Hi Fillipo,
here is my try with your data. The process is explained below. I did not manage to get the fine detail that Gerald's version showed, but in my opinion I have nicer nebula, in particular the core of M42, Horsehead and Flame nebula. For me, the ideal image would use Gerald's Version for the background, and mine for foreground.
Let me know where I can upload the full Image (424 MB) for you.
Georg
Process with some comments:
- you dont need so many different exposure times. Use only 240secs and 30 secs in future. The 30 secs are necessary to resolve the core of M42
- You need flats, darks and bias for proper calibration. Much of the difficulty of processing your data comes from the missing calibration.
- screen images using Blink process. I chose to keep all images, although the ..63 image seems to have tracking issues. But we want to see faint details and in doubt sacrifice some crispness, so we keep all the data. The satellite tracks seen in the images will disappear during integration.
- register images with ImageRegistration. Default settings except: Use one of the 30 second images as reference (hoping that it has well defined stars), Registration model=2d splines, Distortion correction selected (hoping to adjust for lens distortions). Note that your raw images show clear coma, which is probably much stronger than the lens distortions...
- Integrate images with ImageIntegration. Default settings except: Generate 64 bit result selected (due to high dynamic range), Weights= Exposure time, Rejection Algorithm=WinsorizedSigmaClipping, SigmaHigh=SigmaLow=4.0. This weights the images according to their relative strength (=exposure time), and rejects the satellite trail (or so I thought. It is visible in the final image, so I would need to do stronger rejection...)
- Integrate the two 30 second images with PixelMath (Cannot do it with ImageIntegration because only 2 images). RescaleResult, Use 64bit working image, Target Sample Format=64 bit.
- On Full integration:
- Correct for Vignetting with DynamicBackgroundExtraction. Using defaults except Tolerance=5 (due to strong gradient), Correction=Division (because it is the right mode to correct vignetting, not sky gradient). Placing samples automatically, then moving them carefully to avoid any nebulosity or stars (needs a lot of care). Note that this is only a bad replacement for proper calibration, and of course it can not fix the dust particles on your images.
- do second DBE with same settings. Slightly improves it.
- BackgroundNeutralization with defaults
- ColorCalibration with defaults
- On integrated 30 second images:
- same actions as on full integration
- Save both images, then combine with HDRComposition. Default settrings, Binarization Threshold=0.2. This restores the core of M42 as much as possible.
- SCDNR with default settings to get rid of green noise
- Dynamic crop to useful part of image
- Delinearize with sequence of relatively mild HistogramTransforms
- HDRMultiscaleTransform with defaults, to bring out the core of M42
- Denoise with TGVDenoise, defaults except Lightness: Strength=50, EdgeProtection=0.001,Iterations=500, Chrominance: Strength=70
- ColorSaturation to push red
- Final Tweaks with CurvesTransform (RGB, Saturation)
