North America – A way to obtain hubble-palette style colours with a DSLR, using a l-enhance filter

I tried to obtain hubble-palette style colours from data acquired with a DSLR, using a l-enhance filter, imaging the majestic North America nebula

The Subject

Very high in the summer sky, at least from my place in central Italy, lies the majestic North America Nebula and her neighboring Pelican Nebula. These emission nebulas are both strong hα and Oiii emitters, and for this reason are particularly suitable for narrowband astroimaging, even from polluted skies.

My widefield setup allows me to frame the entire complex in a single shot, with a large margin for dithering.

The Shooting

I took the shoot, as usual, from my backyard at the edge of the urban area of Arezzo, where the light pollution is quite evident (around Bortle 6-7). However, the least polluted part in my sky is towards north and east, and the subject culminates almost at the zenith, so the premises are not that bad. I used subs from two different sessions. The current warm summer affects negatively a DSLR without cooling: the shots from the first session are visibly noisier due to an air temperature around 26-27 °C, higher by 5-6 °C than the second session.

Here are the specifications of my widefield rig, I’m using an old but functional Canon EOS 1100D, with a full-spectrum astro modification and an Optolong L-Enhance double narrowband filter.

Given the narrowband shot and the good guiding by the G11 i used 10 minutes subs at ISO1600. This has guaranteed a good exposure on the faintest parts of the nebula, but has  bloated the most brilliant stars in the field.

0. Pre-Processing

For preprocessing I used PixInsight‘s new WeightedBatchPreprocessing script, leaving most of the options as default. I needed an entire set of calibration frames including bias, flat and dark, as per the diagram below. I found I needed the dark frames in order to fully remove a bad amp glow affecting this camera: I shot several dark frames to create a small library, trying to match the temperature of the lights as closely as possible, using the sensor temperature data written in the camera raw files. Also, I used 100 bias frames to create a Superbias using Pixinsight.

The image below compares a 1:2 crop of a single 600 seconds shot (to which only Debayer and Autostretch were applied) with an integration of 29 calibrated 600 seconds shots stacked using 2x DrizzleIntegration (again,  Autostretch was applied and the image was resampled to achieve the same scale despite the drizzling). Thanks to the narrowband filtering and the long exposure time, the nebula is clearly visible with details even in the single shot.

1. DBE and Narrowband channels extraction

I started this elaboration in PixInsight with a simple DynamicBackgroundExtraction in order to neutralize the light pollution gradient, then I extracted the  and Oiii channels from the color image, using a mix of the R, G and B channels. As a basic rule you surely can use the R channel for the  and a mix of the B and G for the Oiii, but I chose a slightly more complex channel mix using PixelMath: since this colour image isn’t calibrated and the elaboration is just for artistic purposes, I used the blend I liked most, based on empirical values I found on the Web and in forums.

These are the two narrowband channels I obtained with autostretch and a DynamicBackgroundExtraction applied. Regarding the L-Enhance filter, while the band of the hα is just 10 nm, that of the Oiii is larger, about 24 nm, and it includes most of the Hβ emission as well. The Oiii channel is noticeably noisier (despite this is not evident in this full image) and has more prominent stars due to the larger band: these problems will be solved in the following steps.

2. Separate narrowband channels processing: Deconvolution

I processed the two narrowband channels separately. As an example I’ll show you the processing steps on the Oiii channel: the processing on the hα channel has the same steps, it’s just a bit easier since hα data is cleaner.

On these undersampled and drizzled data, the first step should be a deconvolution. I first generated a custom Point Spread Function using the DynamicPSF process, then I tweaked the Deconvolution process in order to obtain a good result keeping the noise low. This kind of tweaking requires a lot of trial and error and testing, and a certain level of experience. A very good source of information on this subject can be found in Jon Rista’s website: https://jonrista.com/the-astrophotographers-guide/pixinsights/proper-use-of-regularized-richardson-lucy-deconvolution/

This is the result of the deconvolution on a 1:2 crop showing the Cygnus Wall area, with Autostretch applied. The result is quite impressive.

2. Separate narrowband channels processing: Denoise and Star Reduction

As usual I used a two-steps noise reduction. The first step with TGVDenoise using a low-contrast mask, in order to reduce high-frequency noise, and the second with MultiscaleMedianTransform to further reduce low frequency noise and specs. You can see the result on the sample below. I also reduced the size of the most brilliant stars using a contour mask and MorphologicalTransformation in order to reduce a the blueish halo that would result from the blend with the hα channel.

3. Channel Blending

I tried several channel combinations for the final RGB blending using PixelMath. I found that the most satisfying results came from a simple HOO blend and from a blended channels combination, where the green channel is generated with a 40-60% mix of the hα and Oii channels. Finally I decided to proceed with the simple HOO blend. These are the two blends after BackgroundNeutralization.

4. Star reduction

This star reduction step is important to bring the nebulosity out in a star-rich region. The effect is obtained using MorphologicalTransformation and an adequate StarMask, in two different stages. 

5. Stretch

The best result came with a MaskedStretch, that has the effect of further reducing the stars bringing out the nebulosity.

6. LocalHistogramEqualization

In order to bring out the details and local contrast, I used LocalHistogramEqualization with a conservative configuration, looking also for a more natural effect.

7. Curves correction

The CurvesTransformation process is powerful, and I used it not only for correcting the general contrast of the image, but mainly for altering image colors. Using a Hubble palette image as a reference, I found I can obtain interesting colours altering the a, b and c controls. Exploiting the tonal contrast proper of this subject it’s possible to obtain colors that are similar to an Hubble Palette Image.

Obviously, this remains a two-band image, not a true tri-band Hubble palette picture. Without having any scientific pretensions, this is just an artistic choice, aiming to obtain a pleasing effect. 

8. Black point adjustment and DarkStructuresEnhance

The final steps in PixInsight consist first in adjusting the black point with HistogramEqualization, then in using the script DarkStructureEnhance to further enhance the contrast of the beautiful structures of this nebular complex.

9. Final Color Correction in Photoshop

For the final color correction I switched to Photoshop, in particular the versatile Selective Color tool is very useful for the necessary corrections towards a Hubble-styled image, enhancing the saturation and giving the desired color balance.

This is the final image, at full resolution of 3402×2637 pixels.

Conclusions

I didn’t think I could obtain these colors and details with an old DSLR, a dual-narrowband filter and an undersampled image. A noticeable drawback is the poor, or absent, star color. I tried several times to obtain full color stars from the dual-band shots with the Ballesteros method, but without having much success. A possible solution could be shooting the stars separately wihtout the filter, I should try that in my next project.

In the meantime, thank you for reading, and clear skies!

Leave a Reply

Your email address will not be published. Required fields are marked *