This is my first bicolor in PixInsight, and only my second narrowband in PI. Below are the steps I used, with some pretty detailed notes and some questions. While I’m quite pleased with the image, I’m always up for improving! If you see any mistakes or have any suggestions, please let me know.
http://www.pbase.com/skybox/image/161478325A tricolor, adding in 13 hours of SII:
http://www.pbase.com/skybox/image/161484751 If you want to play with the data, the three stacked images (with DBE already applied – see below) are here:
https://dl.dropboxusercontent.com/u/58468743/SII_DBE.zip1. Batch Preprocessing for calibration and registration. I used darks, bias and flats, and registered everything to one of the Ha frames. I know some folks don’t use darks with Sony chips, but I always find my narrowband images require them.
2. Image Integration
3. Dynamic Background Extraction. I did separate DBEs on each channel, because I couldn’t find a set of sample points that worked well on all three channels. For example, the Ha DBE left large black areas on the SII image. Is there any downside to doing a separate DBE on each channel in this case?
4. Linear Fit. I used Ha as the reference. I compared using OIII, but the Ha was a bit suppressed that way. I really don’t know why, since the OIII signal looked as strong as the Ha.
5. Color merge with Pixel Math. I blended 20% Ha with 80% OIII for the synthetic green.
a. In this approach, isn’t the Ha suppressed from the luminance data? It’s only being used for “1.2 of 3” channels. If so, is there a better way to deal with that? Making an artificial luminance with 50/50 Ha and OIII?
b. Is there any advantage to doing the blend I did vs Ha:OIII:OIII? If I understand correctly, the blend I did simply skews the Ha slightly orange, and the OIII slightly blue of pure teal.
c. I tried Background Neutralization and Color Calibration, but felt they didn’t contribute anything – and maybe sucked out some of the Ha background – so I just skipped them.
6. Deconvolution, extracting a Dynamic PSF from the image. I extracted a luminance of the image, stretched that, clipped the black point, and used that as a mask to protect the background against sharpening. I didn’t create a local deringing support. I usually do, but here I found I could get away with just increasing the global dark setting a bit and adding a third layer of wavelet regularization, following this walkthrough by Alejandro Tombolini.
http://pixinsight.com/forum/index.php?topic=5148.msg35658#msg356587. Multiscale Linear Transformation for noise reduction. I use the k-sigma panel for noise reduction rather than the standard NR panel, again following a tip from Alejandro. I don’t do any NR later in my processing.
8. Histo Transformation. I did some preliminary stretching, until I got the stars about as bright as I wanted them.
9. “Star-bloat controlled stretching,” following Gerald Wechselberger’s detailed tutorial.
http://dl.dropbox.com/u/57910417/Howto_enhance_nebuala_without_pushing_stars.wmvThis is the only “star removal” technique I’ve come across that I’ve found really effective. I’ve looked at a bunch of techniques over the years, and never found anything that didn’t involve a ton of cleanup work. Gerald’s approach looked complex the first time I saw it it, but after doing it a couple times it’s pretty straightforward. I still have challenges with very bright stars deeply embedded in nebulosity, so I dealt with those separately (step 11). Saved the image.
10. The result from step 9 was a bit blown-out in the two brightest spots, so I went back to step 8 and stretched that a bit more with HistoTrans and CurvesTrans until the blown out parts were where I wanted them to be, and saved the image.
11. To deal with the really bright stars, I stretched the result from Step 10 to match the brightness of the result from Step 9, then used Morphological Transformation with a star mask to shrink the bright stars a bit. Saved the image.
12. In Photoshop, I layered the results of steps 9-11, using the bright spots out of step 10 and the bright stars from step 11. Is there a better approach to solving these two problems in PI?
13. Back in PI, I used deconvolution again with a parametric PSF, using a mask to protect the background. I tried to mask out the faint stars embedded in the nebulosity, but I couldn’t manage to make a star mask that would pick up those stars.
14. A final Curves Trans stretch, to bring up the OIII envelope and the faint Ha background. I used an inverted luminance mask to protect the stars and bright nebulosity from further stretching.
