TRI-COLOR with CCD IMAGERS
The growing popularity of CCD imagers has many interested in the task of making color images with these devices. In some places these are touted as the road to getting 'true color' images of astronomical objects. While the use of these imagers does give the user far more control over the final colors obtained, the task of getting 'true colors' is still a difficult one and beyond the effort that most will want to invest in the problem.
Whether one wants 'true color' or just truer colors, the road
to
the final result lies with the selection of color filters that
are well suited to the task and with the process of calibrating
the exposures to the filter set that has been selected.
The matter of carefully selecting a set of filters is far
beyond
what can be presented here in a web site dissertation. The
matter of filter selection for tri-color with film ( which is
VERY similar ) is discussed in depth in "A Manual of
Advanced
Celestial Photography" and the complete discussion of the
problem for CCD imagers will be presented in-depth in our next
work "Astro-Imaging: From Silver to Silicon". For the
purpose
of this discussion, one could either select a filter set such as
the KODAK Wratten set ( R=25, B=47B, G=58 ) or with a commercial
set such as those marketted by SBIG. In both instances, an IR
rejection filter is needed for the tri-color work due to the
fact that these filters are nearly completely transparent in the
IR as well as the fact that CCDs have a very high sensitivity to
light beyond 700nm where they eye and film are insensitive. A
tri-color shot of a color test chart taken with the KODAK filter
set and no minus-IR filter will show the chart as just an array
of grey swatches but redoing the same shot with the minus-IR
filter will show the chart as the eye would see it if the
exposures were properly calibrated.
Note: The B,V,R photometric filters also yields a fairly nice
filter set for tri-color work.
Once a filter set has been adopted, one needs to know what the proper exposure ratio is for each of the three filters in order to be able to reconstruct a color image which has a semi-reasonable color balance and one where the colors presented in the final image are semi-accurate.
For smaller scopes it is a fairly simple matter to perform such an exposure calibration. All that is needed is a neutral grey test target which can be purchased in many well stocked photo supply stores. The ultimate test target is the MacBeth ColorChecker but this is expensive and is far more than what is required. With such a target in hand, one simply takes images through the three tri-color filters and determines the ratio of exposures for each of the filters after illuminating the test target with a light source that is sun-like and has a color temperature of 5500 Kelvin and which is faint enough to allow exposures that are 10 to 100+ seconds in duration.
One can pick a super lamp with such qualities and then dim the intensity by using aperture stops or *neutral* density filters ( KODAK 'neutral' density wratten filters are NOT neutral ! ) but this is tricky to do. The same comment applies to trying to illuminate the test target with the light of the sun. ( One trick that has been used to solve the intensity problem with sunlight and/or lamps is to use a small bundle of fiber optic cable to bring a tiny amount of sunlight into a dark room where the target is then illuminated and imaged ).
However, the simplest solution for getting a nice 5500K light
source is to do your imaging calibration exposures by moonlight
well after sunset. With the light of the full moon (well above
the horizon) one can readily use exposures of 100+ seconds for
the exposure calibration. SO, after taking a 100 second red
exposure one measures the DN level of the grey target and then
take a 100 second exposure in the green filter. The ratio of the
red DN level to the green DN level will yield the exposure
factor for the green filter. ( Well matched exposures will
yield EQUAL DN LEVELS for each of the three TC filters when
imaging a neutral test target ).
RED DN LEVEL --> 600 DN ..... GREEN DN LEVEL --> 525
DN
GREEN EXPOSURE FACTOR : 1.14
Repeating the same set of exposures and measurements for the
BLUE filter will then yield the exposure factor for the blue
filter.
A calibration of an ST-7 using the SBIG filter set (ca
1996)
yields the following set of filter factors:
| FILTER | FILTER FACTOR |
|---|---|
| RED | |
| GREEN | |
| BLUE |
With such a filter set, a 300 sec RED exposure would require
a
GREEN exposure of 330 sec and a BLUE exposure of 870 sec in
order to obtain a well balanced set of exposures for a
reasonably accurate color rendition.
MacBeth Color Chart
For larger scopes, imaging a test target is quite
unpractical.
For larger scopes an acceptable means of calibrating a filter
set is to image a star that is very close to our sun in its
light emitting characteristics. Such 'SOLAR ANALOGUE' stars are
scattered about the sky, but are not as common as one might
think. One of the best solar analogue stars is well placed for
Northern summer imaging, it is 16 Cygni. This star is a double,
but both of its components are very nearly solar in their
characteristics. The process of calibrating using a stellar
target is identical to that of using a test target (although
one will probably need to use a longer FL and defocus the
stellar image in order to get lengthy exposures).
When doing tri-color imaging, one needs to keep in mind that *each* of the three images needs to have a very good Signal-to-Noise Ratio ( SNR ). If one image has a low SNR, then color 'speckling' will be seen in the final image. In the raw image, a low SNR is obvious when the image looks like it is very grainy. When such an image is pulled into a final tri-color image, this 'grain' will appear as a speckly pattern of the color corresponding to that particular image and this color speckle will be quite objectionable. One can either lengthen the exposures of the image set to be sure that all of the images have higher signal levels or one can co-add several image sets of the same exposure durations in order to decrease the noise level.
Another step that can help to reduce the noise level of the
final image is to use a dark frame that is a composite or
average of several dark frames (dark noise is a statistical
process) and to use flat field frames that are composites of
several flat field frames and which are of a good SNR level of
at least several thousand DN in each flat field image ( again, a
flat field image (all images) are a statistical process and
higher DN levels mean lower noise levels ) or about 1/4 to 1/2
full well capacity.
M42
NGC 1975
AE Aurigae
After you have a well calibrated set of images of a favorite deep-sky object, how do you put them back together to make a color image ?? There are several options available for this process and several pieces of commercial software that will take on this task and make it relatively simple to execute this operation. PhotoShop, MIRA, MIPS, IDL and SBIG's own CCDCOLOR software packages will all make this task into a fairly simple process for obtaining images with varying degrees of color accuracy. IDL and MIRA allow the user the greatest degrees of control for image manipulation and color control and thus these two packages have the greatest potential when one is seeking to reconstruct colors at the highest accuracy.
PhotoShop allows the user to manipulate image positions and colors to a great degree, but the user can not keep the image in its original format and the user does not have exact control over the image when it comes to color and contrast manipulations. PhotoShop can not import an image in the 'native' format of the imager and it can not read files in FITS format either. In order to use PhotoShop one needs to save the raw images in a format that PhotoShop can read such as 16 or 8 bit TIFF with the 16 bit TIFF being the preferred format as no data is compromised in this format.
With PhotoShop, the three images are opened up and adjusted for best overall viewing. This is done using the ADJUST LEVELS menu and the three images are all brought in using *identical* settings for offset, gamma, and brightness range so as to not badly distort the final color balance. Once this is done, convert each of the images into an 8-bit GREYSCALE by using the options in the IMAGE MODE Menu. It is important that all 3 images be given the exact same stretch and that an identical overall dynamic range is adopted for all three images. If this is not done then the image color balance will not be preserved.
After the three raw images are up, a new image that is of RGB format and of the same dimensions is then opened up and the DISPLAY CHANNELS menu is then opened up. Now, make the RGB image active and click on the RED CHANNEL, then go to the RED raw image and cut and paste that image into the RGB RED layer. Repeat this cut and paste for each of the three raw R,G,B images. Once the final RGB image is pasted together, you can zoom in and see if the images are all well registered. If not, you can activate each layer individually and use the DRAG function and the arrow keys to register the final RGB image. Once the registration has been completed you can begin to make small changes to the final color balance, intensity and saturation of the merged RGB image.
You can register images to sub-pixel levels by doubling or
tripling the image size when you get close to the final
registration shifts. Once the images are well registered you can
then shrink the image back to its orignal size and preserve the
high quality image registration.
In the next month or so ( when the clouds clear )
we will begin using a specially selected set of interference
filters that were designed for optimal color
rendition over the entire visible spectrum and which were
designed to allow correct color renditions of monochromatic
light sources such as those found in nebulae. Will these
special filters produce renditions that are markedly different
than what we are used to seeing in color photographs and the
calibrated renditions shown here using the SBIG filter set ?
The ultimate ability to make accurate color renditions of
objects such as these lies in the selection of the filters used
for the tri-color exposures. The differences (if any) that
this special filter set produces should be interesting so check
back in to see what this experiment shows.