Pictured here are Bode’s Galaxy, M81 and the Cigar Galaxy, M82. I spent seven nights imaging these galaxies in February, March and April, accumulating over 17 hours of exposure time from which I culled the best 13.6 hours of subframes. I’m impressed with how bright these galaxies are. They were visible as a fuzzy patch through my 50mm finder scope making them easy to find night after night. Seeing was pretty lousy on all nights but one, but I am thankful for the time I had under clear skies during those nights since a persistent deck of clouds has loomed over my fair city for over a month since.

I must say the IDAS-D1 filter works admirably in my orange zone light pollution environment. My previous work on galaxies with the CLS-CCD filter took a big bite out of yellow that was difficult to recover in post-processing (see my M33 - yikes!). With this filter however, I feel yellow was better preserved and star halo colors have nice variety.

I am pretty well convinced that my subexposure times were too long. For a while there I was trying the “expose to the right” philosophy (actually I exposed to 50-60%), but many star cores and even M81’s core was overexposed so I had to use some post-processing tricks to make them look presentable. That philosophy worked well for imaging nebulosity from my home base, but at the expense of overexposed star cores. For my f/7.5 scope with the IDAS-D1 filter and my light pollution levels, 10 minutes is too much (BYEOS histogram peak for highest channel around 50%) and 5 minutes is too little at ISO800, so next time out I’ll give 7 minutes a shot.

One new thing I tried with these imaging sessions was using my laptop screen for flats using the BYEOS AV mode and +2EV on the camera rather than my homemade flat box. Unfortunately the flats exposure length was not enough because at least one RGB channel was underexposed enough to increase noise in along the edges. After calibration and integration, the master light frame had a noisy red boarder around all edges extending for probably 15% into the image. Quite disappointing! A recent thread on CloudyNights suggests that the mirror shadow can cause a horizontal band of vignetting along the bottom of an uncalibrated image which is consistent with what I saw on the top and bottom of my frames given that the orientation flipped 180 degrees at every meridian flip. Correct flats should have resolved this, but they did not so I need to establish the correct flat exposure time and investigate if light leakage through my DSLR eyepiece is a contributor.

To overcome this flat fielding failure, I implemented David Ault’s synthetic flat technique prior to DBE. Indeed it worked wonders, but since I used excessive clone stamping to develop my synthetic flat, I did not attempt to reveal any IFN during post-processing. Even though IFN permeates this region, I would not trust the morphology because my hasty synthetic flat would have manipulated the result. I guess that’s my penance for botching the flats.

For linear noise reduction, I used TGV Denoise and MMT, but with stronger masks than usual to protect the galaxies’ details. HDRMT worked well to enhance the spiral arms of M81 and the H-alpha structure in M82. I only blended 50% with the original image to avoid noise enhancement. I went to grand efforts to retain star colors - during stretching I expanded the top end of the dynamic range twice then afterwards used Vicent Peris’ method to recover star halo colors and Jerry Lodriguss’ method to stuff some color back into the star cores. The Peris method took some extra work to repair the pink and purple star cores prior to combining with the original image. As usual, I increased the star color saturation separately from that of the galaxy to avoid oversaturation. I took several passes at increasing color saturation of the galaxy, at times applying chrominance noise reduction with ACDNR before increasing it more. The final color saturation is about to my taste which is a good thing because the data quality cannot support any more (I may have even pushed it too far with respect to data quality).

Even with fairly aggressive deconvolution, two passes of MLT for sharpening and a touch of unsharp mask, the details are still mushy upon close inspection due to the poor seeing I experienced and noise reduction. I am happy with the level of details available when viewed without zooming in however (LOL). This is in part a testament to the struggles of imaging in light polluted environments. After 13 hours of integration, I’m still craving more photons to tame the SNR so I can push the data harder. And wouldn’t it be nice to have Arizona seeing to? Dare to dream! Still, I’m glad that I can get an image of this quality from my home base. Galaxy season is tough, but these were fun targets.

-------------------------------------------------------------------------------------------

Processing Workflow (PixInsight)

-------------------------------------------------------------------------------------------

1. Initial crop (Dynamic crop).

2. Flatten background due to poor SNR in flats along image borders (David Ault synthetic flat technique, reference below).

2a. Clone image, name synthFlat

2b. Remove all layers but residual (ATrousWaveletTransform, 7 layers)

2c. Use CloneStamp to remove residual stars

2d. Repeat previous two steps

2e. Flatten background of original image with PixelMath expression, $T/(synthFlat/median(synthFlat))

3. Remove remaining gradients, hardly necessary after previous step (DynamicBackgroundExtraction, subtract).

4. Neutralize background (BackgroundNeutralization).

5. Set white balance (ColorCalibration disabling structure detection).

6. Set luminance coefficients to 0.333333 for RGB channels (RGBWorkingSpace).

5. Non-linear noise reduction using David Aults’ recipe for mask generation and process

7. Deconvolution with lightness mask

8. Reduce background noise (TGVDenoise with inverted luminance mask).

9. Reduce background luminance noise (MultiscaleMedianTransform to Luminance)

10. Reduce background luminance noise (MultiscaleMedianTransform to Chrominance)

11. Increase color saturation (CurvesTransformation to Saturation with lightness mask).

12. Non-linear stretch, lower midtones slider of HistogramTransformation

13. Increase top end of dynamic range, increase high range slider to 1.1 in HistogramTransformation

14. Stretch image

14a. Non-linear stretch, lower midtones slider of HistogramTransformation

14b. Increase top end of dynamic range, increase high range slider to 1.1 in HistogramTransformation

14c. Non-linear stretch, lower midtones slider of HistogramTransformation a little less this time

14d. Increase top end of dynamic range, increase high range slider to 1.1 in HistogramTransformation

14e. Non-linear stretch, lower midtones slider of HistogramTransformation a even less this time

14f. Lower blackpoint slightly with HistogramTransformation

15. Expand dynamic range of details in galaxy

15a. Clone image

15b. Apply HDRMultiscaleTranform with To Lightness and Lightness mask selected.

15c. Blend into original image with PixelMath expression, 0.5*$T_2+0.5*M81_HDRMT using inverted star mask.

16. Recover star profile shapes and colors using Vincent Peris method (reference below).

16a. First off, repair pink and purple star cores. Duplicate image.

16b. Extract red channel of duplicate image (ChannelExtraction) and then use RangeSelection to select the pink star cores...easy to do because they stand out wildly in the red channel.

16c. Extract H, S and I channels of duplicated image.

16d. Apply pink star core mask from step 16b to the S channel and use CurvesTransformation to drop the S value of the star cores to zero.

16e. Recombine H, S and I channels using ChannelCombination. It took some experimenting with the fuzziness and smoothness parameters in RangeSelection to get the mask correct which blended the repaired portion of the star cores with the non-saturated perimeter of the stars.

16f. I used this repaired image for the large scale images used by the Peris method to recover star profile shapes and colors. I applied the final PixelMath step in his method 8 times.

17. Reduce green cast (SCNR to green, Amount 0.9)

18. Increase color saturation of galaxies where color saturation is low (CurvesTransformation to Saturation, boosting the low end, but not the top, applied 2 times with star-subtracted mask selecting only galaxies).

19. Increase color saturation of galaxies (ColorSaturation increasing red-orange-yellow and blue with star-subtracted mask selecting only galaxies).

20. Push color from star halos back into star cores (Jerry Lodriguss method)

20a. Duplicate image

20b. Blur duplicated image with Convolution

20c. Increase color saturation of duplicated image (CurvesTransformation to Saturation).

20d. Blend duplicated image with original through a star mask using PixelMath expression $T*.6+.4*M81_clone

21. Increase color saturation of stars (CurvesTransformation to Saturation).

22. Reduce star sizes (MorphologicalTransformation with star mask; morphological selection, iterations 2, amount 0.5, selection 0.2, structuring element size 5).

23. Reduce chrominance noise (ACDNR to chrominance with lightness mask).

24. Increase local contrast (LocalHistogramEqualization, Kernel Radius 32, Contrast Limit 1.5, Amount 0.50, Circular Kernel deselected; with star-subtracted mask selecting galaxies)

25. Decrease background brightness (apply inverted range mask and increase midtones slider of HistogramTransformation slightly)

26. Increase color saturation of galaxies (CurvesTransformation to Saturation with star-subtracted mask selecting only galaxies).

27. Increase color saturation of stars (CurvesTransformation to Saturation with star mask selecting stars).

28. Contrast curve to bring out M81’s spiral arms (CurvesTransformation with star mask).

29. Increase color saturation of galaxies (CurvesTransformation to C channel with star-subtracted mask selecting only galaxies).

30. Reduce chrominance noise (ACDNR to chrominance with lightness mask).

31. Sharpen galaxies (MultiscaleMedianTransform,5 layers total; biases of +0.1 on layers 3 and 4, respectively w/ lightness mask selecting only galaxies).

32. Sharpen galaxies (MultiscaleMedianTransform,5 layers total; biases of +0.1 and 0.05 on layers 4 and 5, respectively w/ lightness mask selecting only galaxies).

33. Sharpen everything slightly (UnsharpMask StDev 2.0, Amount 0.44 w/ lightness mask barely selecting galaxies and stars - I wanted this to be a very mild effect).

34. Decrease background brightness (apply inverted range mask and increase midtones slider of HistogramTransformation slightly)

35. Small contrast curve to make galaxies brighter and background darker (CurvesTransformation).

36. Reduce background brightness to final value (apply inverted duplicate of the maine image and increase midtones slider of HistogramTransformation slightly.

37. Final crop (DynamicCrop)

38. Have a beer.

References:

“Synthetic Flats with PixInsight” by David Ault

http://trappedphotons.com/blog/?p=756

“M42 PixInsight Tutorial” by David Ault for linear noise reduction

http://trappedphotons.com/blog/?p=642

“Dynamic Range and Local Contrast” software tutorial by Vicent Peris:

http://pixinsight.com/tutorials/NGC7023-HDR/index.html