6 replies to this topic

[Karlp295]

I wondered what techniques and tools you use to reduce light pollution in your astrophotography images? What is the best way to do this?

 

I've just made a video and compared several tools and methods icluding GraxPert, Siril and so on:

 

Go take a look and I'd love a comment and a question:

 

https://youtu.be/8KcEDORLCAc

 

I hope you find this helpful!

[SgrB2]

Use duo band filters and it'll do it for you.

[BQ Octantis]

Removing light pollution gradients is different than removing the light pollution signal. To remove the gradients, I use the Background Extraction tool and the Banding Reduction tool in Siril. I find it works very well for very wide angle shots in my Bortle 6 sky—like this 25° shot from my Nifty Fifty on my APS-C DSLR:

 

Starting point (click for full size)

 

RBF subtraction (click for full size)

 

Vertical banding removal (click for full size)

 

Horizontal banding removal (click for full size)

 

Indeed, I've found Siril works better for my RGB images than GraXpert, which leaves ugly residual color blobs in my RGB images.

 

https://www.cloudyni...xpert-problems/

 

Like you, I use the black point in a histogram tool to remove the light pollution signal. This is effectively signal subtraction while holding the white point constant, which results in a higher slope in the gain than the original capture.

 

The more interesting question is, at what point do you clip the blacks, and how far? In an sRGB-calibrated image, I find that the red histogram is significantly wider than the blue and green. If I used an automated black clip, it invariably chops off too much of the reds:

 

Histogram black point auto-clip (click for full size)

 

This robs me of shadow noise improvement when subsequently synthesizing a luminance from the three camera raw channels. So I generally manually clip to the start of the tail of the red for the luminance, followed by another clip on the synthetic luminance:

 

Manual black point (click for full size)

 

Luminance histogram black point auto-clip (click for full size)

 

I then do the auto histogram clip for the CIE (xy) chromaticity synthesis. This gives me a smoother image with more saturated colors.

 

BQ

Edited by BQ Octantis, 08 December 2024 - 07:56 AM.

[BQ Octantis]

I'm sure you're wondering how I get rid of that set of reflection rings that the darks and RBF didn't mitigate. This was actually my reason for trying GraXpert. But I've found nothing as effective as a soft brush in Photoshop set to add or subtract 1 pixel. A strong histogram stretch reveals the problem areas and allows for selecting the right size brush. It makes for a usable image at 100% scale out to the corners (star distortions not withstanding).

 

Coma Berenices Plucked from the Gradient (click for full size @ 30% scale)

 

Contrast that with a default processing of one of the subs through ACR:

 

(Click for full size)

 

Cheers,

 

BQ

Edited by BQ Octantis, 08 December 2024 - 10:07 AM.

[Tapio]

That ring could be the result of light leak/reflection.

[rj144]

Siril's BE per sub gets rid of almost all of it and I'm in Bortle 8/9.

[idclimber]

Thank you for your contribution. Here are my thoughts on this topic. 

 

First the horsehead image you spent most of your time on is not light pollution but vignetting from the optical train and a failure of the flats to correct it. Typical light pollution gradients are a lot more gradual across the image. Exactly like your first image of the galaxies. 

 

The images you pick are relatively easy to fix with modern software as there is a ton of background space. This gets a lot harder when the image is the North American nebula up close and 90% of the image is nebula. 

 

If is far better to address gradients early. This only addresses it in post after integration. A Sky gradient caused by light pollution will change and move as the night progresses. Each sub exposure will have a slightly different orientation. When the scope is pointed East it will be on one side. In the morning when it is pointed west it will be reversed. Each individual sub has a relatively simple gradient. Integrate them all together and the gradient is a mess the is much harder to remove. 

 

This can be handled by PixInsight in WBPP when you calibrate and integrate your data. This is Local Normalization. 

 

Step 1 is to review all your subs and identify the ones that have the least gradient that is the most uniform. Pick one to three frames and make note of them. 

 

Step 2 In WBPP select Interactive Mode in Local Normalization. Then select 1 to 3 frames in that tab and let this run. WBPP will then prompt you later to select the frames noted in step 1. Enter then and let it finish. the resulting integration should match the best frames you have and be much easier to work on from there.