Two Years In - What I've Learned

Dec 2010 - The Greatest Astrophoto Ever Taken

I'm coming up on 2 years of deep sky astrophotography now, and as is the way of my people, it brings about some reflection on where I've come from and where I'm going with this whole thing.  You may remember that I did this a year ago on these very pages when I debuted the undisputed Greatest Astrophoto Ever Taken.  It's visible at left in case you've forgotten the awesome stupendousness of that image.  As technically perfect as that image is, it's even more interesting to me that it was the first image I'd ever taken.  Seeing the image come together as I stacked up the frames and attempted to do some basic levels adjustments lit a new passion for me that continues unabated to this day (which is a Wednesday).

Dec 2010 - My first "true" deep sky image.

Orion has been a common theme in my first couple of years as a deep sky imager.  I thought I'd use this entry to take a look at those images chronologically.  I'll document as well the processing steps that I took on my final M42 image.  To the right is my first attempt at a deep sky object.  I shot the M42 region with a cheap telephoto lens piggybacked on top of my telescope from my Goodyear, Arizona backyard.  

Time to Get Serious

Jan 2011 - This is what happens when you allow
someone to PhotoShop unsupervised.

 My piggyback astrophotography phase lasted about two weeks before I took the plunge and added a Hyperstar to the imaging rig.  Seriously, I don't think it lasted that long.  I captured exactly four images as a piggyback astrophotographer before this obsession drove me to longer focal lengths.  My first night out I shot (naturally) M42.  This time it was with the DSLR strapped to the front of the CGEM 1100HD and Hyperstar.  By this time I had invested in PhotoShop CS5 and had learned just enough about image processing to mangle even the most pristine data.  As you can see, I knew how to stretch the data and even a little about how to layer mask poorly.  You can also deduce that I either knew nothing yet about color balance or that I am color blind.  Thankfully, it was a lack of understanding on how to balance color. 

Apr 2011 - The same data after some learning.

For the technically interested, the image consisted of 32 subexposures of 120 seconds each and 32 subexposures of 10 seconds each.  MaximDL was used for capture, calibration and stacking and all subsequent post-processing was done in PhotoShop CS5 using my extremely limited image processing talents.  For the artistically interested and all those praying for a happier ending to that photo, I took some classes.  I read through PhotoShop Astronomy which was somewhat helpful.  I picked up Ron Wodaski's Astro Zone system which was more helpful.  Incidentally, I think OPT is the only place that you can get it anymore.  What I found most helpful though was signing up and going through Warren Keller's IP4AP tutorials.  It was his tutorials that helped me turn the same data into something more realistic.  This image is the same data, just processed much better. 



Nov 2011 - Cooled CCD and a year of experience
make for a much better image.

You can see though that the color is still lacking.  This was being caused by my unmodified DSLR and I soon took the plunge into the realm of the CCD camera and bought a QHY8PRO from Astrofactors.  After breaking it in on a bunch of other images, I shot M42 once more with the Hyperstar and CCD in November of 2011. Then, as now, I wanted to compare what I could do with my earlier attempts.  I think this was a vast improvement.  I had learned much more about layer masking, color balance and getting more out of the data.  Let me first say that I don't think there's a better way to get into deep sky imaging than Hyperstar.  It makes it so much easier to get a lot of good data quickly.  It's so fast that it allows you to add the other pieces of the puzzle like autoguiding and drift alignment one at a time.  Having said that, it was at this point that I began to feel limited by the Hyperstar.  A month or so later, I began shopping for a refractor and longer focal lengths.

2012 - It Ain't Broke; Fix it!

It really wasn't broke, but my long term goal had always been to get to longer focal lengths and high resolution deep sky imaging.  Over the course of 2012 I have managed to overhaul the entire imaging rig.  I changed the mount to an Astro-Physics AP900.  I changed the imaging telescope to an Explore Scientific ED127 Triplet.  I sold the one shot color CCD in favor of mono with an SBIG ST-8300m.  I added an off axis guider and 8-position filter wheel.  Into the filter wheel I stuffed a full complement of Astrodon broadband filters and Astrodon 3nm narrowband filters.  Early in the year I tried out the PixInsight platform and fell in love with it for image processing.  I have now reached a point where all of my post-processing is done in PixInsight.  I haven't used PhotoShop CS5 for any part of the processing in probably six months.  Of course, I can't leave well enough alone so I added CCDAutoPilot to the mix in an effort to completely automate the imaging rig which has been a great success.

My middle daughter had surgery to remove her tonsils and adenoids at the end of December 2012 and I took some time off work to stay home and be some moral support for her and to help her recuperate.  At 16 years old it's not as simple of a procedure as it is when you're much younger.  She did great though and is back to her 100% self eating hamburgers and all kinds of other awesome food with me.  She had mentioned that the Orion Nebula was her favorite deep sky object so I thought it would be appropriate to capture it while she recuperated.  I wanted to do it at least a little bit different and I was shooting from my light polluted backyard so I went with a narrowband image and processed it in the Hubble palette.  I shot 4.5 hours worth of data and put the image together.  I realized then that I had a really nice image going so I extended the project to collect significantly more data.  The final result is shown below.

M42 - 19 hours in the Hubble Palette

I collected 19 hours of narrowband data between December 17th and January 4th and then put it all together into the image shown above.  What really impressed me with this image was the amount of detail that I was able to record using just a 127mm refractor.  The detail in M43 and the dust lanes that separate the two Messier objects draws my eye each time I look at it.  My daughter calls it my best work yet.  She immediately found the "soap bubble" in the right center of the image - something I hadn't noted previously.

PixInsight Processing Steps

1. Data Calibration - I use the method described in Vicent Peris' tutorial to create master calibration frames and calibrate the data.  I calibrated 106 subexposures in total with subexposure times of 30 seconds, 3 minutes and 30 minutes.
2. Image Integration - I stacked each filter/exposure separately and saved the files in a FITS format.  I had 9 files total - 3 subexposure lengths each of HAlpha, OIII and SII data.  As typical in the Hubble palette, I mapped SII data to the red channel, HAlpha data to the green channel and OIII data to the blue channel.
3. Channel Combination - I created an RGB Master of each subexposure length.  So I ended up with a 30 second RGC, a 3 minute RGB and a 30 minute RGB master.
4. Dynamic Crop - I cropped the image to remove the ragged edges from stacking a set of dithered frames.
5. HDR Composition - This is just an awesome tool built into PixInsight.  It does a linear fit of the three images and then combines them in much the same way that you'd do it manually with a layer mask in PhotoShop.  Truly awesome stuff here.
6. Create a Luminance Master - Extracting the luminance and saving it as a separate master, I proceeded to process the image as a normal LRGB.
7. Multiscale Median Transform - I created an inverse luminance mask and then did MMT solely to do noise reduction while the data is still in a linear, unstretched format.  This noise reduction allows me to push the image more when I stretch the data.
8. Histogram Transformation - PixInsight's fancy word for stretching the data.  It's crazy simple to use and ridiculously effective. 
9. HDR Multiscale Transform - Compressing the data in specific wavelet layers does a great job of pulling out the data that's being overwhelmed in the Trapezium.
10. Local Histogram Equalization - Using a luminance mask and running this process restores some contrast that is lost in the wavelet processing above. 
11. Multiscale Median Transform - Again, using a luminance mask designed only to allow the process to work on the highest signal areas, I used MMT this time to do some sharpening on the luminance frame.
12. LRGB Combination - Finally, I combined the luminance with the RGB master and did some final tweaks to the histogram, color balance and saturation.

Synthetic RGB

After I'd completed the image, I did a final step with the Pixel Math tool in PixInsight and recombined the channels in an effort to create a realistically colored image using only the narrowband data.  I could probably play around a bunch more and get it really close.  This will have to suffice though.  I'm really happy with it. 

Red Channel = 70% HAlpha + 30% SII
Green Channel = 25% SII + 75% OIII
Blue Channel = 70% OIII + 30% HAlpha

If you've stayed with me this long, thanks!!  After this one, I think I'm done shooting M42 for awhile.  There's a lot of other objects out there that I want to shoot.  I'm really happy with my progress in the imaging arena and I feel like I'm starting to turn out some really good astrophotos.  Next time, I'll offer advice I wish I could go back and offer to myself.  I hope you'll stick around for that and that you'll find it helpful as well.