Greetings,
In this memo I will try to set the stage for a productive mini-workshop
on SXT image deconvolution. Please read, comment, and respond as
appropriate. Under "PAST EFFORTS" I am relying on my memory so please
forgive and correct me if I've gotten it wrong.
PURPOSE
To define a path toward a trustworthy and practical method of
correcting SXT images for the smearing effects of the instrument point
response function and scattering wings (hereafter simply termed PSF). This
includes (1) an evaluation ofwhat we know about the PSF and how best to
characterize it for deconvolution purposes, and (2) to discuss and compare
deconvolution codes, techniques and experience.
There are many reasons to desire a robust, reliable, and speedy
deconvolution code for SXT. My personal objective in organizing this
meeting is the hope of coming up with a feasible way to deconvolve the PSF
from our new SXT Science Composite (SSC) images in order to improve the
photometric accuracy, especially of the faint corona. Others may be more
interested in correction of partial frame images (PFI) in order to improve
the accuracy of filter-ratio temperature analysis of flares.
We are not concerned here with super-resolution techniques, except
as they may provide useful insights into the inflight point response
function.
PAST EFFORTS
A number of workers, e.g., Hara and Akiyama, LaBonte, Foley and
Lemen, McKenzie, have developed routines for PSF correction. The
approaches vary in how they handle temperature (wavelength) dependence of
the core and wings of the PSF. Several people, most recently Hara and Akiyama,
have attempted to estimate the scattering through different filters from the data.
A flare mode experiment has been running for some time in an attempt to better
define this important parameter.
Super-resolution work, for which the experimental PSF is not required,
has been done by Sylwester's group (ANDRIL), Roumeliotis, and Karovsk.
PROBLEMS
- We need an "adequate" temperature-independent PSF for full-frame (FFI) deconvolution.
I fear that full rigor, with all of the PSF pixel-location and temperature dependence
included would present an unmanageable computational deconvolution task for SSC production.
- For flare work with PFIs better calibration of the scattering wings is needed for the
different filters. For PFIs the core of the PSF appropriate to the location of the PFI
should be used. SXT_PSF.PRO provides this capability.
- In the case of the thicker filters, where the entrance filters provide a significant
part of the entrance opacity, the scattering wings are not cylindrically symmetric --
because of increased flux onto the mirror from failed sections of the entrance filters.
This pattern of assymetry changes with each entrance filter failure event. Present
software does not reflect these assymmetries.
- Flux must be conserved and areas of low signal (negative values in the image because
of statistics) must be properly handled.
- The experimental error must be carried through the deconvolution process.
PROVISIONAL PLAN
This workshop will follow the path of the very successful HXT image synthesis
workshop.
I propose that the first day of the meeting be devoted to informal presentations.
The morning to PSF discussions and the afternoon to deconvolution techniques and
experience. We are fortunate that Leon van Speybroeck, of AXAF experience and fame,
has agreed to join us. On the second day we may break up into experiment teams to
follow up, on LaBonte's computers or your laptops, ideas and arguments from the first
day's work. The last 2 hours of the second day will be reserved for discussion and
wrap-up.
David McKenzie and Barry LaBonte are collaborating on a workshop
web site. Please send abstracts of what you plan to present or issues you plan to
raise to David McKenzie
for posting to this site. Especially, please
list events or data sets that you are working with so that others can have a look
at these before the meeting. We'll confirm the URL in a later message.
REFERENCES - PSF & scatter
[These have been moved to a separate page and expanded.]
Loren Acton,
December 7, 2000