Ad Astra Observatory

The pretty pictures that show up on this website take a little work to produce.

Just getting the camera to take a photo is quite a process, with balance and alignment and focus and guiding (and now platesolving 🙂 ) and &c. But that’s another workflow for another time.

Right now, I’d like to talk about what happens when you’ve got a disk full of images, and you want to turn them into one (hopefully good) image. This, it turns out, is an art unto itself.

I use a DSLR to capture the photons, and (at the moment) I have the camera capture in RAW mode (in Canon’s .CR2 format), because it will produce better-quality data to manipulate.

The RAW data has lots of information in it; some of it is the object being captured, but then there’s light pollution, camera noise, optical defects, and other nasty stuff, all hanging around making one’s life miserable. It’s the job of the image processing workflow to reduce the “bad” data as much as possible, so that only the actual ancient photons remain.

So, for each individual frame, you can:
– subtract a “dark” frame (an exposure the same length as the sky photo, but with the lens cap on), to remove any problems that the camera itself adds (hot pixels and other heat-related issues)
– divide by a “flat” frame (a short exposure of as blank a background as possible), to remove any problems caused by the optics (dust bunnies and vignetting)
– … there are also “bias” frames, which I have never fully understood, and then, if necessary, you might have to subtract a different set of “darks” from your “flats”, as the level of dark noise will be different because of the shorter exposure length… you can really wrap yourself around the axle with this stuff

Ok.

So, if you’re doing it right, you have several hours’ worth of actual exposures (call these “lights”), then several hours’ worth of darks (same exposure time as the lights, remember?), then several minutes’ worth of flats (have to shoot these during the day, unless you have a way of producing an evenly-illuminated white surface on command, search this blog for “flat box” for more on that), and perhaps a handful of other types of images, if you’re into that sort of thing. For each of these types, you’ll want to have something like 10-20 frames, so that you can stack for better signal::noise.

I used to do all my image processing on Windows, and there’s a wealth of different software packages (including at least one that’s free) that can do the reducing and stacking of all these calibration frames, and then apply them to the lights before stacking all those, too.

I have switched to Mac for image processing, and the options are quite a bit more limited. My workflow until now had been:

– Open all the RAW files in “Camera Raw” in Adobe Bridge (Camera Raw is only available if you also own Photoshop, keep reading), use this to do an initial pass at normalizing the image histogram (because of the astro-modification done to my DSLR, the white balance is way off, so the camera has huge differences in sensitivity between R, G, and B), then save each image as a .TIFF

– Stack the .TIFF images in Nebulosity 4 (about the only Image Stacker under OSX that I could find) — it knows how to handle TIFF and FITS (an astro-format), but not RAW, hence the Bridge thing. Save the now-aligned stack as a single TIFF (TIFF is a lossless format, so it’s better than PNG or JPG at this point)

– open the single TIFF in Adobe Photoshop, realign the histogram, levels and curves to boost the contrast, then crop, add a text label, resize to 1500×1000, and save as a PNG. This is the file that gets uploaded to the blog.

This method is a little clunky; I particularly hate that I have to manipulate the images before bringing them into the stacking software (the reason why is soon to be revealed), and you’ll notice that I made no mention of darks, flats, flat darks, biases, or any other calibration frames.

Why not, I hear you ask? Well, several reasons.

First, I don’t get a lot of dark noise in my images; the DSLR stays pretty cool for one reason or another (also, I’m usually shooting in sub-freezing temperatures, given the way the clear nights work around here), so dark frames have never really been a thing for me. I tried shooting some darks, and didn’t really notice much improvement in my images.

Second, in the absence of dust bunnies, I don’t much care about vignetting. If you look at the various images I’ve taken and then posted to the site, you’ll see that, of all the artifacts and problems that I have with my images, a slight dimming at the corners is the least of my worries.

Finally, it’s a real pain in the neck to go through shooting all those extra frames, when I can barely get light frames done. Not to mention that I have sometimes found that over-calibrating my images leads to worse results than just living with dust bunnies and hot pixels.

So, I often skip the whole calibration frame business, and just post stacked and contrast-stretched light frames, and call it a day.

Then I took the camera off the telescope, and forgot to put it in its hermetically sealed Ziploc bag for storage, and suddenly I have a bad dust bunny problem. Cleaning optics has its own level of complication, so I decided to try my hand at flat frames, to see if I could remove them “in software”.

I lucked out and got a nice, cloudy but dry, day after several days of clear nights (and a nice backlog of images to process). I pointed the scope at the most evenly-clouded section of sky I could find, and rattled off 3 sets of flat frames (one ended up being in FITS format, because I couldn’t figure out how to make the software shoot them RAW, and then one set of RAW @ISO800 and one @ISO100, because I was having exposure length problems, don’t get me started).

So, workflow.

– Bridge, lights, RAW->TIFF
– Nebulosity, flats, FIT->stacked into master flat, save as TIFF also
– Nebulosity, process lights with master flat…

…and here’s where I started running into problems. Nebulosity threw an error, something about “color types do not match”. OK, so try a variety of things, then:

– Bridge, flats, RAW->TIFF
– Nebulosity, flats, TIFF->stacked in to master flat, save as TIFF
– Nebulosity, process lights with master flat, worked this time, woot
– Nebulosity, stack pre-processed lights, save as TIFF

At this point, I took the file into Photoshop, and I noticed that the dust bunnies were, in fact, a bit less prominent. However, I also noticed that the stacks were a lot more sensitive to individual frame quality; I had to restack several of the photos to remove iffy individual frames,… I spent a lot of time trying to get the shots looking right.

I also noticed that I had a sort of blown-out red channel in several shots, which made it difficult to get the contrast working. I couldn’t figure out what was going on, and had to walk away from the problem for awhile to think.

Then it came to me.

That initial step that I did? Bridge->tweak the histogram? Yeah, you need to apply the calibration frames to the *RAW* image, not a processed one. Also, the calibration frames need to be RAW, as well.

So, I will have to go back through and see if I can re-process all of these with uncooked lights and flats… perhaps that will yield better results.

I hate calibration frames.
I hate the OSX workflow.

…

but I hate Windows even more.

Back to the drawing board.

Don’t cook your RAWs.