Package: stsdas.analysis.restore
Help file updated: Mar93
opt=sys help pages is recommended.
Psets are used to input information to the task. There are separate psets dedicated to the Point Spread Function (PSF), filter parameters, signal model parameters, noise parameters and an optional low-pass filter.
The PSF can be input from either a separate file,
or from the input image to be deconvolved. In either case, the PSF need not be
centered in the field. You may specify the center coordinates px0
and py0 of the PSF in the PSF image section, or, leaving either one,
or both, as INDEF, instructing the task to find them automatically. In
this case, the pixel with maximum intensity in the PSF image section will be
taken as the PSF center.
A circular mask with low leakage can be used to
isolate a suitable star in a crowded field. The radius is specified by
task parameter mask (in pixels), and must be wide enough to not affect
the point source image, but narrow enough to eliminate field stars. The
PSF must be background-subtracted before using it with this task. When
extracting the PSF directly from the input image, this image must have
the background already subtracted.
Besides the PSF, the Wiener filter needs additional information about the undegraded image and noise statistics. This information is derived using the signal model and noise parameters in the corresponding psets. The meaning of these parameters is described in detail in the system-level help by typing "help wiener opt=sys".
After deconvolving the input image, the task may optionally apply a
low-pass filter to the deconvolution result. This low-pass filtering
step is particularly important in the case of inverse filter deconvolution.
Use task parameter lowpass in the appropriate pset to specify filter
type, and fwhm parameter to specify the width of the smoothing kernel in
image pixel units.
A circularly symmetric or an axially symmetric (in frequency
domain) filter can be selected, or you can specify that no filter is to
be used at all.
The square filter has proven to be effective at eliminating rings around the
deconvolved point sources. The filter is an apodizing function which
minimizes side-bands.
The output deconvolved image is normalized to the same total flux as in the input image.
Any image size and aspect ratio may be used as input for this task. The input and PSF images may have different size axis. However, for efficiency purposes, the input image axis sizes must be even. If the input image contains an odd-size axis, the last row or column will be stripped off. The Fast Fourier Transform algorithm used by this task is faster when the axis sizes are composite numbers (i.e., non-primes), faster yet when rich in factors of 2, and even faster with exact powers of 2. One limitation in input image size may come from the amount of memory available to the task. For a 512 X 512 image, the task will use about 5 MB of memory for data storage.
The filter function itself can optionally be written to an output image file.
This image will contain the amplitude squared of the complex filter
function, eventually multiplied by the low-pass filter function.
The filter task parameter specifies the file name to be created; if
left as a null string (""), no filter output is generated. Similarly,
if task parameter output is set to null, then no deconvolved image is
generated. This option saves some execution time, and is useful when
running the task repeatedly, for example, when you are interested in the
filter function form only.
The task can process an image template or list of files passed to
input. In this case, output is either a matching list of images or a
directory. Both psf and filter must be always single images.
History records with filter parameter information are appended to the output image header.
Typical CPU times (Sparc 2, 512 X 512 image): 45 sec for inverse filter; 90 sec for geometric mean filter with external image signal model.
psf image section. If either
one, or both, are left as INDEF, the task will locate the maximum pixel
value in the PSF image section, and use its coordinates instead.
Filter type to be used.
Theoretical (undegraded) signal model to be used in Wiener and geometric
mean filter forms. When none of the first five options is entered, the
task attempts to open and read an IRAF image with the name given by the
signalm parameter.
signalm=input). If set as INDEF, signal model will be computed
from input and psf images.
Noise model.
Type of low-pass filter.
lowpass value.