Really struggling to get a good image of the Horsehead & Flame Nebulae

[smr]

I cannot understand why I'm getting an underwhelming stack on my Horsehead project. I've got 6 hours in total - 3 hours with my L-enhance filter and 3 hours without after I realised that the stack with the filter resulted in a completely red nebulosity - blocking the other colours in this region of Space.

I have three stacks in total, one 6 hour stack combining both filter and non-filter stacks, and two seperate stacks 3 hours each - filter and non filter.

Just applying simple stretches to reveal what's there and how much I can stretch (linearly) before too much noise is introduced, well they're all pretty underwhelming. I'm in a Bortle 5/6 zone and my Camera is the 2600MC Pro.

I'm seeing other posts on astronomy groups with 2 hours, OSC and same Bortle, and much more detail than I'm able to pull out.

My stacking method and the data :

For the first imaging session with the l enhance filter on, I didn't take any flats, and applied no calibration frames other than some dark frames (these dark frames were taken 28th Feb 2021 - could it be that I need to take new dark frames?) - the dark frames match the length of the subs (180 seconds)

For the second stack, I removed the filter for a full RGB image, which is how I'm going to keep adding the data, and took another 3 hours - this time with flats and flat darks, and the darks applied.

Adding all of the data results in a stack of 6 hours which looks very noisy even though I'm not stretching it that much, and there's no where near the nebulosity I am seeing in other images with less than half of this imaging time.

If anyone can shed any light as to why, or if you could look at the stacks if you want to see why I'd really appreciate it.

 

Horsehead 5h57m 119Lights.TIF

Horsehead Gain 100 Offset 50 No Filter 2h51m 57lights Darks and Flats and Flat Darks.TIF

3h6m 62 Lights only Darks applied Gain 100 Offset 50.TIF

[JSeaman]

I'm not seeing an issue with it, just a DBE in PI and then crop plus levels/curves in photoshop gave this:

[smr]

2 minutes ago, JSeaman said:

I'm not seeing an issue with it, just a DBE in PI and then crop plus levels/curves in photoshop gave this:

That's what I'm getting, it just doesn't look much different to my 6 hours with an unmodified DSLR. I would have thought I'd be able to get a more detailed and cleaner image with my 2600MC Pro.

I have actually just converted it to black and white and can see quite a bit more detail but it's very noisy.

 

[JSeaman]

I see, what duration subs are you taking?

 

[smr]

1 minute ago, JSeaman said:

I see, what duration subs are you taking?

 

3 minute subs

[JSeaman]

What's the limiting factor, could you push 5/10 minutes?

[smr]

On 10/01/2022 at 17:57, JSeaman said:

What's the limiting factor, could you push 5/10 minutes?

Would it be beneficial to go for longer subs than 3 minutes? I think I've read somewhere that as long as the data isn't clipping and off the LHS of the histogram then there's really no benefit with the IMX571 sensor doing longer 5 or 10 minute subs?

[ONIKKINEN]

2 minutes ago, smr said:

Would it be beneficial to go for longer subs than 3 minutes? I think I've read somewhere that as long as the data isn't clipping and off the LHS of the histogram then there's really no benefit with the IMX571 sensor doing longer 5 or 10 minute subs?

Narrowband especially from darker skies will still benefit from longer subs.

Broadband from light pollution swamps read noise in a few seconds with an IMX571 which is where the short subs are ok - idea comes from.

[smr]

I've just started imaging the Horsehead, Bortle 5/6 without filter, 180 second subs. 

Edited January 11, 2022 by smr

[vlaiv]

Well, for start - use 32bit floating point format instead of 16 bit format with camera that is already 16 bit.

You are truncating a lot of quality data this way.

 

[smr]

4 minutes ago, vlaiv said:

Well, for start - use 32bit floating point format instead of 16 bit format with camera that is already 16 bit.

You are truncating a lot of quality data this way.

 

Not heard of this? I use Deep Sky Stacker and Photoshop. I think PS only supports editing in 16 bit ?

[vlaiv]

Just now, smr said:

Not heard of this? I use Deep Sky Stacker and Photoshop. I think PS only supports editing in 16 bit ?

Well, it is 2022 - time to switch to software that supports 32bit float point images. Gimp does it. DSS can export 32bit float point data without problems.

[The Lazy Astronomer]

2 minutes ago, smr said:

Not heard of this? I use Deep Sky Stacker and Photoshop. I think PS only supports editing in 16 bit ?

1 minute ago, vlaiv said:

Well, it is 2022 - time to switch to software that supports 32bit float point images. Gimp does it. DSS can export 32bit float point data without problems.

In addition to GIMP, some examples of popular astro-specific post process software which supports 32-bit include (in order of cost): Siril (free), Startools, Pixinsight.

On another note, 3 min subs is probably ok - what scope are you using and what gain are you using?

[adyj1]

For 32-bit there's also Affinity photo, which is not a lot of money and has some useful tutorials on the 'tube.... 

[vlaiv]

I quickly did process of that no filter version and here is what I came up with:

Now, I did not mess with saturation and all that - just basic levels and some noise control.

This was done in Gimp.

My recommendation would be:

1. Use 32bit float point

2. Think of binning the data as it is oversampled.

3. Apply careful noise reduction

 

[smr]

Am I really losing a lot of data by using 16bit TIFFs instead of 32bit? I just had a look to see if PS does support 32bit and saved a stack as 32bit but you can't use curve adjustments so clearly it doesn't. 

The thing is all of my action tool sets and my experience and knowledge is in Photoshop. I wouldn't mind giving PixInSight a try but I'd need a new computer first. 

When you say binning the data do you mean binning 2x2 in Astrophotography Tool? 

[Ags]

What is your local light pollution?

[vlaiv]

20 minutes ago, smr said:

When you say binning the data do you mean binning 2x2 in Astrophotography Tool? 

I usually bin in ImageJ where I do some of my linear processing (like background removal and such).

Some other free software also supports binning.

I think it is better to bin later in processing stage (when the data is still linear) than at imaging time as you loose flexibility to decide if you need to bin and by how much.

22 minutes ago, smr said:

Am I really losing a lot of data by using 16bit TIFFs instead of 32bit?

Your camera is already 16 bit - so each sub you record is 16 bit.

When you stack two subs - you create their average. Say you have number 7 in one pixel in first sub and number 6 in same pixel in second sub. Their average is clearly 6.5 - which you can write in float point precision - but need to round up or down in 16bit integer precision (as it can store only whole numbers).

There you go - we just introduced error of 0.5 by rounding in above example - error that we don't need to introduce if we use 32bit float point format.

In fact, you can stack hundreds of subs in 32bit float without error - with even full 16 bit cameras, and if you stack even more than that - error that you introduce is exceptionally tiny due to nature of float point precision (floating point thing makes sure error is tiny).

If you don't use 32bit - you'll struggle to get smooth faint areas after stretch  - imagine you have pixel value of 3 and pixel value of 4 and nothing in between - that is just two shades, but in float point you can have 3, 3.001, 3.002,..... up to 3.999 (in fact not only 3 digits - but more than ten digits) of shades. This means that you'll have smooth transition after stretch rather than posterized in that value range.

[vlaiv]

Here - look at histogram of 32bit float point version after binning (binning adds fractional parts same as stacking) and after initial stretch:

versus same image, same stretch - only 16bit version:

Above histogram is "full" with smooth variation numbers and bottom one has distinct few values.

[Laurin Dave]

A quick Pixinsight process gave this...  some good detail but possibly a bit noisy for the integration time (compared to my Bortle 4/5) , what scope were you using and where was the Moon?

Edited January 11, 2022 by Laurin Dave

[gorann]

Just to let you know that using PS and 16 bit does the trick more than well enough and that is what a big bunch of us are using and are happy with, whatever Valiv tells you, and it can even give you APODs and IOTDs. If you know PS stick to it, PI is a very different animal (I do cherry pick some processes in PI), or what do you think @ollypenrice.....

[smr]

1 hour ago, Laurin Dave said:

A quick Pixinsight process gave this...  some good detail but possibly a bit noisy for the integration time (compared to my Bortle 4/5) , what scope were you using and where was the Moon?

Very nice editing though, colours look great, which stack is that with? It is actually looking much better than what I managed with my DSLR (to be fair it was an unmodified DSLR) but as you say quite noisy but there's detail there..  I suppose ultimately it's about stacking many more subs to pull more data with a smoother and cleaner image. Moon wise, this was on the 4th and 5th of Jan last week and the Moon is 70 percent tonight, so last week it really wouldn't have been an issue.

I would say my sky conditions are Bortle 5/6, I live on the very outskirts of a village so to the North it's relatively dark (a good Bortle 5) as it's farms and countryside for miles until the next village - no houses or light pollution, the other direction to the South is the centre of the village, so the Horsehead is low in the sky and in this direction, and more Bortle 6 I would say.

[Budgie1]

I took your 5h57m stack and put it through my normal workflow in PI. After stretching, I removed the stars and processed the background & nebula separate to the stars, then put the background through Topaz AI Gigapixel to further reduce the noise and sharpen, before adding the slightly reduced stars back into the image.

Here's the results.

[smr]

1 minute ago, Budgie1 said:

I took your 5h57m stack and put it through my normal workflow in PI. After stretching, I removed the stars and processed the background & nebula separate to the stars, then put the background through Topaz AI Gigapixel to further reduce the noise and sharpen, before adding the slightly reduced stars back into the image.

Here's the results.

What kind of sorcery is this!! Wow!

[Budgie1]

14 minutes ago, smr said:

What kind of sorcery is this!! Wow!

It's all in there, you just need to tease it out.

When I started back into AP in mid-2020, I was processing my images in PS CS3 because I had it on my PC already. Then I started looking for something better suited to AP and got the triallicence for PI, watched a load of tutorials and started the learning curve. I was really surprised at the results I got quite early on, when re-processing images I'd done previously in PS, so I bought the full version of PI. 

Sure it can be a difficult program to learn and I'm no where near compitant with it yet, there's so much too it, but there's good tutorials and plenty of people to help on here. The route I took was to learn the processing side of the software first, it's only been the last few months that I've started using the stacking side.

I know it's also not the cheapest software for AP but I figured that, if I'm spending the best part of £2k on a mount and camera, the least I can do is get software made to produce the best quality results from my images.

Everyone has thier favourate processing software, mine's PI, but as long as you're happy with the results you can produce then I think that's what counts.