Está en la página 1de 4

astro imaging

Eclipse Photography
in the Digital Age
The potential of eclipse photography from the deck of a rolling ship is all
the more exciting when we consider the digital-processing techniques possible with todays home computers.

By Gerald L. Pellett
computers and digital processing is used
to extract detailed coronal structure from
these brief exposures.
A goal of my low-budget eclipse trip
to the Bolivian Andes in 1994 was to take
a range of exposures on 35-millimeter
slide film. I planned to use then-emerging image-processing techniques to
squeeze available detail from each slide
and create a composite view in which the
original range of exposures was compressed into a single image. Because I
planned to carry only a lightweight tripod with no motor drive, my exposures

Thin cirrus clouds interfered slightly with the


authors eclipse photography from this 12,500foot site some 40-kilometers south of Potos,
Bolivia. Nevertheless, the cloudss effects were
rendered virtually invisible as a byproduct of
his image-enhancement techniques.

were limited to 18-second or shorter to


avoid image blur. In order to capture the
Suns faint outer corona I selected
Kodaks fast Ektachrome 1600 film. Unfortunately, this emulsion is grainy. It
also suffers from low dynamic range,
which meant that only a limited span of
coronal brightness was adequately captured at each exposure duration. I now
believe that using a fast negative film
would have been a much better choice
on both counts.
After my trip, I was pleasantly surprised to find Eclipse Photography Goes
Digital in this magazines November
1994 issue. Computer expert Steven Albers created a remarkable five-exposure
composite of the July 11, 1991, total
eclipse using photographs made by Dennis di Cicco and Gary Emerson. Albers
wrote about 5,000 lines of Fortran code
while developing his routines for isolating optimally-exposed coronal regions in
each image and creating his composite
Considering the marginal quality of the
original photographs, writes the author, a
surprising amount of detail was extracted
using digital techniques. His image processing was done with a home computer and commercial software. The original images were
obtained during the November 1994 eclipse
with a 600-mm f/8 mirror lens and Kodak Ektachrome 1600 film.

ALL IMAGES BY GERALD L. PELLETT

housands of tourists are


heading to the Caribbean to
witness next months total solar
eclipse. Many will be photographing the event from the decks of
cruise ships. While conventional wisdom
from the past tells us that serious photography of the Suns corona warrants a stable platform, times have changed. Its now
possible to obtain excellent results from a
rolling ship thanks to high-speed film and
fast telephoto lenses. The potential of
shipboard eclipse photography is even
more exciting when the power of home

1997 Sky Publishing Corp. All rights reserved.

Sky & Telescope January 1998

117

astro imaging

An original scan (upper left) of a 160 -second exposure is seen after processing with a 10 radial-blur filter (upper right). The blurred image
was used to process its corresponding original
image. The seven-image composite (left) melds
information captured in exposures of differing duration. It gives a visually realistic view,
while sacrificing detail shown in the other
frames with this article, which were combined by a separate method. The detailed operations of both techniques are explained in
the box below.

view. The pioneering project required


about 500 hours of effort spread over a
year and a half.
In the same article, astrophotographer Kazuo Shiota showed a similar image
constructed from six exposures obtained by a team from the University of
Tokyo. Shiota, however, prepared his
composite with Adobe Photoshop, a popular graphic-arts program available for
Macintosh and Windows-based desktop
computers. Each image was processed
with an unsharp mask made using the
spin mode of Photoshops radial-blur filter. The article indicated that this bit of
technobabble is for those familiar with
Photoshop who might wish to experiment. Shiota estimated that future projects would require about a month of
processing for each photograph used in
the final composite.
Armed with this bit of information
and 15 reasonably good 35-mm slides
with exposures between 18 and 1500 second, I started on a similar path. I soon

Notes for Photoshop Users

hile the actual steps I used to create my images


are straightforward they are most easily understood by people familiar with Photoshop. Each
image was centered, sized, and displayed on an identical
Photoshop canvas before I began detailed processing. I did
this by first recording the four vertical and horizontal tangent
points on the Moons limb and translating their averaged values left-right and up-down to common, centered-pixel locations. Each Moon image was then scaled to a common size
and the canvas size cropped to a common dimension. I used
the programs despeckle filter to soften the apparent graininess. Copies of each image were blurred with the radial-blur
filter using a 10 spin. This operation replaced the brightness
value of every pixel at a given radius from the image center

118

January 1998 Sky & Telescope

with the average of all pixel values at the same radius in a 10


arc. It took one to two hours to process each image, but
todays computers can accomplish the task in about one quarter the time.
The pixel brightness values of each blurred image were subtracted from the corresponding pixel values in the original.
This created a view with a small overall range of pixel values.
These difference images were scaled by 1 (a required Photoshop operation) and offset by adding 128 to all pixel values to
place the values near a midrange brightness.
To create a detailed composite, the low-contrast difference
images were added sequentially. As each image was added,
the result was scaled by 1 and offset by subtracting 128 from
all pixel values. A scale divisor larger than 1 causes progressive loss of detail because the effective range of useful pixel
values is diminished with each addition-division step. As as
final operation, the result obtained by combining 7, 10, or 15
images was multiplied by the pixel values in the 18 -second
exposure.
To create a more visually realistic but less-detailed composite, each low-contrast difference image was multiplied by the
pixel values in its original image. Doing this resulted in images
with a much wider range of pixel values. Subsequent additions
of these images (followed by negative offsets) required a scale
divisor greater than 1 to avoid overflowing the available 0-to255 range of pixel values. The smallest usable divisor depended on the contrast of the difference images and the number of
images in the final composite. For example, I used a scale divisor of 1.4 for my seven-image composite. Because each step
reduces previously recorded detail, the order of image addition also affects the result.

1997 Sky Publishing Corp. All rights reserved.

Top row: Processing the visually realistic


seven-image composite with radial-spin filters having (left to right) 10, 15, and 20 spin
angles offers varying levels of detail enhancement. The author found angles of less
than 10 unsatisfactory. Bottom row: These
10-image composites were processed with
radial-spin filters having 10 (left) and 20
spin angles.

discovered, however, that several conceptual and practical details are important
for successfully extracting coronal structure from the originals and retaining it
in the final composite. Furthermore, two
significantly different techniques were
discovered that produced notably different results. One maintained maximum
detail, while the other produced a visually-realistic image. With time and patience, anyone can perform either process with Photoshop and a desktop
computer.
In the midst of my project the March
1995 issue of Sky & Telescope arrived
containing Roger Ressmeyers computerprocessed image of the 1994 eclipse. He
also used Photoshop to combine six exposures ranging from 1 to 1500 second

that were obtained with a 600-mm telephoto lens and Fujichrome Provia 100
film. His final image file was 576
megabytes in size! He stated, At the beginning, I spent two hours with 1-megabyte images working out my approach,
then 18 hours compositing the final version. Although his method was not discussed, his composite appears similar to
my visually realistic example, but it also
displays some symmetric artifacts, which
I believe result from contrast and brightness enhancements of the composite
image.
Conceptually and operationally, my
processing techniques are quite simple.
Each of my slides was digitized with a
Leafscan 45 system attached to a 1993vintage Power Macintosh 8100. The ef-

fective resolution, after optical magnification, was about 2,200 dots per inch on
the original, which was more than adequate for extracting all the available information from my grainy slides. Each
slide was thus transformed into a 12megabyte TIFF file, which contained 8
bits (256 levels) of brightness information in red, green, and blue color channels. My processing was done with the
Macintosh and Photoshop 2.5 (later upgraded to version 3.0). Details of the
methods I used to manipulate and combine the individual exposures are given
in the box on page 118.
To produce a final image, the Moons
disk, which had become slightly brightened during processing, was handpainted using standard Photoshop tools.

The effects of compositing additional images are apparent in this sequence made with (left to right) 7, 10, and 15 originals. All were processed
with 10 radial-blur filters to increase contrast and visible detail.

1997 Sky Publishing Corp. All rights reserved.

Sky & Telescope January 1998

119

Advertisement

Finally, I made minor overall adjustments to brightness, contrast, and hue.


Except for retouching a few dust specks,
coronal details were left exactly as obtained. In fact, a second seven-image
composite was generated from another
set of image files made with the slides
turned 90 in the scanner, and the result
was identical!
My image files were sent to a highend output device that can produce
prints, negatives, or positive transparencies. It is also possible to photograph
the computers monitor directly or have
prints, slides, or negatives made from
the digital files at a modern photolab
equipped to handle such files. With digital imaging becoming increasingly important in the world of conventional
photography, even small photolabs are
beginning to offer electronic services.
The number of images combined and
the degree of radial spin used for the
blur filter both affect contrast of the
composite image. This was particularly
true near the edges of large equatorial
structures recorded in my original photographs. With a 20 spin, the contrast of
major features appeared too exaggerated,
and details near the coronal brushes were
overwhelmed. A 15 spin was better, but
10 appeared near optimum for composites of seven to 15 images. Spin angles
less than 10 degrees caused progressive
loss of coronal detail.
It would have been nice to include
original exposures of 11000 and 12000 second,
which should have prevented the obvious washout of bright detail, including
prominences, near the Suns limb.
The edge-detection filter I developed
with this method works best with objects
having radial structure and a wide range
of brightness. Zoom blurs and simple
unsharp masking were much less satisfactory. By using fast negative color film,
it should be possible to do good science
from the deck of a rocking ship! One of
the new optically stabilized telephoto lenses might offer some additional advantage. A positive benefit of the compositing technique is that it helps remove
cloud structure, which, unfortunately, was
present in my original slides.
A senior research scientist at NASAs Langley
Research Center in Hampton, Virginia, Gerald Pellett pursues amateur astronomy as
a hobby. He wishes to acknowledge Susan
Bowen, who coached his image-processing
strategy.

120

January 1998 Sky & Telescope

1997 Sky Publishing Corp. All rights reserved.

También podría gustarte