Background mottled...Need more data?

[Star101]

Long story short...I compared my image with one that managed IOTD on Astrobin. Decided the IOTD had almost perfect smooth background with its 11 hours of data. While mine has about 2 !

Would more data improve the background or is there some technique I could use?

 

Thanks in advance.

 

Dave.

[pete_l]

A cooled camera is a definite advantage. Necessary but not sufficient is a deft hand at processing (you need the basic raw material to be good)

[vlaiv]

There are couple of things that can create this sort of background.

First is way the image was stacked - or rather, algorithm used to align subs before stacking. Any sort of alignment of frames requires some sort of interpolation algorithm to be used (actual pixel values in aligned frame fall in between pixels in original frame due to sub pixel shifts and rotation). Different interpolation algorithms have different properties. Some of them cause unacceptable artifacts and are avoided (like nearest neighbor interpolation), some are straight forward to do but cause "pixel blur" (like bilinear interpolation - which is same as bin x2 under some cases). Often used one is bicubic interpolation - that one smooths data too much and noise that should be "pixel" level noise is spread over couple of adjacent pixels and noise grain increases in size. Best to use is Lanczos interpolation as it preserves much of noise signature.

Second thing that can create background like that is excessive use of denoising algorithms based on similar smoothing of data like bicubic interpolation. Selective gaussian filtering and such tend to create such large grained background. Whenever denoising algorithm fails to detect background star it will make it into a bit brighter "blob" rather than smooth it out.

Third thing that can cause such background is using scope that is not diffraction limited, or oversampling a lot on star field that is dense. In this case, even if you have very good SNR star image will be "smeared" over larger number of pixels. If you oversample a lot (like in poor seeing) - most background stars will look like that, instead of being small concentrated light points and background will contain large grain structure that will be composed out of blurred stars rather than blurred noise (as in previous two cases).

To get nice "high fidelity" background (I actually don't mind a bit of noise in background as long as noise looks right - pixel to pixel variation and very small variation in intensity - small scale grain) - you need to avoid things that alter background.

Use good alignment resampling algorithm - Lanczos 3/4. Don't use aggressive denoising that will blur things out. Try to match sampling rate to scope / seeing as best as you can. And, yes, finally under these circumstances more data will certainly make background look smoother and better.

More data will not make large scale grain go away if it is caused by things that I have listed.

[Star101]

Sorry, I should have said this was taken with the ZWO 183MM Pro cooled to -30C. 15 x 300s Ha and 12 x 300s OIII. Gain 200.

 

[vlaiv]

Just now, Star101 said:

Sorry, I should have said this was taken with the ZWO 183MM Pro cooled to -30C. 15 x 300s Ha and 12 x 300s OIII. Gain 200.

 

What scope was it used on?

[Star101]

Thank you vlaiv for such a detailed explanation. I will definitely check my processing settings.

 

TS Optics TS 65/420 Quadruplet scope

Image was processed in Pixinsight

Full image here

 

Dave

Edited July 24, 2019 by Star101

[vlaiv]

8 minutes ago, Star101 said:

TS Optics TS 65/420 Quadruplet scope

 

Image was processed in Pixinsight

Full image here

 

Dave

Could be due to some of the issues I mentioned - image looks like it was denoised a bit?

At 1.18"/px it is probably a bit oversampled as well. In general that is not considered resolution that is oversampling "by default" - up to 1"/px can be "pulled off", but you need aperture for that. With only 65mm star FWHM is going to be at least 3" - 3.5". That is more in 2"/px territory. Do you happen to have stats on FWHM in your subs?

[wimvb]

7 hours ago, Star101 said:

Image was processed in Pixinsight

The noise in your image consists mostly of colour variations = chroma noise. Try mmt noise reduction on chrominance only. Use up to 8 layers and a mask that will protect the main target. This should give you a neutral background without large scale colour mottle/noise.

Btw, in my experience, cooled cmos gives better results if you use a large number of subs. Never mind the "rule of dimishing returns".

Good luck

Edited July 24, 2019 by wimvb

[ollypenrice]

More data would help considerably in my view. Did you dither guide? This would also help.

When I feel the need to process my background sky I try to avoid ready made noise reduction routines because they all work by blurring, meaning pixel to pixel communication. This gives the oily 'vaseline on the lens' look. Instead, in Photoshop, I prefer to zoom in to pixel scale to see what is going on and then select 'families' of pixels for adjustment using the colour select tool. I try to allow each pixel to retain some of its abnormality so that the sky still looks natural. In this case I lowered the colour saturation on most of the background pixels and then picked a middle-brightness background pixel and placed it on the graph in Curves. I then flattened the curve somewhat just above and below it. Even working on a JPEG screen grab this worked reasonably well. It's over-done here but it shows you the general principle. By using Photoshop layers you can choose how much of the modification to apply to the final image. If doing it for real I'd take more time to protect the stars.

Olly

 

Edited July 25, 2019 by ollypenrice

[rubecula]

That's an amazing result Olly!  From your description I can understand what you did, but not how you did it. Oh, I've got so much to learn. 

[Paul73]

23 minutes ago, rubecula said:

Oh, I've got so much to learn. 

I know how you feel.....🤯🤯

[wimvb]

I played a bit with the image in PixInsight. But unlike Olly, who worked on the provided jpeg, I took the liberty of downloading the full version from Astrobin. I hope you don't mind.

As I wrote before, there is colour mottle in the background, but also a lightness (luminance) variation. I tackled both with MMT.

The image showed very little single pixel noise, so I used noise reduction only on structures ranging from 2 to 8 pixels in Luminance, and 2 to 16 pixels in Chrominance.

The result is not as smooth as Olly's. But the image you posted didn't include the main target, the Veil complex. So here's a section with nebulosity in it, processed the same way. When applying noise reduction it's important to consider the entire image, and make decissions based on structures (signal) you want to keep, and structures (noise) you want to reduce.

For completeness, here's the process container that you can load into PixInsight.

MMT_noise_reduction.xpsm

[Star101]

On 25/07/2019 at 07:12, ollypenrice said:

More data would help considerably in my view. Did you dither guide? This would also help.

When I feel the need to process my background sky I try to avoid ready made noise reduction routines because they all work by blurring, meaning pixel to pixel communication. This gives the oily 'vaseline on the lens' look. Instead, in Photoshop, I prefer to zoom in to pixel scale to see what is going on and then select 'families' of pixels for adjustment using the colour select tool. I try to allow each pixel to retain some of its abnormality so that the sky still looks natural. In this case I lowered the colour saturation on most of the background pixels and then picked a middle-brightness background pixel and placed it on the graph in Curves. I then flattened the curve somewhat just above and below it. Even working on a JPEG screen grab this worked reasonably well. It's over-done here but it shows you the general principle. By using Photoshop layers you can choose how much of the modification to apply to the final image. If doing it for real I'd take more time to protect the stars.

Olly

 

Thanks Olly,

 

That is amazing!

I will have a go at trying this method. 

 

No, I did not dither as i was imaging with two scopes and SGP and Artemis do not talk to one another. PHD2 therefore will dither while one of the cameras are still imaging. I did try a recently talked about free program that would wait for both cameras to finish before dithering but I ran into driver issues, so gave up on that. As I do not often use both cameras images, I will go back to dithering when only planning using one camera.

 

Thanks again.

 

Dave.

Edited July 27, 2019 by Star101

[newbie alert]

Dithering would help at the data collecting point...

[ollypenrice]

14 minutes ago, Star101 said:

Thanks Olly,

 

That is amazing!

I will have a go at trying this method. 

 

No, I did not dither as i was imaging with two scopes and SGP and Artemis do not talk to one another. PHD2 therefore will dither while one of the cameras are still imaging. I did try a recently talked about free program that would wait for both cameras to finish before dithering but I ran into driver issues, so gave up on that. As I do not often use both cameras images, I will go back to dithering when only planning using one camera.

 

Thanks again.

 

Dave.

We run a dual rig as well so I know the problem.

Olly

[fireballxl5]

20 hours ago, wimvb said:

For completeness, here's the process container that you can load into PixInsight.

Nice, just tried this on my current project and it works really well, even with no mask applied. I need to learn more about masking in PI 🤔

Thank you! 

CS, Andy

PS, before and after applying Wim's MMT process...

Before:  After: 

Edited July 27, 2019 by fireballxl5
added images

[geordie85]

On 25/07/2019 at 07:12, ollypenrice said:

More data would help considerably in my view. Did you dither guide? This would also help.

When I feel the need to process my background sky I try to avoid ready made noise reduction routines because they all work by blurring, meaning pixel to pixel communication. This gives the oily 'vaseline on the lens' look. Instead, in Photoshop, I prefer to zoom in to pixel scale to see what is going on and then select 'families' of pixels for adjustment using the colour select tool. I try to allow each pixel to retain some of its abnormality so that the sky still looks natural. In this case I lowered the colour saturation on most of the background pixels and then picked a middle-brightness background pixel and placed it on the graph in Curves. I then flattened the curve somewhat just above and below it. Even working on a JPEG screen grab this worked reasonably well. It's over-done here but it shows you the general principle. By using Photoshop layers you can choose how much of the modification to apply to the final image. If doing it for real I'd take more time to protect the stars.

Olly

 

That's an amazing result. 

Would you mind posting a few screen shots of the process please? I find it easier to learn things when I can see them. 

[Starflyer]

Following, I always struggle with blobby noisy backgrounds.