM must be able to the theoretical rms noise of the input dirty
map(s), and will, by default, attempt to reduce the residuals to have
the same rms as this. If the true rms noise is different from the
theoretical, you may give the factor to multiply by to convert from
theoretical to true rms noise.
The theoretical rms will usually be an optimistic estimate of the
true noise level. The true noise will be increased by calibration
errors, confusion, poorly understood distant sidelobes, etc.
The rmsfac factor gives some `fudge factor' (usually greater than 1)
to scale the theoretical noise estimate by. Either one or two values
can be given, with the second value corresponding to the single dish
input.
For a mosaic, the theoretical rms is position dependent, and is
determined from information save by INVERT (the mostable table).
For a single dish image, the rms is assumed to be constant across
the field, and given by the "rms" item in the image. If the single
dish input does not contain this item, then this must be added
before using MOSMEM. This is easily done: for image xxxx, use
puthd in=xxxx/rms value=????
where "????" is the rms noise in Jy/beam.
Key: factor
The flux calibration factor. This is only relevant when doing a
joint deconvolution of a mosaic and a single-dish image. It gives the
factor which the single-dish data should be multiplied by to convert
it to the same flux units as the mosaic. The default is 1. If the
``dofactor'' options is used (see below), MOSMEM solves for this
parameter.
Key: flux
An estimate of the integrated flux of the source. This parameter is
generally not useful if there is an input single dish image.
Giving MOSMEM a good value for the integrated flux
will help it find a good solution. On the other hand, giving
a poor value may do harm. Normally MOSMEM will NOT constrain the
integrated flux to be this value, but see the ``doflux'' option below.
The default is image-dependent for measure=gull, and zero for
measure=cornwell. A value can be given for each plane being
deconvolved.
Key: options
Task enrichment parameters. Several can be given, separated by
commas. Minimum match is used. Possible values are:
doflux Constrain the solution to have the correct integrated flux
(normally the integrated flux is not constrained). The
integrated flux is determined from the "flux" parameter or
(if no flux parameter is given) from the default image.
This option cannot be used if a single dish input map is
also given.
dofactor Solve for the flux calibration factor.
verbose Give lots of messages during the iterations. The default
is to give a one line message at each iteration.