Exposures, both at f/2.0 with custom white balance:

Background sky: 20 sec, ISO3200

Foreground: 210 sec, ISO400

This was originally intended to be a test of my new Rokinon 16mm lens but it has been so cloudy around here that I’m digging in the junk drawer for data to play with. Glad I did, because this turned out pretty well. The tricky part was figuring out how to apply the in-camera custom white balance after I converted my image to .xisf format. Rather than replace the foreground with the ISO400 image, I averaged the two together in the foreground so the difference in noise levels between the sky and foreground isn’t as sudden. My foreground range mask is imperfect due to the plant tops which were tricky to capture with range mask. With that said, I couldn’t process the background and foreground separately with much vigor because the imperfect mask would give away the trick. So I relied on contrast curves much of the time to brighten the foreground. You may notice the orangish color in the foreground that was caused by ambient light pollution. Hampton Roads was to my back. I toyed with balancing the colors there but decided to keep this rendition because it is more true to the scene. Altogether, this was a fun project and unlike all DSO work I’ve done before.

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

Processing Workflow (PixInsight)

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

Note: Both ISO3200 and ISO400 images follow steps 1-3. Noise reduction in step 4 is only applied to the ISO3200 image. They are blended in step 5.

1. Correct hot pixels, save .CR2 as .xisf (CosmeticCorrection, CFA, Auto Detect, Hot Sigma 1.3)

2. Apply in-camera custom white balance

2a. Using exifftool, extract Red_Balance, Blue_Balance and Average_Black_Level from .CR2 metadata. Since PixInsight reads the .xisf file as 16-bit, divide each of these values by 65536.

2b. Open .xisf image from step 1 as Pure Raw.

2c. Scale the image to account for the black level by applying PixelMath expression: ($T - Black_Level)/(Saturation - Black_Level). Here, I assume Saturation is the maximum 14-bit value (0.25 after dividing by 65536).

2d. SplitCFA

2e. Apply PixelMath expression to CFA0: CFA0*Red_Balance

2f. Apply PixelMath expression to CFA3: CFA3*Blue_Balance

2g. MergeCFA

3. Non-linear stretch (MaskedStretch)

4. Reduce noise (TGVDenoise in CIE L*a*b* mode, lightness edge protection 1e-4, chrominance strength 3 and edge protection 2e-3, iterations 200, local support extracted L image; applied w/ inverted lightness mask).

5. Blend foreground and background images in the foreground region

5a. Extract blue channel of ISO3200 image (greatest contrast between foreground and background in this channel) and use range mask to select foreground, clone stamping as needed.

5b. Apply PixelMath expression: 0.5*($T+foreground) using inverted range mask selecting foreground.

6. Contrast curve in the sky (CurvesTransformation w/range mask selecting sky.

7. Add color to star cores

7a. Duplicate image, blur with Convolution and increase saturation with CurvesTransformation saturation channel.

7b. Blend with original image through a star mask, PixelMath: 0.3*$T + 0.7*duplicate

8. Increase star sizes using star mask, MorphologicalTransformation, Dilation, Amount 0.7, Size 3, Diamond structure.

9. Increase star sizes using star mask, MorphologicalTransformation, Dilation, Amount 0.7, Size 3, Diamond structure, inverted way.

10. Neutralize sky at the very top of the image (BackgroundNeutralization using preview at top; range mask selecting sky)

11. Brighten foreground (HistogramTransformation, lower midtones slider to 0.45; range mask selecting foreground)

12. Increase saturation (CurvesTransformation, C channel)

13. Brighten image (HistogramTransformation, lower midtones slider to 0.39)

14. Increase saturation (CurvesTransformation, saturation channel)

15. Contrast curve (CurvesTransformation w/star mask protecting stars)

16. Brighten foreground (CurvesTransformation; range mask selecting foreground)

17. Increase contrast in foreground with a micro-contrast curve (CurvesTransformation; range mask selecting foreground)

18. Darken image (CurvesTranformation w/star mask protecting stars)

18. Remove chrominance noise in background (MultiscaleMedianTransform to Chrominance w/ inverted luminance mask)

19. Sharpen image (MultiscaleMedianTransform, bias +0.03 to layer 4 of 5)