IC 1848 / Sh 2-199 - Soul Nebula in SHO

[Herbert West]

 Weather's crap in Poland, so it's time to redo my old data and do it well this time arround.

Gear: 

EQ6PRO 

TS65Q 

ASI1600MMP 

ASIAir V1 

EAF 

EFW 

Filters: Baader Ha, OIII, SII, Astronomik DeepSky RGB 

 

Subs: 

Ha: 103x300s 

OIII: 107x300s 

SII: 92x300s 

RGB: po 60x30s 

Darks, Flats, Dark Flats. 

 

Workflow (100% Pixinsight): 

A. Linear processing S, H, O for RGB channels. 

A.1. Mure Denoise on S, O (variance 0.3, cycles 30) and H (variance 0.2, cycles 30) 

A.2. DynamicCrop on S, H, O 

**** DynamicBackgroundExtraction should be here but I, uhh, forgot about it :-P *** 

A.3. LinearFit H and O to S. 

A.4. TGVDenoise with mask on O, S- medium intensity, and H- low intensity. 

A.5. MultiscaleMedianTransform denoise with strong mask on O, S and H- with even brighter mask. 

 

B. Luminance layer processing (Ha stack) 

B.1. Copy Ha stack from A.3 stage. 

B.2. Very mild TGVDenoise with mask. 

B.3. Very mild MultiscaleMedianTransform deonoise on wavelt layers 1, 2, 3, 4 with strong mask. 

B.4. Deconvolution round 1 with custom global dark/bright settings, no local support or mask, 15 iterations. 

B.5. Deconvolution round 2 with custom global dark and increased global bright settings, increased wavelet regularization, 15 iterations, with local support. 

B.6. Deconvolution round 3 with slightly increased global dark and again increased global bright settings, slightly increased wavelet regularization; 10 iterations, with increased ammount of local support. 

B.7. Starnet2 to remove stars. 

B.8. Subtle MultiscaleLinearTransform sharpening on wavelet layer 3 only. 

B.9. Pixelmath to add stars back. 

B.10. Very mild MultiscaleMedianTransform denoise with very bright, low contrast mask on wavelet layers 2, 3, 4 to smooth out some middle-low frequency noise from deconvolution. 

B.11. HistogramTransformation. 

B.12. HDRMultiscaleTransformation. 

B.13. Starnet2 to remove stars. 

 

C. Channel combination 

C.1. Pixelmath to combine channels: R=.7*S+.3*H, G=H, B=O. 

C.2. Invert image. 

C.3. SCNR to remove magenta cast arround stars and in the background (it resulted from LinearFitting the S, H, O stacks). 

C.4. Invert image. 

C.5. Starnet2 to remove stars. 

C.6. HistogramTransformation. 

C.7. LRGBCombination with Luminance layer prepared in step B.13. 

C.8. SCNR with mask to remove a bit of green from some areas (oxygen dominated area left untouched). 

C.9. CurvesTransformation- a lot of it, with a bit of SCNR, with the following masks: 

- 3 masks made from H, S and O stacks with boosted contrast, removed wavelet layers 1,2,3 with MMT and denoised strongly with ACDNR, 

- 3 masks made from the above mentioned, but with more contrast, 

- mask made with pixelmath: max (S, O), boosted contrast, 

- range mask made with RangeMask script, 

- color mask made with ColorMask script. 

C.10 A little bit of LocalHistogramEqualization with very low contrast and low amount (scale 16 & 64). 

C.11 Very delicate ACDNR with range mask to protect the background (yes, the reverse of what’s usually done, but my background is smooth enough) on chrominance and luminance. 

 

D. Stars 

D.1. Pixelmath to combine R, G and B stacks. 

D.2. PhotometricColorCalibration with background neutralization. 

D.3. HistogramTransformation. 

D.4. LRGBCombination with stretched luminance from step B.12. The aim is to tighten the stars up with deconvolved luminance made of Ha stack. 

D.5. Another round of PhotometricColorCalibration. 

D.6. Starnet2 to extract stars only. 

D.7. Some CurvesTransformation tweaks. 

D.8. Delicate Convolution to soften the edges. 

D.9. A bit of ExponentialTransformation to make the bright stars pop a bit more.  

 

E. Final steps 

E.1. Pixelmath to add stars from step D.9 ($T + .9 * STARS_IMAGE - to clip it a bit) 

E.2. HistogramTransformation to brighten it up a bit for people with overly dark screens.

E.3. Frantic work on the background (curves, SCNR with masks) to make up for the lack of DBE :-)

E.4. Horizontal flip to make it look less like a pig. 

E.5. Save as PNG (full res).