restore image resolution using deconvolution techniques
arestore infile psffile outfile numiter [psf_x_center] [psf_y_center]
[method] [clobber] [verbose]
`arestore' restores images degraded by a blurring function
by using one of several deconvolution
techniques.
An image of astronomical source observed from space
can be mathematically described as a convolution of the
intrinsic brightness distribution of the source with a
blurring function. The blurring of the source image may be
introduced by the telescope optics and
the detector, including the limited size of detector
pixels, and by an additional
noise component. The blurring function, often called Point
Spread Function (PSF), can be expressed as
a convolution between the telescope image of a point source
and the point response function of the imaging detector.
Various deconvolution methods can be applied to restore
the degraded resolution in the image, assuming a known
blurring function (PSF).
We have currently implemented the Lucy-Richardson
deconvolution algorithm (method=lucy).
This algorithm requires that the spatial/spectral
characteristics of the input PSF are very well known (from
observations or from simulations).
The Lucy-Richardson method performs well for bright
point sources.
However, for low-count observations
and especially for extended low-intensity sources,
the algorithm has a tendency to concentrate the intensity
into points
rather than restoring the smooth low-intensity regions in the
image.
As in the case of many other deconvolution techniques,
this method does not
produce reliable convergence and
uncertainty information. Users should be very
cautious in interpreting and evaluating the results of the
deconvolution, especially when
applied to
faint and/or extended sources (few hundred counts).
arestore my_img.fits my_psf.fits my_outfile.fits method=lucy numiter=100
Will deconvolve an image (my_img.fits) using a PSF
from a file supplied by the user (my_psf.fits) using
the Lucy-Richardson deconvolution (lucy) method with 100 iterations.
arestore my_img.fits my_psf.fits my_outfile.fits psf_x_center=130
psf_y_center=128 numiter=50 method=lucy
Will deconvolve an image (my_img.fits) using a PSF
from a file supplied by the user (my_psf.fits) with the PSF
center defined by its logical coordinates:
x=130 pixels, y=128 pixels and using 50 lucy iterations.
Input response file (eg. PSF).
The PSF file that describes the instrumental smearing of the
data.
To carry out deconvolution of Chandra data with arestore
user needs to provide an accurate model of the PSF.
The HRMA PSF can be
simulated with ChaRT and the instrument effects, including
detector pixelization, can be simulated with MARX. Before applying
arestore, users should consult the
ChaRT caveats page
and review the PSF calibration
information.
See the
ChaRT pages on the CXC web site
for more information on Chandra PSFs.
The restored output file.
Number of iterations
The number of iterations to perform; typical numbers vary by
deconvolution technique.
For "lucy" deconvolution, users should run few dozen to
a hundred iterations
to avoid an overly grainy reconstruction.
X center of the PSF file.
The location of the center of the PSF in the X-direction. This
should be in image coordinate, eg from 1 to N where N
is the number of
pixels in the X-direction.
A value of 'INDEF' will use the value (floor(N/2)+1).
Y center of the PSF file.
The location of the center of the PSF in the Y-direction. This
should be in image coordinate, eg from 1 to N where N is the number of
pixels in the Y-direction.
A value of 'INDEF' will use the value (floor(N/2)+1).
Deconvolution method
Determines which deconvolution algorithm to use.
Currently, the only
available option is lucy, for the Lucy-Richardson deconvolution
algorithm.
Remove output file if it exists
The Lucy-Richardson deconvolution algorithm implemented in
arestore was developed
independently by Lucy (1974, Astron. J. 79, 745) and
Richardson (1972, J. Opt. Soc. Am., 62,55).
This is an iterative technique that applies maximum
likelihood fitting using nonnegative data corrupted with
Poisson noise, eg. Chandra data with photon-counting statistics.
The advantage of the Lucy-Richardson algorithm is that it is
relatively fast and easy to implement, and performs well for sources
which are concentrated in a few points.
|
![[CIAO Logo]](../imgs/ciao_logo_navbar.gif){kind=logo}