The problem of green in narrowband SHO palette

[Bobinius]

Hi guys,

I was wondering what would be the correct and scientific way of dealing with green after combining the channels/masters in narrowband images using the SHO Hubble palette. Since we assign the Ha data to the green channel, when we use SCNR to decrease the green aren't we actually clipping Ha data? While this is normal when using an OSC camera with 2 green pixels for the Bayer matrix, it seems incorrect for narrowband.

The problem is that in order to obtain for example a correct blue colour that represents the Oyxgen, I am almost always obliged to decrease the green using SCNR. I generally try to use equal total exposition times for the 3 channels, but even after stretching the three channels in order to obtain equal median pixel values using Statistics, the green is still too strong and unbalanced. That is the case even when I have less total exposition for Ha! Or when I combine the images in Linear using linearfit. Astrophotographers using equal duration exposures still present the classical appearance of the nebula, with less green than 'normal'.

The question is, am I doing something wrong? I like Pixinsight especially for its scientific way of processing images, so I don't want to end up cosmetising things in order to obtain the result that I am looking for by clipping data and misrepresenting reality.

Thanx.

[dave_galera]

Hi Bob,

I never use SCNR to remove the green cast, I use colour masks and curves as per this video....

https://www.youtube.com/watch?v=RZZ4UlIQk0s

However, I do use SCNR -> Remove Green to sometimes get rid of pink stars.

Hope this helps.

[Bobinius]

Thanx, I'll check it out.

But why do you think that some green should be removed in the first place? (We may like more or less green, but since it is Ha, how do we decide ?) I feel it is like taking out green from an apple to make it look more yellow when it is actually green. I don't know of processing steps that take out red from a HOO palette (when it is the same Ha signal).

[dave_galera]

The green should not actually be there or is required, however, it is the way the different channels react.....its very complicated, Vincent Peris did explain it to me with lots of diagrams but it was 5 years ago and I have forgotten, basically you should not have any green in a narrowband image, unless, of course you want some.....that's all up to you.

To remove red use color masks then curves on the red channel.

[wvanreeven]

Correct me if I am wrong, but the SHO Hubble palette maps S to red, Ha to green and OIII to blue right? That's how the green ends up in the image. Since Ha usually is the strongest component of the three, these kind of images tend to have a green hue that can be removed in various ways with SCNR being one of them.


HTH, Wouter

[dave_galera]

Ooooops........You are right Wouter.
I remember now....it was pink stars that Vincent was explaining not green casts LOL

I do find using colour masks you have a lot more control

[Niall Saunders]

Any method you choose to create your final colour image from your NB data is perfectly valid.

After all, you didn't start with 'colour' in the first place.

You are free to use any processing steps you wish to get the image you desire - and there are no requirements to 'fill the Green channel' with data whatsoever.

Food for thought?

[Bobinius]

The green should not actually be there or is required, however, it is the way the different channels react.....its very complicated, Vincent Peris did explain it to me with lots of diagrams but it was 5 years ago and I have forgotten, basically you should not have any green in a narrowband image, unless, of course you want some.....that's all up to you.

To remove red use color masks then curves on the red channel.

Thanx for the video, very useful as I did not use this mask technique for processing. Probably the mask is more flexible than SCNR. SCNR is really fast for removing the magenta around the stars (due usually to the O3 filter bloating): you just invert the image and apply a SCNR at 1, that's it.

The idea with the red is that we usually do not decrease the reds for example in a HOO image (or do you?) It would be analogous to decreasing green in SHO. How do you know IF it should be removed or HOW MUCH to remove? (other than esthetic criteria, "I like it more this way" kind of justification).

[Bobinius]

Any method you choose to create your final colour image from your NB data is perfectly valid.

After all, you didn't start with 'colour' in the first place.

You are free to use any processing steps you wish to get the image you desire - and there are no requirements to 'fill the Green channel' with data whatsoever.

Food for thought?

h
I agree that the color assignment is arbitrary and the SHO palette is there mainly for contrast reasons. Once you assign Ha to green, the Ha data is Green ( I was talking specifically about the SHO palette). If I am trying to take a real picture of a celestial object, that's representative, am I not going to lose data and misrepresent that object if I clip the green data? Should this difference in intensity be compensated by longer total exposure for O3 and S2 for example?

[dave_galera]

As Niall indicated above as we didn't start with colour in the first place it is totally down to you how much red you remove, it all depends on what looks pleasing to you

[dave_galera]

I agree that the color assignment is arbitrary and the SHO palette is there mainly for contrast reasons. Once you assign Ha to green, the Ha data is Green ( I was talking specifically about the SHO palette). If I am trying to take a real picture of a celestial object, that's representative, am I not going to lose data and misrepresent that object if I clip the green data? Should this difference in intensity be compensated by longer total exposure for O3 and S2 for example?

You won't clip the green data, but if you think you are clipping just check the histogram before and after and that will show you if any clipping is taking place. Yes you normally take longer exposures of SII and OIII (if there is any there), or even bin them 2x2, because these are much weaker than Ha, also you can multiply OIII and SII by a factor (say OIII * 1.5 or OIII * 2 etc) when you combine the channels.

...and SHO is not a real picture from a colour perspective, it is totally false.

Experiment........

[Bobinius]

Dave,

Probably you are right with the histogram, I usually look what's going on and indeed there is a difference between the intensity transformation (the green pixel value is less but not zero) and a clip (which would totally delete the pixel).

For example, do you know if there is a standard protocol at Hubble for producing the images - I mean I hope that the M66 aspect that Hubble produces does not depend if Sandra from NASA who did the post-processing Saturday really loves pink hues or not. Hopefully there are some objective rules for treating the image.

[dave_galera]

Dave,

Probably you are right with the histogram, I usually look what's going on and indeed there is a difference between the intensity transformation (the green pixel value is less but not zero) and a clip (which would totally delete the pixel).

For example, do you know if there is a standard protocol at Hubble for producing the images - I mean I hope that the M66 aspect that Hubble produces does not depend if Sandra from NASA who did the post-processing Saturday really loves pink hues or not. Hopefully there are some objective rules for treating the image.

I think it is down to the processing team, i.e. Sandra LOL

[Geoff]

As Niall indicated above as we didn't start with colour in the first place it is totally down to you how much red you remove, it all depends on what looks pleasing to you

Well actually we do start with colour: Ha is a deep red, SII is a deeper red and OIII is blue-green (teal). Just an idle remark.

[dave_galera]

As Niall indicated above as we didn't start with colour in the first place it is totally down to you how much red you remove, it all depends on what looks pleasing to you

Well actually we do start with colour: Ha is a deep red, SII is a deeper red and OIII is blue-green (teal). Just an idle remark.

I think Niall was refering to the fact that its false colour not real colour