how can I gradient-remove this ?

[glowingturnip]

So I'm starting work on processing my take on the Elephant's Trunk and have immediately run into issues about how to gradient remove it.

Ordinarily, I'd use DBE in Pixinsight and carefully choose sample points that are definitely clear of any nebulosity or stars. However, in this case, it's all nebulosity !

The top photo is mine - raw with nothing done to it at all apart from an autostretch so you can see it.  That's the full DSLR frame, and has been calibrated with 30 flats.  There is still some gradient there, especially to the bottom and left.

The photo underneath is one I've borrowed from some helpful person on the internet (if it's yours, thank you ,and don't worry I won't be using any of your data).  It's been star-aligned to correspond to mine (it's a small segment of a widefield).  As you can clearly see, it's all nebulosity, and there's actually no safe place on mine that I can put a DBE sample point.

Any bright ideas for alternative ways I can do a gradient removal ?

Cheers

[Leveye]

Gradient Xterminator always works for me.

[Macavity]

I'm often troubled by gradients.  Seriously though, video astronomy

integrating cameras produce quite a lot of "amp glow" or whatever...

With no access to Photoshop, I often play with the "clunky but fun"

IRIS software. You could create an (semi) artificial "mask" or smooth

something? Maybe use "Straton" (qv) to separate out the stars and

nebula prior to processing? Of course none of this may be useful!

[Stub Mandrel]

This is what I do in photo paint.

work in 16-bit mode as you will roughly halve the dynamic range of your picture by doing this, in 16 bit it won't notice.

Stretch the image without clipping it.

study the gradient - is is radiating from a point or linear?

edit the fill pattern to be a 'negative' version of the gradient, usually going from black to white. remember you can use edge pad and mid-point to tweak the pattern.

Create a filled rectangle rather larger than your image and paste it as an overlying object

Set transparency to less than 50% and move around and resize the rectangle until it roughly cancels the gradient.

Play with transparency until the shades at darkest and lightest points on the gradient look similar. You should make aim to use a transparency rather less than 50% to minimise loss of dynamic range.

Don't worry if everything looks a bit washed out, but do worry about any light or dark intermediate bands - if they appear go back and change the fill a bit.

Run jaggy despeckle a few time on the rectangle - this is the only blur tool that works in 16 bit and it helps obscure any steps in the gradient.

Combine the rectangle with the background image.

Do a histogram stretch to bring back the dymnamic range.

I can't see a way of doing this in other packages, but it must be possible.

I found a pretty grotty 8-bit image of M33 and used this approach on it:

Before (stretched ready for gradient removal):

After gradient removal and a slight hue rebalancing (still pretty crummy, but you get the point):

[glowingturnip]

trouble is though that is that it's all nebula, so both DBE and gradient xterminator are going to try and flatten the nebula, which isn't what I want.

Will have a look at IRIS, not something I've played with before.

Stub, don't think your post came over properly ?

[Stub Mandrel]

I've edited it. The principle is to create a 'negative' of the gradient as a layer and then blend that with the original.

[StargeezerTim]

trouble is though that is that it's all nebula, so both DBE and gradient xterminator are going to try and flatten the nebula, which isn't what I want.

Will have a look at IRIS, not something I've played with before.

Stub, don't think your post came over properly ?

Stuart,

I've been having a play myself today trying to get rid of a gradient in an image. I tried DBE and Noels actions. Have a look at this thread for the images.

http://stargazerslounge.com/topic/255111-crescent-nebula-reprocessed/

DBE is the method that left most of the nebulosity intact but it it did 'flatten' it to an extent. I've also found that it works best after you have stretched the image. When I tried on the autosave output, it removed virtually all of the useful nebulosity from the image. It would be good to hear of a method that preserves the good stuff!

Tim.

[pipnina]

Perhaps you could work on making it a mosaic?

[Pompey Monkey]

I had a very similar picture to yours of the trunk. Have you tried ABE instead of DBE?#

Oh, and keep it 32 bit to avoid posterisation. 

[glowingturnip]

thanks all - a few things to try there, and I have a sneaky idea I'm trying too, will report back with progress

[MattJenko]

If it is all nebulosity, how are you going to know if there is a gradient or not anyway? Wouldn't it be better to use flats/bias and leave out DBE altogether unless there is a really obvious gradient?

Also, in PI, you could colour channel split and see if there is something that could be done in individual channels and then apply that to the L channel making some assumptions about the gradient being the same across all wavelengths.

Matt

[glowingturnip]

If it is all nebulosity, how are you going to know if there is a gradient or not anyway? Wouldn't it be better to use flats/bias and leave out DBE altogether unless there is a really obvious gradient?

Also, in PI, you could colour channel split and see if there is something that could be done in individual channels and then apply that to the L channel making some assumptions about the gradient being the same across all wavelengths.

Matt

hmm, that's not a bad idea actually.  It is all nebulosity across the field, and I'm pretty sure that I have a gradient, darker towards the bottom-left corner even though I've used flats & bias, but yes, I'll have a play with trying a DBE on just the green and/or blue channels, and see if I can come up with a dummy gradient from those to apply to my red, or my lum - cheers, let's see how I get on

[glowingturnip]

is there much OIII nebulosity in this target that I might be seeing in my G & B channels ?

[glowingturnip]

I had a very similar picture to yours of the trunk. Have you tried ABE instead of DBE?#

Oh, and keep it 32 bit to avoid posterisation. 

Pompey Monkey wins !!

ABE absolutely nailed it, almost from the default settings.  Here's before, gradient, and after with a deviation of 0.65.

Well happy, now to get on with processing the rest of it !