M97 - Owl Nebula
M97, or the Owl Nebula, is a planetary nebula approximately 2,030 light years away in the constellation Ursa Major. It was formed from the outflow of material from the stellar wind of the central star as it evolved along the asymptotic giant branch (a massive red giant star). The nebula is arranged in three concentric shells, with the outermost shell being about 20–30% larger than the inner shell.
source: Wikipedia
NGC/IC:
Other Names:
Object:
Constellation:
R.A.:
Dec:
Transit date:
Transit Alt:
NGC 3587
Owl Nebula
Planetary Nebula
Ursa Major
11h 14m 48s
+55º 01’ 08”
03 Apr
88º N
Introduction
This object has been photographed on two different occasions, from different locations, using different equipment. The first version is referred to as M97, while the second version is referred to as M97b. The goal of the second image was to capture the faint ring of nebulosity that surrounds the core of this planetary nebula.
Conditions
For the first image, frames were taken on two consecutive nights from the backyard in Groningen, the Netherlands. M97 is circumpolar, so visible all year around. But in late winter, early spring, it is at its peak and visible most of the night around the northern sky. During the two sessions the moon was almost full and only around 50° separated from M97. So it was hardly dark with SQM values around 18.0 mag/arcsec².
The second image was taken the following year (2024) from the remote observatory site IC Astronomy, in Oria, Spain. Imaging took place at about the same time in the season and also with plenty of moonlight around. However, at this dark site, overall SQM values were significantly higher, something essential to image the faint nebulosity.
Equipment
Planetary nebulae are typically small in size and M97 is no exception. Therefore the Field of View of any telescope/camera combination is not so critical. the first image was taken using the Takahashi TOA-130 (FL=1000mm) in combination with the small sensor of the ASI533. The second image was taken using the Planewave CDK14 (FL-2563mm) and the full-frame Moravian C3-61000 camera. Interestingly the field of view of both setups is very similar, with the difference that the first is a square format, where the second is a rectangular format. The target can be very well imaged using broadband LRGB filters. But imaging with just H-alpha and OIII filters is also a feasible option. Given the strong interference of the moon, the latter option was chosen.
Telescope
Mount
Camera
Filters
Guiding
Accessoires
Software
Takahashi TOA-130, FL67 flattener, Pegasus Astro Motor Focus kit v2
10Micron GM2000HPS, EuroEMC S130 pier
ZWO ASI533MM Pro, cooled to -15 ºC
Astrodon 1.25” Ha/OIII (5nm) mounted, ZWO EFW 8-position
Unguided
Fitlet3, Linux Mint 21.1, Pegasus Ultimate Powerbox v2, Aurora Flatfield, Pegasus Astro Flatmaster 150MBox
KStars/Ekos 3.6.3, INDI Library 2.0.0, Mountwizzard4 3.0.1, PixInsight 1.8.9-2
Telescope
Mount
Camera
Filters
Guiding
Accessoires
Software
Planewave CDK14, Optec Gemini Rotating focuser
10Micron GM2000HPS, custom pier
Moravian C3-61000 Pro, cooled to -10 ºC
Chroma 2” Ha (3nm) and unmounted, Moravian filterwheel “L”, 7-position
Unguided
Compulab Tensor I-22, Windows 11, Dragonfly, Pegasus Ultimate Powerbox v2
Voyager Advanced, Viking, Mountwizzard4, Astroplanner, PixInsight 1.8.9-2
Imaging
Default settings for narrowband imaging were used for each setup. For the first image this was 5-minute exposures at gain 100. For the second image this was 10-minuite exposures at gain 2750. The first image consisted of a total of 13h of imaging. The second image on purpose was exposed for as long as possible, in this case 20h.
Resolution
Focal length
Pixel size
Resolution
Field of View
Image center
2800 × 2100 px (5.9 MP)
1000 mm @ f/7.7
3.76 µm
0.77 arcsec/px
36' x 27’
RA: 11° 14’ 46.959”
Dec: +55° 01’ 01.72”
Resolution
Focal length
Pixel size
Resolution
Field of View
Rotation
Image center
9576 × 6388 px (61.1 MP)
2585 mm @ f/7.2
3.8 µm
0.30 arcsec/px
48' x 32’
RA: 11° 14’ 47.956”
Dec: +55° 01’ 21.61”
Processing First Image
All frames were calibrated using Darks (5), Flats (25/filter) and Flat-Darks (50), followed by registration and integration, all using the WeightedBatchPreProcessing script in PixInsight. The OIII master had a much higher signal value than the H⍺ master. LinearFit was used to adjust for the difference. An HOO image was created using ChannelCombination. The resulting image was cloned so that one imaged could be processed for the stars, and the other for the nebula.
Stars: Getting star colour correct in an HOO image is almost impossible. But there are a few things you can do to get at least a bit closer than just white stars. First is to apply SpectroPhotometricColorCalibration. This adjusts white balance to have the best overall match of each stars’ colour from the database with the corresponding star in the image. To maintain colour during stretching, a combination of ArcSinghStretch and HistogramTransformation was applied. There are a few stars located within the nebula that can easily be obscured when boosting the nebula. To make sure they remained visible in the final image, they were stretched a little bit further, by means of a mask generated using the GAME script, essentially covering the nebula. A saturation boost with CurvesTransformation completed the processing of the stars. StarXTerminator (SXT) extracted the stars to create a Stars-only image.
Nebula: First step was to remove the stars using SXT. The image looked pretty clean with very little gradients. Still a DynamicBackgroundExtraction was applied to remove any residual gradient present. The nebula does show some structure, but it can be easily obscured by too much stretching. BlurXTerminator turned out to be a very good instrument to enhance and sharpen any existing structure. With the stars removed, the PSF value could not be measured automatically. Instead, the strength was set to 0.7, and the PSF diameter was adjusted to have a maximal effect without creating artefacts. Stretching was tried both using the GeneralisedHyperbolicStretch (GHS), as well as the more traditional HistogramTransformation (HT). In comparing the two results the HT version appeared more pleasing than the GHS version. Theoretically this is different than what one would expect, so this was probably more a function of operator skills than anything else. Now came a very tricky part. At this stage, the regular HOO image had a very teal-like look to it. And the red areas of H⍺ around the edges were largely overwhelmed by the OIII signal. Bill Blanshan has developed a very nice tool to properly balance the colours in narrowband images, the NarrowbandNormalisation script. This gives a lot of control to the final look of various colour palettes.
Finding a pleasing colour palette using the NarrowbandNormalization (NN) script. The initial result (left image) of the HOO channel combination resulted in a very teal-looking image, with not much detail in the Halpha outer layers. Applying the NN script in its default HOO-setting, using Halpha for lightness gives a much better overall blue hue to the nebula, but the Halpha areas became a bit yellow. Interestingly, selecting the HSO palette and boosting with the OIII and SII sliders resulted in an image that had a nice blue center area of the nebula, while bringing out nice red colours in the Halpha outer areas. Using Halpha for lightness still meant a bit of a dim image, but that could later be boosted with regular CurvesTransformation.
With the colours now in place, an extra boost was given to brighten the image up a bit, using CurvesTransformation. A final step of mild noise reduction using NoiseXTerminator created the final Nebula-only image. Nebula-only and Stars-only were now put together using Pixelmath. The object is pretty small compared to the Field of View, but other than cropping from a square to a 4:3 image ratio, the final composition was left as is. Only some finishing touches, mainly contrast enhancing, using CurvesTransformation were applied. When exporting the image and looking at the final JPEG it occurred to me that the stars in comparison to such a small target were a bit more bloated than desired. So BXT was used to only adjust the stars and make them a bit tighter. This resulted in the final image as shown above.
Processing workflow (click to enlarge)
Processing Second Image
All frames were calibrated using Darks (50), Flats (25/filter) and Flat-Darks (50), followed by registration and integration using the WeightedBatchPreProcessing script in PixInsight. Ha and OIII channels were combined in a HOO palette and stars removed using StarXTerminator. Structure in the nebula was emphasized by deconvolution using BlurXTeminator. At this point in time many different stretching methods were tried that would keep some structure and detail in the heart of the nebula, while bringing out the details of the faint nebulous ring that surrounds the nebula. The problem however was the gigantic dynamic range. The following approaches were tried.
GeneralisedHyperbolicStretch. This allows very fine control over where in the image you want to add contrast. Unfortunately, despite several attempts no satisfying results could be obtained.
HDRMultiScaleTransform. This gives the ability to first stretch for the fainter areas, and then bring back the structure and definition in the bright areas. But also here no satisfying results could be obtained, with generally weird artefacts popping up at the strengths that the effect was required.
iHDR is a new script and works the other way around compared to HDRMultiScaleTransform. You first stretch the image to make the brightest parts look good, and then apply iHDR to stretch the darker areas, while masking the brighter areas. This was getting fairly close, but still not to the extend I was hoping for.
A complicating factor in all the above attempts was that any use of masks failed. Range masks, GAME masks, colour masks, convoluted or not, all gave strange artefacts at the edges. The difference in stretch however is so huge any mask would show up in the final result as strange edges, or unnatural looking transitions.
The solution that was eventually applied, was ImageBlend. This is a recent script developed by Mike Cranfield. Like the name indicates, it allows to blend two images together, but in a Photoshop-like way and with an enormous amount of control. First there are a ton of different blend modes to choose from. For my situation, the LinearDodge gave the best results, although Screen and Colordodge also gave reasonable results. Besides the blend mode, various parameters can be further adjusted to create the best looking result. This includes opacity as the base, but also black point, midtones and highlights sliders give great flexibility
The ImageBlend script gives a lot of control over the blending process between two images. Blending mode and opacity are two obvious ones. But, as was important in this case, also black point, midtones and highlights can be adjusted in the process.
Two images were created to be blended together, each focused on one part of the image. One image was stretched with HistorgramTransformation to show detail in the center. Another image was stretched using multiple iterations of GHS and CurvesTransformation to bring out as much as possible signal from the nebulous ring around the nebula (halo). The colour of both images was defined using the NarrowbandNormalisation script, with blend mode 2 and Preserve lightness selected. The halo image was very noisy at this stage, so a separate run of strong noise reduction was applied using NoiseXTerminator. Interestingly enough this also created some extra definition in the halo. After blending the Center-focused image and the Halo-focused image, a decent compromis was reached.
In order to maintain both detail in the nebula, while showing the faint nebulous ring around the nebula, the ImageBlend script was used. First, two images with different focus were created. One image (left) stretched to show detail in the center. One image (middle) stretched to show as much as possible the faint nebulous ring around the nebula. The blended result (right) shows more or less the best of both worlds.
The stars were created from a fresh HOO combination that was first colour calibrated using SPCC. It turned out that there was still some green in the image, which was removed with SCNR. The stars were tightened up using a deconvolution step with BlurXTerminator. After that, the stars were removed using StarXTerminator. Stretching was done with GHS, where the first stretch was a 60% blended colour stretch to maintain star colours during the stretching process. Subsequent stretching cycles used the regular RGB method.
As a final step, the stars were added into the blended starless image, using PixelMath. A bit of a boost in saturation and adjusting the background to a value of 0.07 completed the final HOO image.
Conclusion
Processing an image like M97 is a challenging task because of the huge dynamic range. Many of the choices are made based on personal preferences. The original image did not show much signal in the raw images of the halo of nebulosity around the nebula, and no attempts were made to bring out this nebulosity. This did mean that all focus could be on the center of the nebula. And to be honest, I prefer that rending of M97. But bringing out the nebulous ring around M97 is a challenge in itself, and it is rewarding in its own right that this was achieved with this new dataset that was acquired under better (darker) conditions, with better equipment.
