SharpCap Forums

Reflections on my progress with this object and imaging in general.

February 2019

Captured with a no-name refractor of focal length 400mm using an Altair 183m and 183c. Blobby stars due to poor focus and over-stretching back then.



Dave

February 2025.

The same 183m camera with a Vixen refractor and filter wheel.

Influences since post #1:

• More attention paid to polar alignment & mount alignment.

• Focusing with a Bahtinov Mask.

• Addition of guiding.

• Calculated exposures and higher gain to cater for the QE curve of the camera.

• Serious exposure lengths for narrowband - now using 600s for Ha.

• The influence of the books Lessons from the Masters & Inside PixInsight and now having enough experience to understand what the authors are talking about.

• Starting to understand the use of masks to control which parts of an image to enhance and also avoid introducing noise when stretching

This image was weak in OIII plus it was captured as the object was sinking into the light pollution of the Manchester sky.

Progressing slowly .........

Dave

Hi Dave,

Nice work. I always get the impression that - aside from the planetary nebula where the high temperatures can make it the dominant emission- there are only a few objects where the OIII signal is even 5% of the HA and that avoiding adding too much noise into an otherwise good and detailed HA image takes spending disproportionately much more time on the OIII. It depends ultimately what you want to depict i.e. 1) A colour-balanced image which shows nicely where the OIII is located (as in fact are most images on Astrobin where most all of the HII nebulae OOH images photogenically show lots of blue to balance the red irrespective of the strength of the OIII signal in the particular nebula) or 2) An image that reflects the actual relative emission strength of HA v OIII but which then often renders the OIII contribution so subtle as to be barely discernable.

However - in many (but not all) of the objects the OIII doesn't actually add anything in terms of structural detail and all the OIII signal just underlies a stronger HA signal everywhere in the image - which makes scientific sense because OIII emission is more temperature-dependent and strongest at the centres of HA nebulae in the immediate surrounds of the ionizing stars. So - to me - collecting lots of OIII seems virtuous but ultimately rather dull in terms of what it adds to the final image. For things like the rosette I ended up falling into a compromise position wherein I'd collect enough OIII to be able to stretch it up to get a reasonable blue patch to locate it and add not too noisy colour information but not use it's contribution to luminance.

So, overall -- for some objects anyway, and probably including the rosette -- I personally think it is justified to collect as much OIII as necessary to provide colour information that you can stretch up to indicate where the OIII is but to then just use the HA luminance to preserve the nice detailed structure without contamination from the OIII noise. Just my take on it anyway - I was wondering whether you have also done the same above ?

Tim

Thanks Tim.

For the Rosette, I captured 30x600s of Ha and 20x420s of OIII. The OIII was probably captured too near to the Manchester skyglow but I am left feeling that I would have been better off with 20 more frames of Ha. All my struggles in PixInsight only seemed to degrade the image when OIII was included. To date, I haven't had much success with OIII. I really need to capture my images in the east where I have darker skies.

The Rosette is quite large for the equipment I have and it really warrants a 2-panel mosaic which would give a better framing when the object is surrounded by some darker skies.

Dave

I was never satisfied with the HOO image in post #3. I have found a PixInsight tutorial which discusses colour calibration of narrowband images at https://www.youtube.com/watch?v=LB22afHqIoI. Following the advice there, I have reprocessed the data and ended up with:



I managed to achieve a better gradient removal model this time.



This is the gradient model Autostretch (boosted) - thank you Manchester skyglow:


The problem with the data is that there is not enough OIII and the position of the nebula was heading into bad light pollution.

The PixInsight processing steps were:

• ChannelCombination: HOO -> RGB

• ScreenTransferFunction: autostretch, unlink, autostretch

• GradientCorrection: with structure protection

• SPCC colour correction: narrowband settings from video

• ScreenTransferFunction: link, autostretch

• ScreenTransferFunction: set sky level background to 0.15

• ScreenTransferFunction + HistogramTransformation: de-linearise image

• MultiscaleLinearTransform: noise removal by removing layer 1 (small scale structures)

• Create RangeMask

• Curves/Colour Saturation using RangeMask

• TGV: denoise

Dave

Hi Dave,

Thanks for sharing those details. I use many of the same steps - although using RC Noise Exterminator at the end rather than the MLT and TGV denoise combination.

The gradient removal area is getting interesting. On extended objects like the Rosette accurate gradient removal has been impossible up to now at longer focal lengths. I recently downloaded the latest MARS database onto PI and there is much more coverage now and for those objects -and especially those covered at both broad band and NB - it is revolutionary. I was able to go back and re-process old images of the heart nebula for example and repost pictures on Astrobin having much more confidence that they are correct.

OIII remains problematic though for many NB objects - some seem to have almost none. It is usually just so weak --10-20X and more - relative to the HA as well as inherently more subject to bright sky gradients. Imaging OIII for the lengths of times that would bring its signal up to match the SNR of the HA is unfeasible -- not to mention very dull. So - as far as I can see - it always gets stretched up much more than the HA to provide some sort of colour balance but that process inevitably brings in much more noise to the final image and spoils the HA detail.

Philosophically the way I look at it is that - since OIII is a much weaker signal that for many emission nebulae entirely underlies HA (obviously not in all - not Wolf Rayet for example) - it is - in these cases- valid to forego the the small OIII contribution to luminance (and large amount of noise) and to just use the nicely detailed HA luminance and the LRGB process for the entire starless image (i.e. do all the work on the starless images). That way you'd end up with an image with the detailing of the HA image of Feb but with diffuse bluey/ whitish regions marking the OIII locations. I don't know if this is general or accepted practice -- maybe cheating? -- but NB seems to be an area with no 'correct' method.

Tim

Tim

Thanks Tim. The PI steps are pretty much derived from the official PI videos on Youtube https://www.youtube.com/@PixInsight. I find these excellent - short duration (10 - 15m), well scripted, well structured and free from ERRMM.., UUMMM.. and LIKE.

I haven't tried the MARS database yet as I am still on PI 1.8.3. The upgrade version, 1.9, requires a CPU with AVX2 and FMA3 instructions - my old (was Win 7) HP workstation now showing it's (15 year) age. I have sourced a more modern PC (ex-corporate Lenovo) and am gathering myself for the gruelling migration process (unless I opt for a Linux installation).

I agree with you about OIII - it promises so much but seems to deliver so little and eats good sky time. At the moment I am trying to image galaxies using LRGB and having an Ha component (e.g. arms of the Pinwheel Galaxy).

This winter I am going to try NB filters in an SHO palette. Good candidates look like NGC 281, IC 410, Rosette Nebula, Heart & Soul Nebulae, M27 [silently regrets having purchased a 5 slot filter wheel instead of a 7 slot ;-( ].

I now have an Atik 314L+ which has set point cooling, so cloudy nights have been spent working on a darks library.

Dave

Another attempt using the steps in post #5. This time a gentler application of the later processing steps.



A comparison with the post #5 result. Time spent with PixInsight starting to show benefits.



Dave

Another attempt at processing using PixInsight/Starnet++ to create 'nostars' and 'starsonly'. These were processed independently, then recombined using PixelMath.

Image is HOO, Ha 20x600s, OIII 20x420s giving a total integration time of 5h 40m. The camera used was an Altair 183M.

This is the Ha only version - Ha 20x600s, total integration time 3h 20m.

Dave

Restacked all the data and reprocessed using knowledge/tools acquired since last April. This version is HOO + Ha stars, captured with an Altair 183m camera plus Altair Ha (7nm) and OIII (6.5nm) filters. ~6h integration. Processed with PixInsight (NarrowbandNormalisation script) plus StarNet++.



The PixInsight processing steps were:

• ChannelCombination: HOO -> RGB

• ScreenTransferFunction: autostretch, unlink, autostretch

• GradientCorrection: with structure protection

• ScreenTransferFunction: set sky level background to 0.20

• ScreenTransferFunction + HistogramTransformation: de-linearise image

• Export to 16-bit TIFF (StarNet requires this)

• StarNet: create starless HOO image

• NarrowbandNormalisation script: process starless image with Mode2 colour scheme

• ACDNR (lightness & chrominance): noise reduction, Amount = 20 applied twice

• Repeat above steps for Ha image

• Pixelmath: Ha - Ha_starless = Ha_stars

• Pixelmath: HOO_starless + Ha_stars = HOO_final

• HistogramTransformation: adjust black point and mid-point

• Curves/Colour Saturation

• UnsharpMask: mild sharpen

• DarkStructureEnhance script

Looking at the unprocessed stacks I probably can reduce the exposures with this camera, particularly in the OIII case where there is evidence of star bloat.

A far cry from the images at the beginning of this post, always worth reprocessing at intervals as new tools and increased knowledge can improve existing data.

Dave