Jump to content

Banner.jpg.b89429c566825f6ab32bcafbada449c9.jpg

Help with these gradients


Recommended Posts

52 minutes ago, Allinthehead said:

I dither very 5 frames too, seems to do the trick. If i were you i'd aim for at least 5 hrs per filter especially Oiii and Sii as i've found these to be noisy with the Asi 1600. 

I downloaded your jpg above and used SCNR green and DBE on it and got this. I reckon your attempt is actually over stretched. Try to hold back to about 60% in each channel and see what you end up with, less is more sort of deal.

souls_DBE.thumb.jpg.cdc0cf10f2d61bc2a8a891803073e9c8.jpg

I guess doing anything with the jpg image is useless because I've obviously done too much wrong with it. But you my friend have raised the bar super high with what you were able to do with the image. Absolutely gobsmacked. 

I totally understand the need for more subs which is what I'll aim to do come this winter but because we had shorter nights and hardly any astro darkness, I had to complete while we had a clear spell. 

I normally apply STF to the image and then use the STF to the Histogram and apply what I'm given. Should I stretch the image manually you reckon? 

Also not entirely sure what you mean by hold back 60% on each channel? 

Link to comment
Share on other sites

  • Replies 46
  • Created
  • Last Reply
8 minutes ago, Allinthehead said:

Yes you should learn to stretch manually. That's what i mean by holding back to about 60% maybe even 70% of the Stf stretch. The gradients are in the background and can be hidden better by stretching a little less.

Cheers mate, makes more sense now. I'll definitely try to give it a go with that tonight. 

Have you tried to play with the xsif data in PI by any chance? 

I'm not sure if you can tell but I so want to see if anyone can take out these gradients in PI. 

Link to comment
Share on other sites

On 11/07/2018 at 19:05, souls33k3r said:

Just a bit of a background to this image. I ran DBE on all individual channels, then star aligned, ran a bit of noise reduction and combined them. Used STF, used the stf value to HistrogramTransformation Tool and applied. Used ColourMask script for Magenta (to reduce purple stars), Green to tone down the Green a bit and then Blue to up the blue a bit. 

Since narrowband has no relation whatsoever to "natural" colours, and the exposures may be completely different. There's no need to combine in the linear stage. Rather, stretch the narrowband masters, then combine using RGB-combination or PixelMath

 

5 hours ago, souls33k3r said:

Only darks. I've heard bias don't work or are not recommended  (one or the other)

Bias is not needed if you use dark flats for calibrating your flats. Flats are ALWAYS needed imo.

I'll have a go at the files.

 

edit:

First impressions, the gradients are partly caused by flat fielding. Normally that's no match for PixInsight. But since there is such a large variation in overlap between the subs, the gradients mix, and are more difficult to remove. The only sure way to avoid this in the future, is to use flats.

Link to comment
Share on other sites

In PixInsight:

Crop on 3 masters

DBE (first division, then subtraction for Oiii)

Histogram stretch

RGB combination with SHO palette

Contrast enhancements

No noise reduction, but Oiii would have needed it.

(Oiii and Sii are generally much weaker than Ha, and need a lot more data to bring the noise under control.)

Image22.thumb.jpg.b37d477ca57852b807f97cf7c3fbc6f8.jpg

(click on image to enlarge)

Link to comment
Share on other sites

3 hours ago, souls33k3r said:

because we had shorter nights and hardly any astro darkness, I had to complete while we had a clear spell. 

Ha and Siii are red colours and can be taken when the sky is blue, ie no astrodarkness or with moonlight. But Oiii is blue/green, and is a no go when you lack true darkness.

Compare to this daytime image that was taken with a deep red filter on broad daylight (blue sky). The red filter dramatically darkened the sky (further enhanced during processing). This would have been a completely different image if taken with a blue filter (such as Oiii).

http://anseladams.com/celebrating-90th-anniversary-ansel-adams-making-monolith-face-half-dome/

Link to comment
Share on other sites

16 hours ago, souls33k3r said:

Unfortunately it is too late but then i've always avoided flats. I only have an iPad which i've tried it in the past but as soon as i place the iPad over the top of our WO Star 71, the white torch light application goes all wonky even after putting 3/4 sheets of thick white paper over it. 

Is there a way to deal with this issue without flats? :(

You can take BIAS frames at any time! and re-stack! 

I use 250 frames now, will add a bit later (the more you use, the better - PI even has SuperBias tool, but I have not noticed any change in the end result, so I do not use it)
the only parameter which has to be the same - Binning, the rest do not have any impact as exposures are very short.

If you kept camera on the scope, you can do flats later also, - it does not need to be in superb focus.

 

also, please look through this post (noise reduction before blending) https://pixinsight.com/forum/index.php?topic=3184.0

 

Link to comment
Share on other sites

10 hours ago, wimvb said:

In PixInsight:

Crop on 3 masters

DBE (first division, then subtraction for Oiii)

Histogram stretch

RGB combination with SHO palette

Contrast enhancements

No noise reduction, but Oiii would have needed it.

(Oiii and Sii are generally much weaker than Ha, and need a lot more data to bring the noise under control.)

1

Hi wimvb, can you please explain why you Divided OIII with DBE first? Is the way you tried slightly to remove vignetting (instead of Flats)?

Link to comment
Share on other sites

12 hours ago, wimvb said:

Since narrowband has no relation whatsoever to "natural" colours, and the exposures may be completely different. There's no need to combine in the linear stage. Rather, stretch the narrowband masters, then combine using RGB-combination or PixelMath

 

Bias is not needed if you use dark flats for calibrating your flats. Flats are ALWAYS needed imo.

I'll have a go at the files.

 

edit:

First impressions, the gradients are partly caused by flat fielding. Normally that's no match for PixInsight. But since there is such a large variation in overlap between the subs, the gradients mix, and are more difficult to remove. The only sure way to avoid this in the future, is to use flats.

@wimvb this is a really important information. I've always combined the image after a little bit of processing before combine and then stretched the combined image. Will try to follow the rule of stretching the NB masters first and then combine.

 I've generally avoided using any filter other than Ha when the moon is up but handy to know that i can still go with SII. OIII i knew about but i took the chance as my target was completely in the opposite direction of the moon so thought i'd try my luck. Never again

Flats is something i'm starting to think i need to work on now. Need to figure out the best way to do them.

Your go at the image is very very good and i really like it, really do appreciate you taking your time and have a go at it. I am somewhat following the same workflow but like i said i will try the manual stretch of the master channels first and then combine.

Like @RolandKol i would also want to know why use divide first? Also i've only ever used subtraction only.

Last night i had a few minutes to play around with the data so followed @Allintheheadsuggestion of manually stretching so my slight different approach to this was:

- Ran LinearFit on SII and OIII using HA as the master frame

- Combine the three channels using PixelMath

- Manually stretched the channels

- Background Neutralization & Colour Combination.

- Ran slight noise reduction

- Used ColourMask to manipulate the colours a little

- Sharpened the image a little

Now will try to use Ha as my L data to bring out more nebulosity out of the image. (not sure if there's any other way i can bring out the nebulosity because it's seems to be hiding in there)

945540147_mysoul.thumb.jpg.cf1f48391a3c138bf8bf11e8dca40fa4.jpg

 

 

Link to comment
Share on other sites

Wonderful result. A massive improvement on your original image. You should always try to let the data guide you and your data is limited due to conditions and integration time so if you need to stretch less then so be it. Flats are a must for a decent image. A flats panel might be a worthwhile investment. I can't recommend one but i use an led panel with good results. 30 euro at my local electrical wholesalers. I don't think the experts recommend them however.

Link to comment
Share on other sites

7 minutes ago, Allinthehead said:

Wonderful result. A massive improvement on your original image. You should always try to let the data guide you and your data is limited due to conditions and integration time so if you need to stretch less then so be it. Flats are a must for a decent image. A flats panel might be a worthwhile investment. I can't recommend one but i use an led panel with good results. 30 euro at my local electrical wholesalers. I don't think the experts recommend them however.

Cheers mate, muchas gracias. I really can't thank you enough for your valuable input and giving me the suggestion of stretching it manually.

I will have to give it another go from the start using @wimvb suggestion of stretching each channel separately and then combine because like i said on the above image i combined first and then stretched it.

I do have a flats panel (It's one of those ceiling LED panels, i believe it's the same one you say experts don't recommend) but it's super big and heavy so can't really put it on the top of my scope but it's dimmable. What i can put on the top of my scope is my ipad so will have to work out the best way to do it.

How do you think i can bring out nebulosity in this image? because right now it's so dim. These are new waters for me so learning the ropes from the start. Use Ha over it?

Link to comment
Share on other sites

21 minutes ago, souls33k3r said:

I do have a flats panel (It's one of those ceiling LED panels, i believe it's the same one you say experts don't recommend) but it's super big and heavy so can't really put it on the top of my scope but it's dimmable. What i can put on the top of my scope is my ipad so will have to work out the best way to do it.

The cheap and simple solution for FLATS 

https://ebay.co.uk/itm/A4-LED-Stencil-Board-Light-Box-Artist-Thin-Art-Draw-Tracing-Copy-Plate-W-USB-UK-/152587835143?hash=item2386f19707

Just do not through the box away, - cut the hole for your scope in it and use it as a light box all together.
I placed 2 or 3 white paper sheets inside the box, as the Stencil Board has a pattern.

It can be dimmed and USB powered, - nice toy for £13.00! :)

Link to comment
Share on other sites

18 minutes ago, RolandKol said:

The cheap and simple solution for FLATS 

https://ebay.co.uk/itm/A4-LED-Stencil-Board-Light-Box-Artist-Thin-Art-Draw-Tracing-Copy-Plate-W-USB-UK-/152587835143?hash=item2386f19707

Just do not through the box away, - cut the hole for your scope in it and use it as a light box all together.
I placed 2 or 3 white paper sheets inside the box, as the Stencil Board has a pattern.

It can be dimmed and USB powered, - nice toy for £13.00! :)

Cheers mate, like i said i already have a panel so will use of that. Just need to work out the logistics :)

I guess this is a discussion for another topic :)

Right now i'm looking in to making this image better from where i am right now

Link to comment
Share on other sites

21 minutes ago, souls33k3r said:

Cheers mate, like i said i already have a panel so will use of that. Just need to work out the logistics :)

IKEA to the resque. 30x30 cm FLOALT on its dimmest setting. Four layers of fabric in an embroidery ring, courtesy of my wife.

Laugh all you want, it works.

flatsbywim.jpg.c026c724da2a2cb26227a5776c03685f.jpg

Link to comment
Share on other sites

3 minutes ago, wimvb said:

IKEA to the resque. 30x30 cm FLOALT on its dimmest setting. Four layers of fabric in an embroidery ring, courtesy of my wife.

Laugh all you want, it works.

flatsbywim.jpg.c026c724da2a2cb26227a5776c03685f.jpg

Don't worry mate, i'm not laughing at what you're doing because on my recent trip back home i bought myself 4 of these embroidery rings to stretch the white t-shirt ... i thought i was the only one in the world to think like that :D I probably won't need 4 layers of white t-shirt because the LED i have dims quite a bit. 

Link to comment
Share on other sites

DBE is all about sample placement.

Here's mine

1345821013_Skrmklipp2018-07-1301_10_32.png.c6c5ecaf710fcdc96ce3c57163c0e79d.png

This placement avoided much of the gradients in Oiii, but caught the vignetting. Vignetting needs to be corrected by division, since it replaces the flats in the calibration process.

On Oiii I then used this

1946841007_Skrmklipp2018-07-1301_13_41.png.96d73eca854cb3c6fb99f0dc78acfa12.png

More samples caught the rest of the gradients. Since this no longer is from vignetting, but most likely from light pollution (moonlight, summer nights), I used subtraction. Note that I use large samples (about 35 pixels), to get better statistics.

Link to comment
Share on other sites

2 minutes ago, wimvb said:

DBE is all about sample placement.

Here's mine

1345821013_Skrmklipp2018-07-1301_10_32.png.c6c5ecaf710fcdc96ce3c57163c0e79d.png

This placement avoided much of the gradients in Oiii, but caught the vignetting. Vignetting needs to be corrected by division, since it replaces the flats in the calibration process.

On Oiii I then used this

1946841007_Skrmklipp2018-07-1301_13_41.png.96d73eca854cb3c6fb99f0dc78acfa12.png

More samples caught the rest of the gradients. Since this no longer is from vignetting, but most likely from light pollution (moonlight, summer nights), I used subtraction. Note that I use large samples (about 35 pixels), to get better statistics.

Awesome stuff mate. Cheers for the explanation. 

You guys have all been nothing short of super help. Massive props to you guys.

One small question, since i have a manually stretched image which has done some nice gradients hiding, how can i bring out the nebulosity using my produced image? Use Ha and a L channel, apply some curves and HT to bring out more nebulosity and then apply it on the RGB combined image?

Link to comment
Share on other sites

2 hours ago, souls33k3r said:

Now will try to use Ha as my L data to bring out more nebulosity out of the image. (not sure if there's any other way i can bring out the nebulosity because it's seems to be hiding in there)

You can process each master (Ha, Oiii, Sii) separately, including noise reduction, local histogram enhancement, star reduction, etc, etc. Then combine at the very end, and tweak. The Ha master can take a lot more beating than the Oiii. I suggest you do noise reduction on Oiii (and maybe Sii), but not so much on Ha. But you can try doing deconvolution on Ha in the linear stage, and HDR transformation in the non-linear stage.

Link to comment
Share on other sites

3 minutes ago, wimvb said:

You can process each master (Ha, Oiii, Sii) separately, including noise reduction, local histogram enhancement, star reduction, etc, etc. Then combine at the very end, and tweak. The Ha master can take a lot more beating than the Oiii. I suggest you do noise reduction on Oiii (and maybe Sii), but not so much on Ha. But you can try doing deconvolution on Ha in the linear stage, and HDR transformation in the non-linear stage.

I'm not sure i follow you mate, how is this going to bring out nebulosity on the image that i have already got? (see below)1162364966_mysoul.thumb.jpg.cd2e06e154d01506917cc0bd8ebece76.jpg

 

Link to comment
Share on other sites

2 hours ago, souls33k3r said:

handy to know that i can still go with SII.

Sii is red, which means it has a good contrast against a moonlit sky. Unfortunately, the signal can be very weak. Depending on your filter (bandwidth, 3nm, 6 nm, 12 nm) you could get widely varying results. This is because a wider filter won't block light pollution as efficiently as a narrow filter. If you have a strong Sii target, imaging with the moon about can work, but if you have a weak Sii target, it may not.

Either way, I think you also need to increase your single sub exposure time for Sii and Oiii from 5 minutes to maybe 10 or 15. If that's not possible, double or triple the number of subs. This will get the noise down. If your integration time for Oiii and Sii is the same as for Ha, you end up with a very clean Ha master and very noisy Oiii and Sii masters. Ideally you should have about the same amount of noise in all three masters, because the noisiest channel sets the image quality.

Link to comment
Share on other sites

5 hours ago, souls33k3r said:

I'm not sure i follow you mate, how is this going to bring out nebulosity on the image that i have already got?

 

It doesn't. You'd start fresh from the 3 masters.

If you've saved your work as a project, back the process for each master all the way to the linear state. Create a clone which you move to a new workspace. Then restore the originals to their final state. Do that for all 3 masters. Then reprocess.

Reworking an image this way, lets you reuse process steps and masks, while trying new approaches.

Link to comment
Share on other sites

Archived

This topic is now archived and is closed to further replies.

  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...

Important Information

We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue. By using this site, you agree to our Terms of Use.