This document is retained for historical interest
This documents discusses the schedule and implications
of the major SkyView hardware and software
upgrade which is currently underway.
Schedule and Hardware Configurations
At noon, May 29, 2007, the skyview.gsfc.nasa.gov
address was pointed to a pair of load-balanced machines with
new hardware and software. For a transition period the name
skys.gsfc.nasa.gov continued to point to the older
hardware and software.
address was moved at noon (EDT), Tuesday, June 12.
In this configuration the skys and skyview addresses
are interchangeable. They do not
point to different machines but to the same load-balanced cluster.
The new machines
will continue to support the old IDL software (though not in the
old locations) through September 21, 2007. The old interfaces
will be removed after that time.
In addition to the SkyView service itself,
the SkyView hardware has hosted a few
other services. These should continue to be supported unchanged under the
Transition to the new Java-based version of SkyView
The new machines use a new Java-based version of SkyView
This new version has many new features and capabilities
and supports almost all of the capabilities of the old system.
Subseqent sections discuss some of the changes and outline areas where
users may see changes in the system.
We hope that for most users the transition is completely seamless
other than the new (and we hope improved) user interfaces.
Please let us know if have problems or concerns.
Only a single Web interface is provided in the new version rather
than the three Non-astronomer, Baick and Advanced interfaces
in the old SkyView
. This interface
complex and advanced options easily, but only when needed.
The Non-Astronomer interface has been replaced by a discussion
of how lay users might use the standard features of the system.
If you find this interface confusing or would like additional
capabilities please send us a note.
were discovered for Internet
Explorer users after our initial transition.
We believe that these issues have largely been rectified. To support
in the longer term, a version of the interface which does
We have compared image results pixel by pixel for most of the
surveys. We have found
the results are identical or have differences due to differing
roundoff errors. Note that when resampling using the
nearest neighbor algorithm, roundoff errors will occasionally
cause a different pixel to be sampled.
For nearest neighbor resampling
the pixel values will normally be identical but may be a few pixels where
an adjoining pixel was sampled. For other resampling methods
one typically gets small differences in the values of all
By default on the Web users will get 4-byte real data as before.
Unchecking a checkbox allows the users to get 8-byte reals which
are what is used in all calculations.
There are substantial differences in the quicklook data.
Quicklook images are now provided as JPEGs rather than GIFs.
These are more
generally supported and often smaller. The underlying software
supports both JPEG and GIF images and we can provide GIF as
an option if there is a significant demand. GIF images
are supported in Batch mode requests (as well as TIFF, BMP and
In appearance the most significant
change is that by default images are now black and white.
Users can select any of the old and many new color tables if desired
but we feel that the black and white images most accurately
reflect the intensity scaling. Some novice users were confused about
the meaning of the colors in the old versions. The old system used the
Stern Special color table.
Users can invert any color table. Only gray-scale images could
be inverted in the old system.
The algorithms for logarithmic and histogram equalization scaling
are provided by ImageJ and are somewhat different than what is used
in the old SkyView system. A SQRT scaling is now provided as well.
The overlays may look slightly different. There are many more
options for contours. Smoothing is available as a separate
options for the contour maps and RGB overlays.
The color for plot overlays can be specified by the user.
When mosaicking component images of a survey together, a
more robust algorithm is used. This should eliminate small regions
that ran past the edge of the source images that were sometimes
seen in previous versions. The new algorithm should generally
find the 'best' image for each pixel in the image.
The old interface had a link which showed the user
which images were used in constructing an output
image. This information is now included in the
FITS header of the output file.
New Resampling Algorithms
A new set of 'high-order'
resampling algorithms is available. These
allow for use of Lanczos resamplers with kernels of a user specified
size, and for second to fifth order spline resamplers.
An efficient exact-area resampler has been developed to replace the
very slow and buggy triangular decomposition resampler used in the IDL
version. This is useful when exact conservation of flux is required
or where the original data is poorly sampled or incomplete.
Currently supported samplers include:
Nearest Neighbor (NN)
Bi-linear interpolation (LI)
Splines from 2-5 order (Spline or Spline3, Spline2, Spline4, Spline5)
Lanczos interpolation with a specified kernel size
(Lanczos, or LanczosN where N >= 2, the default is 3)
exact-area resampling using Sutherland-Hodges clipping (Clip).
The default is Nearest Neighbor but the Clip sampler is relatively fast
and users should seriously consider using it.
Faster Image Generation
The Java-based system uses less memory than the IDL version, so that larger
images can be generated and large images are generated more quickly.
For most surveys images of up to at least 5000x5000 pixels
can be generated while the
previous version was limited to about 2000x2000 pixels. Please
only ask for large images if you need them.
New Edge Adjustment Algorithm
A new edge adjustment algorithm
is available whose invocation can
be controlled by the user. Edge-adjustment attempts to hide the visible
boundaries between source images caused by differing background levels in
the source image. The older version used a median test where the medians
of images were matched. The new version matches medians along
the edges associated
with pairs of images and can be turned on or off (but is turned on by
default for the DSS and 2MASS).
Support for catalogs has been substantially upgraded. Users
can request data from HEASARC and Vizier catalogs as well as
any URL that supports the Virtual Observatory Cone Search Protocol.
HEASARC and Vizier catalog names or the Cone Search URLs can be entered
in a text box, and HEASARC catalogs are also available from a menu.
Users can now add criteria to their selection, e.g., 'vmag < 15' to
select only the brightest objects.
Catalog query results are now stored in a separate page which can be
viewed or downloaded and is linked to the image display.
The catalog query results are no longer appended as a FITS extension to
the image file. Nor are there links to a similar (though not identical)
HEASARC query that was available from the results page. Please let
us know if these are useful capabilities to restore.
Deleted and added surveys
A few surveys have been deleted as obsolete:
the very low resolution UV surveys and the PSPC (old) survey.
It is much easier to add surveys and the new code has the
latest (DR5) release of the SDSS, the IRIS improved IRAS data,
and the Very Low-frequency Sky Survey (VLSS) were not available in the old version.
Survey are now displayed in 5 regime areas. Currently within
each regime surveys are displayed in alphabetical order. Previously
some attempt had been made to display surveys in frequency order
The internal directory structure used for the creation
of FITS and quicklook data has changed. Users of the undocumented
RETURN=filename option may need to modify their code.
The Image center at pixel center option is no longer available.
That shifted the center of the image by half a pixel when the number
of pixels was even. The underlying code supports the offset setting
which can be used to emulate this option if needed.
The survey documentation now repeats survey information for each
band rather than combining related surveys. We may address
this in the future. E.g., the IRAS survey is described in full
for each of the 4 bands rather than having the 4 bands refer to
the same document.
The SkyView Java applet has been ported but will no longer
be actively supported. We suggest using Aladin or OASIS.
Direct all questions regarding SkyView to our Feedback Page
SkyView has been developed with generous support from the NASA AISR
and ADP programs (P.I. Thomas A. McGlynn)
under the auspices of the High Energy
Astrophysics Science Archive Research Center (HEASARC) at the NASA/
Astrophysics Science Division.
We gratefully acknowledge the support of NASA and contributors
of SkyView surveys.