August 19
The router I talked about in my last post worked well, and I should be able to monitor my imaging remotely from now on. Other things happened this night that made it less than successful.
Imaging started with NINA doing a slew, center and rotate, or trying to. I use NINA's manual rotator, which has always worked well for me in the past. This night it did not. My first rotation resulted in NINA requesting an even greater rotation in the same direction, and so did the next. I didn't want to sort it out, the image wasn't going to amount to much given the omnipresent overhead smoke, so I redid the sequencer script to just slew and center and pause so I could do a rotation by eyeball. Not at all optimal but it started the collection of light frames.
After a while I noticed that the images were showing fuzzy haloes around stars. It wasn't a focus issue; the stars were sharp. It was a very dewy night so that was my first guess. I put everything on pause and looked at the objective. It was dry. I could have examined all the other optical surfaces but instead I looked at the camera settings. The camera was at the correct temperature (-5 C) but the dew heater on the sensor window was off. I set it back on and decided to wait and see if it would clear; I could monitor this on my tablet indoors. The images slowly improved, but I was running out of time. I decided to call it a night and started breaking down the setup.
The next afternoon I found that there was a setting in the manual rotator that allowed me to reverse the sense of corrections. Having it set the wrong way would have produced exactly the symptoms I saw the night before, so I was confident that things would work next time.
Both issues were probably a consequence of my having to restore NINA after a full lobotomy of my laptop earlier this year; the clean reinstall had put everything back to default settings.
August 20
So, two problems, both probably fixed, and a clear night. The moon was at first quarter and there was way too much overhead smoke, but I wanted to see if things were fixed so out I went to the club's Eagle Lake site. Everything worked and I decided to do a quick and dirty two-pane mosaic of the Veil Nebula. I had planned for a minimum of two hours, one for each pane, and started collecting light frames. The northern pane finished with 60 x 60 s frames, then the sequencer did its thing and went to the southern pane. 45 frames into that the clouds swept in and the night was over; the second pane got short-changed.
An hour per pane was not really enough to for a decent image, but it did give me data to practice mosaic-making. Here is the result:
 |
| Veil Nebula 2-Pane Mosaic (1/2 scale, click to enlarge) |
Comments:
- This needs more time on target!
- Shooting 60 s subs really helps preserve star color and allows essentially total eradication of satellite tracks, Yay!
- I would like to thin out the stars better during the next reprocess.
- About that lighter area to the right of the Western Veil: is it real or is it an artifact of how I flattened the field? It compares very well to this image in which it shows structure; there's also a discussion (and another excellent picture in RGB) of it on CloudyNights. Since the veil is often imaged using LP filters or in narrowband, most images might not show a dim broadband feature like this. I'm pleased that I'm able to draw it out with only an hour of imaging!
- The image was calibrated using dark frames, flats, and flat darks. Lights were dithered, but not sufficiently for drizzling. I had about a half dozen runs at processing this before I found a good way to merge the panes. The workflow I found that worked best is given below.
- Can you find the two seams? I can't despite knowing right where they are!
- [Added 8/28/2026] Three separate cropping operations (see workflow below) reduced the mosaic down from its maximum possible size of 6248 x 7308 (taking into account the original 25% overlap) to 5980 x 7076. That's a loss of 7.9 percent in terms of area, or about 3.8% linear loss. Not too bad, but I'm glad I started with 25% overlap and had room to spare surrounding the Veil.
August 22
The night of the 21st was cloudy, but the 22nd was clear. The moon was now two days past first quarter, and there was still too much smoke aloft; the sky looked very bright. My first target (Sh2-114, the "Flying Dragon Nebula") was such a lost cause that I stopped collecting data early hoping to have enough time for a quick and dirty image of the Cocoon. I got 45 x 60 s of light frames: here's how that turned out:
 |
| Cocoon Nebula (1/2 scale, click to enlarge) |
This would be gorgeous from a dark sky site with no moon, no smoke, and more time on target! There's a lot of entrained gas/dust around the dark lane that extends to the left, and it's only hinted at in my image.
One other glitch did arise this third night of four. The mount stopped unexpectedly well before it should have and signaled it had reached its safe limit. The next day I set it all up and made sure the mount had proper limits and was also correctly configured for meridian flips. I also made sure NINA was ready for flips, too. I'm going to attribute the glitch to the usual guilty party: operator error.
Verifying the change will need to wait for the next new moon, though. I'd also like to see if my setup can do a proper meridian flip.
Mosaic Workflow
This was all performed within PixInsight 1.9.4 build 1695, and I've provided most of my non-default settings for each process or script. PixInsight settings are meant to be explored, and you should find those that work best for your imaged objects and esthetics.
If you don't have the same scripts and processes installed as I use here, you will need to obtain them or find suitable substitutes. I also assume you have at least a rudimentary knowledge of PixelMath; there are two points at which PixelMath will come into play.
Calibration
WeightedBatchPreProcessing (WBPP) is used to do the calibration, and for these images I used the default settings. When it finished I had two master light frames, Pane1 and Pane2. I chose to have WBPP do autocropping even though it sometimes insufficiently crops out edges with weak signal. By the way: I've now stopped using bias frames as part of calibration; they're mainly useful for scaling dark frames with mismatched temperature or exposure time, and I never have need to do that.
Linear Image Background Normalization and Color CalibrationRun BlurXterminator with Correct Only checked; this will take care of aberrations.
Use the ImageSolver script to solve the images.
Apply SpectrophotometricFluxCalibration followed by MultiScaleGradientCorrection.
Apply SpectroPhotometricColorCalibration
If MSCG doesn't give you the results you want there's always ABE and DBE.
At this point the linear image is color-corrected and its background is reasonably flat. If you want, you can apply noise reduction, but in the following workflow star detection will be performed and it's possible that works better with the noise left in.
Preparing and Merging PanesI use star alignment to create registered panes ready for merging. There are other ways to do this, but it worked best for me for this particular data. This mainly entails creating an image full of synthetic stars that is larger than the eventual mosaic will be and then registering the panes to it.
I make a rough guess at the RA and DEC of the eventual mosaic's center. With only two panes it's easiest to just average the RA and DEC of the panes. I also guess the eventual size of the mosaic in pixels, and add some "elbow room" to make sure the panes will fit into it with room to spare. You can find the pane centers by looking at the image solutions using the PixInsight Property Explorer.
For the Veil mosaic the center is 20h 51m RA and 30d 33m Dec; the size is 6400 x 7400 pixels. The rotation is zero. Plugging these into CatalogStarGenerator produces a star map image named CatalogStars. If this image appears black to you, apply autostretch.
As noted above, the automatic crop feature of WBPP sometimes leaves some data-poor edges uncropped. Use DynamicCrop to trim these off to help GradientMergeMosaic work better. Note that this destroys the previous solutions so we will need to re-solve as needed.
Use StarAlignment to register the panes (the correct settings are reference image = CatalogStars, Registration mode = Two-Dimensional Surface Splines, Radial basis function = DDM Thin Plate Spline, Working Mode = Register/Match Images). This results in two registered panes.
The pane backgrounds must be matched, so use
DNALinearFit. This can be obtained at David Ault's
website, and more specifically
here. Please read Ault's instructions for the use of
DNALinearFit.
Next is more cropping! First, creat a temporary composite image Use PixelMath's max function like so: max(pane1,pane2) where pane1 and pane2 are the identifiers of the two registered panes. Remember to tell PixelMath that you want to create a new image, not overwrite one! Use DynamicCrop to crop away all the black portions of the composite and then apply the crop to both registered panes. Save these newly cropped images.
Merge the newly cropped panes using GradientMergeMosaic with type of combination = overlay, Shrink radius = 1, Feather radius = 10, and Blackpoint = 0. If you see pinched stars, you can rerun it with an increased feather radius or you can black out offending stars. At this point you will have a linear mosaic image.
Traditional Single-Pane Linear Post Processing
At this point processing follows your usual workflow for a single linear image.
Do serious color calibration by using ImageSolver followed by SpectroPhotometricColorCalibration.
Do serious background extraction by running SpectroPhotometricFluxCalibration followed by MultiscaleGradientCorrection. I use MGC as it usually gives nice results.
Now you have a choice to continue post processing on the mosaic as it is, or to split it into starless and star images. I went with the latter by applying StarXterminator.
You can do all sorts of things to the starless image; noise reduction and sharpening are probably the most important. I used NoiseXterminator and BlurXterminator (disable automatic PSF, set PSF to about 3 and sharpen nonstellar to 0.5). Apply and fiddle with the settings to taste.
At this point I did see some mild star pinch effects in the starless image and ended up correcting them by adjusting settings and blacking out some of the offending stars.
Because I was seeing some residual green in the star image, I applied SubtractiveChromaticNoiseReduction to both the star and starless images. I did nothing else to the star image.
Stretching
Starless image: I used
MultiscaleAdaptiveStretch with target background = 1.25 and scale separation = 256.
Stars image: the usual STF-HT method wildly overstretched the stars, so I used StarStretch with the stretch amount = 5 and color boost = 2. The stretch amount is where you can get very fiddly. It's worth exploring the entire stretch range of 4 to 6.
Combine Starless and Star images
Combine the nonlinear starless and star images with a PixelMath line ~((~$T)*(~stars)) where stars is the identifier of the stars image.
Final Tweaks
At this point you're free to do more noise reduction, adjust saturation, apply curves, get fiddly with dynamic range; it's up to you.
-------------------------------------
Next time: imaging at the September new moon. Will it be clear? Will it be smoke-free? Will I be local or camped at Lac qui Parle state park?
