I collected several more examples of the Cygnus Wall when I found this extremely clear example, this time drawn back a bit so you could see more of North America. The astrophotographer called it the Cygnus Wall in Diamond Dust. In responding to a question about why image processors called the initial shots data, he gave this incredible talk on what they do to produce a new picture.

I'll try to go over the process of astrophotography briefly, though it is actually quite a complex process... I'll speak in general terms about how my collaborator and I do it for clarity; this is not how everyone does it but it serves as a decent overview.

Astroimaging telescope equipment is set up to track the stars, meaning it has to account for (and cancel out) the rotation of the Earth. That involves a motorized mount, usually very precisely aligned with the Earth's rotational axis (this is known as an "equatorial mount").

Telescopes usually have very precise mounts, mechanically, but even so the mount may not be perfect, and it may drift if the mount's rotational axis is not perfectly aligned with the Earth. This leads us to "guiding", in which a separate telescope and camera, mounted to the primary imaging system, are set up to image a particular star (a "guide star") over and over, and make minute adjustments when its image moves relative to its initial position, effectively fine tuning the pointing of the telescope. Not everyone does this, but it is used to optimize the imaging system to point at exactly the same point in the sky throughout each exposure. This will yield a sharp image.

On the telescope is mounted a purpose-built astro camera, highly optimized for light sensitivity in the spectrum(s) interesting to an astrophotographer, and for precise remote control by computer. The telescope may also have an electric focuser, also capable of being controlled remotely.

Keep in mind that we're shooting photos through the Earth's atmosphere, and the amount of light coming from the atmosphere is significant in virtually every location (ground lighting may be reflected back, the atmosphere diffuses starlight or moonlight, and it even glows a certain amount with its own light). Keeping this in mind, the amount of time for any one photographic exposure must thus be limited, or it will yield simply a whited-out image. Depending on sky conditions, optics, and imaging equipment in practical terms the time can be from seconds to a few minutes to a matter of hours. If the time is set so that the sky glow is not allowed to overwhelm the image, it can be subtracted digitally. But I'm getting ahead of myself; more on that in a minute.

Since we're struggling to capture very, very dim things in the night sky, literally pushing the limits of the photographic equipment, modern astronomers usually take many, many exposures (called "subexposures" or "subs") and combine them digitally (a process known as "stacking"). This tends to average out individual errors (noise) and yield cleaner imagery (better signal). Since things in the heavens don't generally move much, repeated subexposures of the same objects again and again can be combined in this way. One can often even combine subexposures taken YEARS apart.

Many amateur astronomers set up their imaging equipment so that a computer automates the process of capturing subexposure after subexposure. There is specialized software written to orchestrate this process, often integrating the guiding with operating the camera.

This takes us to the data... Each subexposure, when downloaded from the camera into the computer, is written to a file. These subexposures consist of the raw linear readings of light levels from the photosites of the camera's imager - a bit like a digital camera raw file, but stored in an science-oriented file format called FITS (which stands for "Flexible Image Transport System"). This data is usually saved at a high bit depth so as to maintain the highest possible accuracy. It's common for a single subexposure to take upwards of 100 megabytes on disk.

Once a number of subexposures are taken - say several hours worth of 2 minute subs (e.g., 60 subexposures of 2 minutes each) - and the session is complete, specialized software is again brought to bear to interpolate color from (de-mosaic) each of the subexposures and precisely align and stack them into yet another high bit depth FITS file. It's still not an image yet, even though the pixel values have been averaged to minimize noise and maximize signal - the readings are still linear, and the data needs to be "developed" or "stretched" into what we consider an image.

Here's where even more specialized software comes in... The stacked FITS file can be opened into an image editor - I use Photoshop - via a specialized plugin called FITS Liberator that understands the FITS format, and provides sophisticated developing functions for "stretching the linear levels" using tone curves and creating an image we can see in the typical gamma-corrected space of our computers. This must be done one color channel at a time and through careful choice of math functions and parameters, including setting black and white levels, scaling, and choice of formats. Here again I choose a high bit depth format to preserve the highest possible accuracy. You might think this is a bit like developing a raw digital camera image, and there's a similarity, but imagine your raw converter made an order of magnitude more "geeky" and you'll get an idea of what's involved. I have written my own software to process the color channel data delivered from FITS Liberator into a full color RGB image in Photoshop. FINALLY we have something one can actually look at as an image. But we're not NEARLY DONE yet!!!

At this point, we have an image in Photoshop that represents the stacked raw data from the imager. But it's got some problems we need to fix before it's ready for public consumption. It has, for example, uneven lighting because no optical system delivers a perfectly flat luminance field (there's usually vignetting because the aperture of the telescope isn't infinite). Also, there is sky light (from the atmosphere) to be subtracted, and there are hot pixels, noise, and optical inaccuracies.

Again, using specialized software, some of which I've written myself, I flatten the luminance so that there is an even level of lighting all across the image, corner to corner. I then subtract the sky light level so that the background color - deep space - looks nearly black. It's at this stage that the true colors of the image finally start to come out. I hand retouch hot pixels - basically places where the camera's imager has delivered a too-high reading for light level, and which show up as colored bright dots. Since the telescope does not generally track absolutely perfectly throughout an imaging session, hot pixels tend to look like little blobs or streaks that can quite easily be differentiated from stars.

At this point quite often the stars are very bright and nebulae or galaxies are very dim by comparison. This is because stars ARE very bright by comparison to these deep sky objects. Depending on what is being highlighted in an image, I'll adjust curves and even mute the star brightnesses in relation to the background in order to bring out very faint objects.

Final color balancing, image sizing, cropping stacking remnants off the edges, and preparing an image for display and voila, what you see above.

It is truly a labor of love, and done carefully it's truly amazing that so many of the wonders of the universe can be seen so well with amateur equipment and a bit of time.