astro imaging

Thoughts on Super-Resolution
Planetary Imaging
Obtain detailed images of the planets under less-than-ideal observing conditions. By Antnio Jos Cidado

L
ets admit it who has never with a large variety of instruments, such Above: The CCD camera is ideally suited for
dreamed of obtaining a sharp as apochromatic refractors, Newtonian taking high-resolution images of the planets.
image of the famous Cassini Di- reflectors, and Maksutov- or Schmidt- Achieving optimum performance, however,
vision in Saturns rings or cap- Cassegrain telescopes, coupled to web- requires you not only to fine-tune your tele-
turing the pastel colors of the ever- cams or video, digital, or CCD cameras scope setup and carefully match the CCD de-
changing Jovian atmosphere? These (S&T: November 2000, page 142). In ad- tector to your optical system but also to
goals, which are very appealing from an dition, there is plenty of outstanding quickly capture as many individual sharp im-
aesthetic point of view, are readily acces- ages as possible on a given night. The raw
sible to a large number of amateurs images can then be digitally combined and
since imaging the bright planets doesnt enhanced on a computer. The author obtained
require very large aperture telescopes or these highly detailed views of Jupiter, Saturn,
expensive CCD cameras. Even better, and Mars (shown not to scale) from a light-
planetary imaging is well adapted to polluted site under favorable seeing condi-
urban sites since its virtually unaffected tions. All images in this article were made with
by light pollution. Last, a high-quality a 25-centimeter (10-inch) f/10 Meade LX200
planetary image, together with such basic Schmidt-Cassegrain telescope and a Santa
information as date, time, and imaging Barbara Instrument Group ST-5C camera.
conditions, is a valuable record of a plan-
ets activity that may be of interest to Left: Antnio Cidado with his 2-year-old son,
amateur or professional researchers Pedro. Cidado also uses a vintage 11-inch
studying transient planetary phenomena. Celestron telescope on an east-facing bal-
High-resolution planetary imaging is cony. During an imaging session he rolls the
greatly facilitated by the equipment and scope from a small storage shed and aligns
software currently available to amateurs. its platform on permanent dimples he has
Excellent results have been obtained made on the concrete floor.

2001 Sky Publishing Corp. All rights reserved. Sky & Telescope December 2001 127
astro imaging

Processing planetary images with image-editing programs, such as Adobe Photoshop, or with To capture the finest possible planetary
dedicated CCD-camera software, can greatly improve the quality of the final composite. These images you also need a telescope with
portraits of Jupiter were obtained on December 11, 2000, with an AO-2 adaptive-optics acces- clean, high-quality optics that are carefully
sory from Stellar Products. Ten raw exposures each were taken through red, green, and blue aligned. And you need a solid and polar-
(RGB) filters and then registered and averaged to produce the composite on the left. For the aligned mount, an accurate telescope
composite on the right the author applied the unsharp-mask tool of SBIGs CCDOPS for DOS to drive, and a good-quality CCD camera. Al-
the raw images before combining them. South is up; the planet measures 48 across. lowing the telescope to cool down to am-
bient temperature and taking great care to
software, both commercial and free, that The Challenge of achieve perfect focus are also crucial. You
allows us not only to efficiently control Atmospheric Distortion can also use powerful graphics-editing
our cameras but also to process our To reach the spatial resolution that is theo- programs such as Adobe Photoshop to en-
images in order to squeeze out every bit retically possible with your telescope setup, hance your images. All of these are quite
of information from them while mini- an extremely stable atmosphere is required. important, but in most cases the result is
mizing electronic noise. Anyone with Unfortunately, such an ideal observing going to be limited by the atmosphere.
access to such resources should definite- condition is rare. Even the best ground- Professional astronomers use sophisti-
ly try his or her hand at imaging the based images are still limited in resolution cated adaptive-optics technology to coun-
planets. by our planets turbulent atmosphere. teract the blurring effects of the atmos-

There are three important guidelines worth keeping

in mind when you are imaging the planets. First, use
the optimum image-sampling
method; second, use exposures
that produce raw images with a
good signal-to-noise ratio, or S/N;
and third, combine separate raw images to further
improve the S/N of the final processed image.
Jupiter and one of its satellites, Ganymede (upper right), as imaged
on January 7th this year. By applying the super-resolution technique
discussed by the author, ultrafine planetary details can be captured
by the CCD camera. Note the dark marking visible on Ganymedes 1.8
disk (inset). South is up; the planet measures 46 across.

128 December 2001 Sky & Telescope 2001 Sky Publishing Corp. All rights reserved.
phere (see the October issues cover story
on page 30). For amateurs the solution is
not to image only under the best seeing
conditions but instead to try to obtain the
best possible images under available see-
ing conditions. This is the main focus of
my article, and I hope readers will realize
that there are various ways of at least par-
tially beating the atmosphere and produc-
ing decent images routinely.

Useful Guidelines
There are three important guidelines
worth keeping in mind when you are
imaging the planets. First, use the opti-
mum image-sampling method; second,
use exposures that produce raw images
with a good signal-to-noise ratio, or S/N;
and third, combine separate raw images
to further improve the S/N of the final
processed image. In simple terms, S/N is
the ratio of the desired photons from the
target planet to the unwanted photons
from background light pollution and
noise generated by the camera itself. The
higher the S/N value, the better the qual-
ity of the image.
The easiest way to get crisp raw im-
ages is to experiment with integrations
(exposures) that are short enough to
freeze atmospheric motion. You should Advertisement
also try to obtain the images as quickly
as possible to increase the probability of
capturing enough good images in a given
time period. This is particularly impor-
tant when you are combining images of
rapidly rotating planets such as Jupiter.

Sampling the Image

To obtain the highest spatial resolution
in a given telescope-camera combination
the pieces of equipment have to be prop-
erly matched in order to yield the best
possible image sampling. According to
the Nyquist sampling criterion, this oc-
curs when at least two CCD pixels cover
the finest detail resolved by the telescope.
If you use fewer than two pixels youll
undersample your data and lose infor-
mation permanently. For example, a tele-
scope that resolves 0.5 arcsecond must be
equipped with a CCD camera whose pix-
els span 0.25 arcsecond. Conversely, us-
ing more pixels oversamples the image
and doesnt increase the amount of in-
formation recorded. It only spreads light
over more pixels, lowering the S/N and
reducing the images overall quality.
In practice, you reach this optimum
setup by working at the correct focal
length (either adding a Barlow lens or

2001 Sky Publishing Corp. All rights reserved. Sky & Telescope December 2001 129
astro imaging
Matching a CCDs pixel size to the telescopes resolution is essential for obtain-
0.5 30 ing the best results. This nomogram developed by Sky & Telescopes Roger Sin-
400 nott makes simple work of determining the relationships between image
KAF-1001E

y
tar
and SITe chips 20
scale, effective focal length, and pixel size. You can connect any two values

ne
200 with a straight line to get the third. For example, in order for a 9-micron pixel

Pla

3000 4000
KAF-0261E
1 to cover 112 of sky you need a focal length of about 1,300 millimeters (50
15 inches). Under typical amateur observing conditions, you will produce excel-
verage per pixel (arcseconds)

Focal length (millimeters)

14

Focal length (inches)

100
13 lent results with pixels that cover about 112 to 212 of sky, while lunar and

1500 2000
planetary work will greatly benefit from image scales as small as 14 per pixel.

Pixel size (microns)

80 12
70
60 11
Raw Images TC-255 10 pled images, you can obtain a super-
50
KAF-0401E
Digitally combining the in- resolution effect, partially overcoming
9
D e e p sky

1000

40 and KAF-1602E dividual raw images ob- the limitations of your less-than-ideal
2 30 TC-237
8 tained in a short time peri- seeing conditions.
od will increase the signal Some time ago I started to use a man-
Sky co

KAF-3200E
7
20 and decrease the noise of ual image-processing routine, wherein a
500

6 the final image, but this series of raw images are registered to-
3 5
procedure is effective only gether at subpixel accuracy and are sub-
when the raw images sequently averaged. For precise registra-
4 themselves are of good tion, I enlarge the raw images with pixel
5
quality to start with. So it interpolation and then crop them as ac-
makes sense to use only curately as possible using the planetary
your sharpest images when limb as a reference. I perform the whole
focal reducer to increase or decrease the making the composite. procedure with the same software I use
scopes effective focal length, respectively) In practice, you can obtain good S/N to control my CCD cameras, namely,
and by using a camera whose pixels have with an exposure time that allows the Santa Barbara Instrument Groups CCD-
the most appropriate size. To calculate CCD pixels to be filled up to a value of OPS for DOS.
your cameras image scale, divide its pixel about two-thirds of their saturation Interestingly, I later found out that an
size in microns by the telescopes focal level. This can be difficult to achieve, identical manual procedure had been
length in millimeters and multiply the re- however, when you are doing imaging used by Tim Parker of NASAs Jet Pro-
sult by 206. For example, a camera with 9- through color filters, which further re- pulsion Laboratory with Adobe Photo-
micron pixels attached to an 8-inch tele- duce light reaching the CCD and force shop to produce the fantastic super-reso-
scope working at f/20 (4,064 millimeters you to use longer exposures. lution mosaic panoramas and 3-D
effective focal length) will span approxi- From my point of view, the best solu- anaglyphs from the Mars Pathfinder
mately 9/4,064 206 = 0.5 arcsecond. tion is to use an image scale that allows lander camera. The excellent work of
For a planet like Jupiter, 100 to 200 short exposures and, if necessary, to ac- Parker, who is also an amateur astrono-
pixels across the planetary disk are suffi- cept some degree of undersampling of mer and planet observer, can be seen at
cient. This corresponds to pixel coverage the images. Fortunately, by properly http://mars.jpl.nasa.gov/MPF/parker/
of roughly 0.5 to 0.25 arcsecond. combining such individual undersam- anaglyph.html. Currently, an automated

For the authors 10-inch telescope to achieve

Pixel size = 0.25 arcsecond Ideal Situation
optimum image sampling, his CCD camera
should have an image scale of 0.25 per pixel
(green). One of his setups, however, produces
undersampled images (yellow). When imag-
ing a stars 0.5 disk with an SBIG ST-6 cam-
era (roughly 25-micron-square pixels), he
gets either 0.75- or 1.5-square-pixel pat-
tern, depending where the image falls on the
Pixel size = 0.75 arcsecond Actual Situation CCD array. In this example, the best possible
resolution attainable with his camera is
Sharpen and Coregister
then enlarge and average 0.75 (a single pixel). To improve the images
2 with pixel definition, the author digitally processes and
interpolation
combines the raw images. While this process
produces a larger image with pixels still hav-
ing their original size, in practice it is equiva-
lent to emulating an image of the same size
but with smaller pixels (half the original size).
The final image retains the original sampling
0.5 arcsecond
of 0.75 but looks as if it were made at some-
0.75 arcsecond
what higher resolution.

130 December 2001 Sky & Telescope 2001 Sky Publishing Corp. All rights reserved.
 These unfiltered images of Saturn were ob-
Original Sampling tained with the ST-5C on September 8, 1998,
at an image sampling of 0.4 per pixel. Dur-
ing those nights of very stable seeing condi-
tions, this sampling may be insufficient to
capture all the planetary details provided
by the telescope, unless the author applies
the super-resolution processing technique
he describes in the text. Here the original
sampled image is an average of several co-
registered (nonenlarged) raw images. Much-
improved results are obtained when the
2 Sampling
individual raw images are enlarged (2 sam-
pling) prior to being averaged. The insets
highlight deliberately exaggerated, unsharp-
masked close-ups of the rings to show Ring
As Encke Division. South is up, and the planet
(with the rings) measures 44 across.

procedure to obtain the super-resolution As mentioned earlier, digitally combin- ets rotation. On these occasions, which
effect in both deep-sky and planetary ing a series of raw images is important to are more frequent that we would like,
images, called drizzle, is available in the produce a final processed image with marked distortion of the planet is also
outstanding freeware program IRIS, de- good S/N. However, seeing can sometimes likely to occur due to bad seeing, and this
veloped by French amateur Christian be so erratic that, in practice, only one de- fact alone would argue against averaging
Buil. This software can be downloaded at cent raw image can be obtained during the images. So it seems that the only pos-
www.astrosurf.com/buil/. the window of time allowed by the plan- sible choice for these single raw images

Advertisement

2001 Sky Publishing Corp. All rights reserved. Sky & Telescope December 2001 131
astro imaging
Red Blue Green (average of red + blue) R(G)B

This illustration shows how a color view of Mars was assembled from images obtained on April 28, 1999. Since most of the color information
comes from images taken through red (R) and blue (B) filters, the author synthesized the green (G) component by averaging the R and B im-
ages. Note the red planets north polar cap and dark features (Mare Erythraeum at upper left and Mare Acidalium at lower left on the disk).
South is up, and the planet measures 16 across.

(which largely surpass all the others in will be covered by several pixels. By com- RGB Imaging
quality) is to process them individually. bining simple unsharp masking with an The conventional way of producing true-
Youll end up with rather poor results if image-processing technique called wavelet color CCD images is to take a series of ex-
your single raw image was undersampled analysis, its possible to target and effi- posures through red (R), green (G), and
in the first place, but for properly sampled ciently eliminate the major noise source blue (B) filters and then combine them
or oversampled images the results can be in our single raw image. A detailed digitally. Since planets rotate, in some
quite satisfactory. And the main reason explanation of the procedure can be cases quite rapidly, these three exposures
for this is that the annoying high-frequen- found on my Web page at www.astrosurf need to be obtained as quickly as possible
cy noise we want to get rid of exists most- .com/cidadao/single.htm. I use Buils free- in order to prevent heavy misalignment
ly at single-pixel scale, while the smallest ware IRIS to perform both unsharp mask- between the R, G, and B components.
planetary detail in an oversampled image ing and wavelet analysis. This is particularly true when individual

Advertisement

132 December 2001 Sky & Telescope 2001 Sky Publishing Corp. All rights reserved.
raw images are to be averaged to produce
the final image for each color component.
Slight misalignments between the compo-
nents can be easily compensated for, as
described on my Web page at www
.astrosurf.com/cidadao/correction.htm,
but in other cases color fringes will re-
main along the planets limb or near its
central meridian. Unfortunately, poor see-
ing conditions can severely limit the num-
ber of nights when RGB exposures can be
obtained at a rapid enough pace.
Planetary observers have long known
that the most useful information origi-
nates from just the R and B components.
Red-filtered images provide surface data,
as in the case of Mars, and allow us to see
deeper into the atmosphere of gaseous
planets like Jupiter or Saturn. On the
other hand, blue-filtered images show
plenty of atmospheric details, such as
Martian ice clouds and hazes, and allow
us to visualize higher cloud structures
on the gaseous planets. In short, images
through these filters provide complemen-
tary information, while those through the
G filter just stay somewhere in between
the other two. By properly combining red
and blue exposures its possible to synthe-
size a green component that is suitable
to produce true-color-like R(G)B im- Advertisement
ages. Thus, theres no need to waste time
attempting to obtain the nonessential G
component. Its better to use those pre-
cious moments of good seeing to try to
obtain more red and blue raw images, or
simply to obtain them more rapidly. The
result is that the S/N of the final image
will be better and the color misalignment
will be lower, respectively.

My Setup
Finally, Id like to give you a quick
overview of my observing site and
equipment. I do all my planetary imag-
ing from my apartment in Oeiras, Portu-
gal, about 20 kilometers (12.5 miles)
west of Lisbon. Its a considerably light-
polluted urban site, about a kilometer
from the Atlantic Ocean, and it is often
characterized by poor seeing conditions
during the summer. Some periods of
average-to-good atmospheric stability
also occur, mainly in spring and autumn.
My main telescope is a pier-mounted
25-centimeter (10-inch) Meade f/10
LX200 Schmidt-Cassegrain telescope lo-
cated in a roof-window observatory. Un-
fortunately, a lot of fireplace chimneys
surround my window, which consider-
ably degrades my local seeing conditions

2001 Sky Publishing Corp. All rights reserved. Sky & Telescope December 2001 133
 during winter. As my views to the north up new horizons for my imaging capabil-
and east are blocked by the roof, I also ities. Some time ago I acquired a used
use a vintage 11-inch Celestron telescope AO-2 adaptive-optics device from Stellar
semipermanently mounted in an east- Products. (It was reviewed in the April
facing balcony (see page 127). 1993 issue, page 52, but unfortunately
I started shooting the planets with this product is no longer being manufac-
SBIGs ST-6 camera equipped with a tured.) This image-stabilization device
motorized color filter wheel for RGB has allowed me to go beyond convention-
imaging. I now also own two other SBIG al unfiltered and RGB imaging and
cameras an ST-5C (with 10-micron- do time-lapse animations of planet rota-
square pixels and an internal RGB color- tions (see www.astrosurf.com/cidadao/
filter wheel), which has long been my animations.htm).
workhorse camera, and an ST-237 (with You can contact me at Rua D. Antnio
7.4-micron-square pixels), which is also a Lus de Menezes, No. 9, 3A, 2780-048
very good lunar and planetary camera. I Oeiras, Portugal; phone: +351-214412192;
also do specialized imagery with ultravi- e-mail: a.cidadao@mail.telepac.pt.
olet (320- to 390-nanometer) and in-
frared (830- to 1,000-nm) filters, both A physician and assistant professor of medi-
from Murnaghan Instruments, and a nar- cine, Antnio Jos Cidado is a member of
rowband methane (889-nm) filter from the Associao Portuguesa de Astronmos
Custom Scientific. Amadores, the Association of Lunar and Plan-
Last, I must emphasize a very special etary Observers, and the British Astronomical
piece of equipment that literally opened Association.

Advertisement

2
3
5

1 4

Top: The author works most often from a roof-window

observatory (center) on the top floor of his apartment
building in Oeiras, Portugal. Note the fireplace chimneys
surrounding the open window. Above, right: His main instrument is a 10-inch LX200 telescope
mounted permanently on a concrete pier and operated remotely from a computer room
15 meters (50 feet) away. The doors between the scope and computers are always closed to
minimize thermal currents and mosquitoes. Above, left: The authors imaging setup includes an
ST-5C camera (1) with an internal RGB filter wheel; an SBIG CFW-6A motorized filter wheel (2)
with ultraviolet, broadband infrared, and methane-band filters; an AO-2 device (3) fitted with a
small Barlow relay lens; a conventional miniature video camera (4) for remotely centering the
planets image on the AO-2; and a Meade f/3.3 or f/6.3 focal reducer/field flattener.

134 December 2001 Sky & Telescope 2001 Sky Publishing Corp. All rights reserved.