Post processing

[Jm1973]

13 hours ago, wimvb said:

Light pollution is a likely candidate. If you get the histogram so far to the right with only 30 seconds exposures on an unmodded dslr, you probably have a fair amount of light pollution. If this is the case, consider investing in a light pollution filter. To compensate for the added noise, you will need long total integration times, ie lots and lots of data.

Hi. Thanks for the comments. LP filters seem to vary in price from a few quid up to over a hundred. 

How much would I need to spend to get something that would help. Are there any particular brands worth buying? I'm trying not to break the bank.

[Jm1973]

9 minutes ago, wimvb said:

Before you stack the sub frames, you need to calibrate them. For the setup you used to take this image, good flats are essential. This is what the gradient in your image looks like. The circular structure is vignetting. But there is also a colour gradient with a red band at the top. Most software can handle one source of an uneven background quite well, but a combined gradient can be a challenge to remove. Especially if you also have different gradients in the three colour channels, which makes it even more of a challenge to remove.

Here's what I managed to pull out of your data, before the remains of gradients started to pop up. Nothing too fancy. Arcsinh stretch (available in PS) lifted the colour in your image. after that, curves transformation as Olly already wrote about.

(click on the images to see a larger version)

Thanks for the reply. A lot of what you are saying with gradients kind of goes over my head a bit, but you are not the first person to tell me I should be using more flats.

I only used 10 in this image. The next time I get a clear night I will take more. 

You have managed to get my data looking pretty amazing, as far as I am concerned. I would be very happy if I could get to that stage.

I'll have a look at arcsinh stretch. Assuming it is on my rather old version of PS.

[Jm1973]

So I have got a little further with it now. I am bringing out more detail and more colour, but unfortunately I am also blowing out the core and background.

[wimvb]

9 hours ago, Jm1973 said:

Thanks for the reply. A lot of what you are saying with gradients kind of goes over my head a bit, but you are not the first person to tell me I should be using more flats.

I only used 10 in this image.

For some reason, the flats aren’t doing their job. Even with only one flat, vignetting should be greatly reduced, albeit at the cost of increased noise.
 

 

10 hours ago, Jm1973 said:

How much would I need to spend to get something that would help. Are there any particular brands worth buying? I'm trying not to break the bank.

Quality costs. I used Baader a uhc filter before I moved house to my current location, which is much darker than where I lived before. IDAS filters are considered efficient, and colour preserving. But obviously at a price. The very cheapest may also be the lowest in quality, but such an argument doesn’t always hold. A poor filter can give you unwanted reflections, and you will need to do some investigating begore you buy.

[ollypenrice]

I didn't address gradients in my first demo, I was focusing only on the stretch process, but Wim is quite right that they do need a fix. A gradient is simply a gradual drift in brightness across the image. If this drift is strongest in one colour (as it usually is) this will appear as a 'colour gradient.'  Where they come from is anybody's guess. Even at one of the darkest sites in mainland Europe I still get them on 99% of the images I take. (I'd have said 'on all of them' but I recently shot some gradient free colour data!)

Flats won't remove gradients caused by skyglow. As far as the camera is concerned any light going onto the chip is real signal - as indeed it is, even if you don't want it. Flats only remove the effects of uneven illumination caused by the optics, so vignetting and dust bunnies. Gradient removal is done in software by well engineered algorithms. Here are some pointers:

Pixinsight's Dynamic Background Extraction. Stunningly effective. With the image still linear but with a temporary visualization of its stretched appearance on screen you place markers on what is genuinely background sky to tell the software that those places should be at the same brightness in every channel. It makes a map of the gradients it finds and you then subtract the map from the image. A nice by-product is that it also performs a fine colour calibration in the process. Don't use DBE to fix local artifacts like dust bunnies. It won't. Use a small number of markers to find the broad underlying gradient.

Russ Croman's Gradient Xterminator. A reasonably priced Ps plug-in which is pretty good. There are videos online about how to get the best out of it.

AstroArt has a gradient removal tool and is a top quality stacking and calibrating program, far better than DSS, for instance.

Astro Pixel Processor is highly recommended by several friends and clients. I'm certain its gradient tools work because I just processed a 32 panel mosaic from individual subs made in APP. They were admirably flat.

Core: just make a second stretch, almost as hard as the first, but concentrating on the core. Avoid getting that 'stellar' central part and put it as a layer under your present version and use a feathered eraser to remove the saturated bit on top.

Olly

[wimvb]

1 hour ago, ollypenrice said:

Where they come from is anybody's guess. Even at one of the darkest sites in mainland Europe I still get them on 99% of the images I take.

Natural skyglow is one origin. Aurora Borealis may be jaw dropping on rare occasions, but mostly it’s a nuisance. The physical process that creates the aurora is always at work, it’s just too weak to be visible by the naked eye, most of the time. I’ve seen its effect through the eye piece of my scope on a dark and cold winter night, when it was invisible otherwise. I don’t want it on my images, though.
Normally, I apply DBE after RGB combination, because the gradients I have are easy to remove. But for combined gradients, such as vignetting plus sky glow or light pollution, where the resulting gradient can be complex, it is at times necessary to do DBE on the individual masters before combining them. Also, vignetting has to be divided out, as opposed to additive glow gradients. For OSC images, it is sometimes  necessary to split the channels, and apply DBE to each channel.

1 hour ago, ollypenrice said:

A nice by-product is that it also performs a fine colour calibration in the process.

Btw, I don’t use DBE to neutralise the background in my images. I prefer the background neutralization process in PI for that. The math may very well be the same, I just prefer to be in control of this step.

Edited September 30, 2020 by wimvb

[City9Town0]

+1 for Pixinsight DBE and Background Neutralisation, although I found that DBE does take some time to get to grips with.

OFF TOPIC. wimvb... interesting perspective on the aurora.

Having travelled into the Arctic Circle four times for the delights of the aurora I'd never considered the inconvenience in imaging! For me those rare visits are significant moments of my lifetime.

It has once shown up in my images of the Northern horizon at 52 North. Thanks.

[Crackabarrel]

4 hours ago, wimvb said:

Quality costs. I used Baader a uhc filter before I moved house to my current location, which is much darker than where I lived before. IDAS filters are considered efficient, and colour preserving. But obviously at a price. The very cheapest may also be the lowest in quality, but such an argument doesn’t always hold. A poor filter can give you unwanted reflections, and you will need to do some investigating begore you buy.

+1 for the IDAS LPS filters, I have been using a D2 recently and it does a great job of dealing with light pollution (Imaging from Bortle 8 skies) and retaining balanced colours.

It is pricey though, another option would be the Optolong L-eNhance, although I have no personal experience with this one.

[ollypenrice]

1 hour ago, wimvb said:

 

Btw, I don’t use DBE to neutralise the background in my images. I prefer the background neutralization process in PI for that. The math may very well be the same, I just prefer to be in control of this step.

I guess my site means that I never need to use Background Neutralization. After DBE it's always neutral and star colour is good as well. I think it's better to get good flats than to use DBE against vignetting but it can be done. One problem is that we often frame the object (which is bright) in the middle of the image, meaning it's hard to give information to DBE about how the gradient increases towards the centre. (I think this is why Automatic Background Extraction is not so good. I find it often gives a dark ring around a central object like a large PN or galaxy because it has read the outer parts of the object as background and dimmed them when applied.)

Olly

[wimvb]

I never use ABE, just because it’s difficult to control. For some reason, my flats correct dustbunnies well enough, but don’t remove all vignetting, and I have to clean up with DBE. It works good enough if you put samples in each corner and along each side (=8), plus from each corner along the diagonal, at most half way in towards the centre (+4). No samples closer to the centre needed. You also don’t have to avoid stars, because the are excluded from the samples. The sample plot shows which pixels will be included in the calculation of the background model. Black means excluded, white means included. Gray is a measure for weight. As long as stars don’t have haloes, they will be excluded.

@Jm1973: sorry for derailing this thread. Such is life on SGL, it makes this forum so alive and interesting. Welcome to the club. 😀

 

Edited September 30, 2020 by wimvb

[Jm1973]

3 hours ago, wimvb said:

I never use ABE, just because it’s difficult to control. For some reason, my flats correct dustbunnies well enough, but don’t remove all vignetting, and I have to clean up with DBE. It works good enough if you put samples in each corner and along each side (=8), plus from each corner along the diagonal, at most half way in towards the centre (+4). No samples closer to the centre needed. You also don’t have to avoid stars, because the are excluded from the samples. The sample plot shows which pixels will be included in the calculation of the background model. Black means excluded, white means included. Gray is a measure for weight. As long as stars don’t have haloes, they will be excluded.

@Jm1973: sorry for derailing this thread. Such is life on SGL, it makes this forum so alive and interesting. Welcome to the club. 😀

 

No problem. It's all very interesting, though a bit above my pay grade so to speak.

[Jm1973]

7 hours ago, ollypenrice said:

I didn't address gradients in my first demo, I was focusing only on the stretch process, but Wim is quite right that they do need a fix. A gradient is simply a gradual drift in brightness across the image. If this drift is strongest in one colour (as it usually is) this will appear as a 'colour gradient.'  Where they come from is anybody's guess. Even at one of the darkest sites in mainland Europe I still get them on 99% of the images I take. (I'd have said 'on all of them' but I recently shot some gradient free colour data!)

Flats won't remove gradients caused by skyglow. As far as the camera is concerned any light going onto the chip is real signal - as indeed it is, even if you don't want it. Flats only remove the effects of uneven illumination caused by the optics, so vignetting and dust bunnies. Gradient removal is done in software by well engineered algorithms. Here are some pointers:

Pixinsight's Dynamic Background Extraction. Stunningly effective. With the image still linear but with a temporary visualization of its stretched appearance on screen you place markers on what is genuinely background sky to tell the software that those places should be at the same brightness in every channel. It makes a map of the gradients it finds and you then subtract the map from the image. A nice by-product is that it also performs a fine colour calibration in the process. Don't use DBE to fix local artifacts like dust bunnies. It won't. Use a small number of markers to find the broad underlying gradient.

Russ Croman's Gradient Xterminator. A reasonably priced Ps plug-in which is pretty good. There are videos online about how to get the best out of it.

AstroArt has a gradient removal tool and is a top quality stacking and calibrating program, far better than DSS, for instance.

Astro Pixel Processor is highly recommended by several friends and clients. I'm certain its gradient tools work because I just processed a 32 panel mosaic from individual subs made in APP. They were admirably flat.

Core: just make a second stretch, almost as hard as the first, but concentrating on the core. Avoid getting that 'stellar' central part and put it as a layer under your present version and use a feathered eraser to remove the saturated bit on top.

Olly

I might look at Xterminator, as I want to reach the limtis with Gimp and PS before exploring other software.. lest I get (even more) overwhelmed.

I will try your idea for fixing the core tonight. Thanks again!