Colour considerations

[neil phillips]

My recent image of Jupiter. i adjusted colour by eye. I kept thinking maybe i got it wrong. (Because I did at first and adjusted it) So I took my image to registax to look at the RGB Graph. To my surprise it was almost bang on. I then asked registax to do an auto colour balance correction. 

The problem with registax is, if you ask it to do an auto colour balance on a image that is already fairly balanced. It doesn't appear to know what to do.

What the software designer should have done is told the program to do nothing. and repeat the colour that is being presented as accurate.

But i guess the software designer thought this might irritate people that registax was not giving a second choice.

So, it attempts to spit out a correction. even though its actually not correct at all and its own graphs know this lol.

As an example I post two pages one with my colour adjusted by eye. the other with registax showing its poor attempt at trying to correct something that was already correct. to, give the user another choice. You will notice when it does this red goes up from green. blue goes down from green. It effectively separate's the colours. Now with a red colour cast. 

my colour adjusted by eye

 

Registax incorrect second guess. Notice how the peaks up top are more split. But the bottom more correctly aligned. Which is true top or bottom part of the graph ?

 

Preference is one thing. And that's fine. Accuracy is another. Which is also fine

Edited October 14, 2022 by neil phillips

[ONIKKINEN]

Hmm, i think i prefer the second one actually. On my monitor the first one looks a bit too blue with the whites, but then again neither are what i see in an eyepiece so what is right?

Maybe im just used to seeing warmer tones from Jupiter, maybe its my eyes. Cracking image either way.

[neil phillips]

15 minutes ago, ONIKKINEN said:

Hmm, i think i prefer the second one actually. On my monitor the first one looks a bit too blue with the whites, but then again neither are what i see in an eyepiece so what is right?

Maybe im just used to seeing warmer tones from Jupiter, maybe its my eyes. Cracking image either way.

It also may depend on the monitor. But yes it does look more blue. But then again. The other image does look too red. I think the reason it may look a bit blue. Is colour saturation. i use a lot of saturation and it can do this to the image unfortunately. Dropping saturation a bit, does help 

Btw nobody likes white, everyone likes red

Also to be fair i said ALMOST bang on. Actually i think i did add just a smidge to much blue

 

Edited October 14, 2022 by neil phillips

[ONIKKINEN]

14 minutes ago, neil phillips said:

It also may depend on the monitor. But yes it does look more blue. But then again. The other image does look too red. I think the reason it may look a bit blue. Is colour saturation. i use a lot of saturation and it can do this to the image unfortunately. Dropping saturation a bit, does help 

Btw nobody likes white, everyone likes red

Also to be fair i said ALMOST bang on. Actually i think i did add just a smidge to much blue

 

This one is better than either of the first ones to my eyes. Doesnt jump out as too blue anymore, maybe just the saturation that jumped out to me? Small difference, but big difference, if you know what i mean.

[neil phillips]

Just now, ONIKKINEN said:

This one is better than either of the first ones to my eyes. Doesnt jump out as too blue anymore, maybe just the saturation that jumped out to me? Small difference, but big difference, if you know what i mean.

None of this is easy.  really should try and pay more attention. Also, what is confusing is the graph itself the peaks look aligned in my correction. but not in the lower registers. 

The auto corrected version its reversed the peaks look separated. but the lower frequencies look more aligned. 

[neil phillips]

I think the question is should we ignore the peaks and concentrate on the lower registers?  If i knew the answer to this, it would help in the future 

Edited October 14, 2022 by neil phillips

[ONIKKINEN]

8 minutes ago, neil phillips said:

I think the question is should we ignore the peaks and concentrate on the lower registers?  If i knew the answer to this, it would help in the future 

Tricky question, one i dont know how to solve thats for sure. DSO imaging has an easy crutch to lean on in photometric colour calibration, no such luck in planetary i guess and the process is more involved.

[neil phillips]

1 minute ago, ONIKKINEN said:

Tricky question, one i dont know how to solve thats for sure. DSO imaging has an easy crutch to lean on in photometric colour calibration, no such luck in planetary i guess and the process is more involved.

Yeah, it is tricky. But i think it's worth pursuing. Running members images into registax is very revealing. i urge all members to run it through. And it will get them thinking. 

I truly believe accurate colour balance is something worth thinking about. Though i know others will disagree. Each to their own of course. Its interesting you mentioning deep sky. 

How much do the deepsky experts worry about colour balance is it seen as important. its not my world so i know nothing about that ? 

[ONIKKINEN]

2 minutes ago, neil phillips said:

Yeah, it is tricky. But i think it's worth pursuing. Running members images into registax is very revealing. i urge all members to run it through. And it will get them thinking. 

I truly believe accurate colour balance is something worth thinking about. Though i know others will disagree. Each to their own of course. Its interesting you mentioning deep sky. 

How much do the deepsky experts worry about colour balance is it seen as important. its not my world so i know nothing about that ? 

I think it is important, most DSO imagers do too i think. Thing is, there are many opinions and few agree with eachother and most have their own style while trying to go real colour. with narrowband anything goes. Its a hot item so to speak. I think i remember a recent lengthy thread that got derailed immediately and the consensus was that brown is not a real colour or something like that..

In that way its the same with planetary. Range of similar but not quite the same renditions of colours that most push out.

[neil phillips]

12 minutes ago, ONIKKINEN said:

I think it is important, most DSO imagers do too i think. Thing is, there are many opinions and few agree with eachother and most have their own style while trying to go real colour. with narrowband anything goes. Its a hot item so to speak. I think i remember a recent lengthy thread that got derailed immediately and the consensus was that brown is not a real colour or something like that..

In that way its the same with planetary. Range of similar but not quite the same renditions of colours that most push out.

Right that makes sense i guess.  And like all things involving judgement. As you say hotly debated.  A lot of it comes down to experienced guesswork.  the more i think about it the more i start double guessing myself lol. Maybe I lean too much to blue. Others lean too much to red. Not many lean too much to green thats also interesting 

Edited October 14, 2022 by neil phillips

[neil phillips]

Manually lowering the green and blue a bit we get this. which out of the 3 looks best to me 

[CraigT82]

Jupiter has more reds/oranges/terracotta in it’s actual colour so you’d expect the red channel to be higher in the histogram than the other two. If you look at mono RGB captures of Jupiter the R is always brighter/lighter than the G or B. So balancing the overall image histogram so that all three channels are equal doesn’t really make sense to me. 
 

And who’s to say what the true colour of Jupiter is? Yeah there should be a way of doing it but colours will always be distorted, either by the personal tastes of who’s processing the image, or by other factors such as our atmosphere, telescope optics (eg looking through an eyepiece to get the true colour won’t work as your scope and EP will add their own colour cast and so will our atmosphere, often making things muddy). So what about space telescopes or probes, no atmosphere there so scientific imaging should reveal the true colour balance. Trouble is we very rarely see the truly unmanipulated colour images, we see the processed for the public versions that are almost always fiddled with and will contain the biases of whoever processed them. Take this official release Hubble image for example:

 

look at these two Juno images which are presented side by side on the NASA website, both have different colour processing, I’d be tempted to say the top one is more realistic with the bottom one a colour boosted/adjusted version? 

Here is another one that shows the white clouds not actually white :

What we really need is a Juno or Hubble image that is combined from mono RGB data that hasn’t had any normalisation or colour balancing. Then we would have a reference for what Jupiter really looks like. Perhaps someone who has done their own processing on public Juno/Hubble data could do? 
 

Edited October 14, 2022 by CraigT82

[Nik271]

I think we should send a SGLer on the next trip to the ISS with a decent scope and mono RGB camera set-up to answer this for us. 🙂

[Kon]

5 hours ago, neil phillips said:

Manually lowering the green and blue a bit we get this. which out of the 3 looks best to me 

 

Interesting topic Neil. Looking at your colour balance, in your final image you are aligning them at the left shoulder rather than the peak. Do you think it is a better representation of 'true colour' or better attempt to align them as at the peak you are at the maximum for the particular channel which may bias the colour balance? Ie the peaks should be at the highest sensitivity of the sensor. In addition, do you think RGB alignment in Autostakert might influencing how registax is doing it's RGB balance?

( I will have a go at my images when I am back home tonight and see how they turn out.)

[geoflewis]

This is a really excellent thread. I definitely find the colour of the planets tricky and see a lot of variability even from amongst the so called experts, e.g. Damian Peach, Chris Go, Anthony Wesley (Bird), et al. Like @neil phillips, I now usually use Registax colour balance tool for consistency, but as Neil reports Registax can get that wrong if the image is already well balanced, so I sometimes double click (to check what it did) and/or tweak the histogram. As @CraigT82 comments, we are not helped in our endeavours by the significant range in colours in images posted from Hubble and Juno data, as these reflect the preferences of the person processing them. Of course the colours of the planets vary over time so finding / using historical 'reference' images also doesn't work. In the end I think we all end up with somthing that pleases ourselves personally and is hopefully within a 'normal' range and not too 'out there'.

[newbie alert]

2 hours ago, CraigT82 said:

Jupiter has more reds/oranges/terracotta in it’s actual colour so you’d expect the red channel to be higher in the histogram than the other two. If you look at mono RGB captures of Jupiter the R is always brighter/lighter than the G or B. So balancing the overall image histogram so that all three channels are equal doesn’t really make sense to me. 
 

And who’s to say what the true colour of Jupiter is? Yeah there should be a way of doing it but colours will always be distorted, either by the personal tastes of who’s processing the image, or by other factors such as our atmosphere, telescope optics (eg looking through an eyepiece to get the true colour won’t work as your scope and EP will add their own colour cast and so will our atmosphere, often making things muddy). So what about space telescopes or probes, no atmosphere there so scientific imaging should reveal the true colour balance. Trouble is we very rarely see the truly unmanipulated colour images, we see the processed for the public versions that are almost always fiddled with and will contain the biases of whoever processed them. Take this official release Hubble image for example:

 

look at these two Juno images which are presented side by side on the NASA website, both have different colour processing, I’d be tempted to say the top one is more realistic with the bottom one a colour boosted/adjusted version? 

Here is another one that shows the white clouds not actually white :

What we really need is a Juno or Hubble image that is combined from mono RGB data that hasn’t had any normalisation or colour balancing. Then we would have a reference for what Jupiter really looks like. Perhaps someone who has done their own processing on public Juno/Hubble data could do? 
 

Hubble doesn't use any lrgb , so it's all false colour, without searching I'd assume it to be the same for Juno as it's a scientific instrument, not for pretty pictures 

[newbie alert]

9 hours ago, neil phillips said:

It also may depend on the monitor. But yes it does look more blue. But then again. The other image does look too red. I think the reason it may look a bit blue. Is colour saturation. i use a lot of saturation and it can do this to the image unfortunately. Dropping saturation a bit, does help 

Btw nobody likes white, everyone likes red

Also to be fair i said ALMOST bang on. Actually i think i did add just a smidge to much blue

 

Hi Neil, do you add the blue saturation before you colour balance?

[neil phillips]

Yeah its tricky isnt it. I am unsure to be honest how to get closer to a consensus

2 hours ago, CraigT82 said:

Jupiter has more reds/oranges/terracotta in it’s actual colour so you’d expect the red channel to be higher in the histogram than the other two. If you look at mono RGB captures of Jupiter the R is always brighter/lighter than the G or B. So balancing the overall image histogram so that all three channels are equal doesn’t really make sense to me. 
 

And who’s to say what the true colour of Jupiter is? Yeah there should be a way of doing it but colours will always be distorted, either by the personal tastes of who’s processing the image, or by other factors such as our atmosphere, telescope optics (eg looking through an eyepiece to get the true colour won’t work as your scope and EP will add their own colour cast and so will our atmosphere, often making things muddy). So what about space telescopes or probes, no atmosphere there so scientific imaging should reveal the true colour balance. Trouble is we very rarely see the truly unmanipulated colour images, we see the processed for the public versions that are almost always fiddled with and will contain the biases of whoever processed them. Take this official release Hubble image for example:

 

look at these two Juno images which are presented side by side on the NASA website, both have different colour processing, I’d be tempted to say the top one is more realistic with the bottom one a colour boosted/adjusted version? 

Here is another one that shows the white clouds not actually white :

What we really need is a Juno or Hubble image that is combined from mono RGB data that hasn’t had any normalisation or colour balancing. Then we would have a reference for what Jupiter really looks like. Perhaps someone who has done their own processing on public Juno/Hubble data could do? 
 

Good points I will have to spend time thinking about this. Something else I have noticed certain sides to Jupiter at different times. show different hues of white. So even if the processing was exactly the same the hue of white isn't actually white its tinted. Other times i see pure as snow whites.

But if the contention is Jupiter doesn't actually have pure white clouds. Then i am wrong at the outset and might as well try and figure that out best I can. Though I am not yet convinced that there are no pure whites on Jupiter. How do we tell the difference between a red channel being too high. And a stronger red channel sitting where we would expect ? Seems like a reason to fall down a rabbit hole of increasingly red Jupiter's if we are not careful.

If lots of images were being processed bright green. many on here would think ok the colour balance is off. Like unadjusted from raw. But very red Jupiter's and very blue Jupiter's are ignored. As if to nod well we don't know so might as well say nothing type thinking. It's a mess. The art of Jupiter imaging in many ways is a mess.

And what's worse without knowing i could even be adding to that problem. Anyway, good points i am going to have to think about this more. do some in depth research. Not exactly sure where or how at the moment. But feel its worth doing. Good points again Craig 

Edited October 14, 2022 by neil phillips

[neil phillips]

2 hours ago, Kon said:

Interesting topic Neil. Looking at your colour balance, in your final image you are aligning them at the left shoulder rather than the peak. Do you think it is a better representation of 'true colour' or better attempt to align them as at the peak you are at the maximum for the particular channel which may bias the colour balance? Ie the peaks should be at the highest sensitivity of the sensor. In addition, do you think RGB alignment in Autostakert might influencing how registax is doing it's RGB balance?

( I will have a go at my images when I am back home tonight and see how they turn out.)

I am afraid i am finding it hard to answer those questions. Kon. Especially after Craigs input. I am unsure of my own decision making to the point of not knowing how much to trust the histogram on registax. I don't like uncertainty but I feel uncertain at the moment. I don't think RGB alignment in AS/3 is affecting overall balance. More likely just colour alignment Kon. Do what you said you was going to do. See if you can learn anything. if you can then please contribute. 

Edited October 14, 2022 by neil phillips

[neil phillips]

1 hour ago, geoflewis said:

This is a really excellent thread. I definitely find the colour of the planets tricky and see a lot of variability even from amongst the so called experts, e.g. Damian Peach, Chris Go, Anthony Wesley (Bird), et al. Like @neil phillips, I now usually use Registax colour balance tool for consistency, but as Neil reports Registax can get that wrong if the image is already well balanced, so I sometimes double click (to check what it did) and/or tweak the histogram. As @CraigT82 comments, we are not helped in our endeavours by the significant range in colours in images posted from Hubble and Juno data, as these reflect the preferences of the person processing them. Of course the colours of the planets vary over time so finding / using historical 'reference' images also doesn't work. In the end I think we all end up with somthing that pleases ourselves personally and is hopefully within a 'normal' range and not too 'out there'.

It's a nightmare isn't it Geof we need a better consensus. At least an understanding of all the problems we are facing and how we can learn to go forward from them. Just not exactly sure how at the moment. Anyone know anyone at NASA that might have some ideas? Just kidding

[geoflewis]

4 minutes ago, neil phillips said:

It's a nightmare isn't it Geof we need a better consensus. At least an understanding of all the problems we are facing and how we can learn to go forward from them. Just not exactly sure how at the moment. Anyone know anyone at NASA that might have some ideas? Just kidding

I'm not even sure that there is a correct answer, so yes, some sort of concensus, is likely the best we can do 🤔. If we think about DSO NB imaging (not something I do BTW), it seems that the colours are adjusted to help better reveal structures and I think to some extent we do this in planetary processing without even considering it. Even the process of wavelet sharpening in Registax alters the colours, never mind any saturation that we apply to give images more 'pop'. I suspect that the varying bandwidth of different broadband colour (RGB) filters used in mono RGB, or even OSC cameras will yield their own variations. Colours off the 462 chip reflect its greater sensitivity to IR and we know what adding IR to an RGB image does to the colours.... 🙄

[neil phillips]

13 minutes ago, geoflewis said:

I'm not even sure that there is a correct answer, so yes, some sort of concensus, is likely the best we can do 🤔. If we think about DSO NB imaging (not something I do BTW), it seems that the colours are adjusted to help better reveal structures and I think to some extent we do this in planetary processing without even considering it. Even the process of wavelet sharpening in Registax alters the colours, never mind any saturation that we apply to give images more 'pop'. I suspect that the varying bandwidth of different broadband colour (RGB) filters used in mono RGB, or even OSC cameras will yield their own variations. Colours off the 462 chip reflect its greater sensitivity to IR and we know what adding IR to an RGB image does to the colours.... 🙄

All true Geoff. Plenty we don't know. Need to find some things we do know. Would have been helpful if NASA had taken a full revolution here. Not sure if they did or didn't? They do talk about blue clouds though. So there's a start Jupiter does have blue clouds 

Jupiter in True and False Color | NASA Solar System Exploration

 

 

Edited October 14, 2022 by neil phillips

[neil phillips]

Looks like one of my colour images the day after new years eve🤣

Hubble View of Jupiter | ESA/Hubble (esahubble.org)

 

Edited October 14, 2022 by neil phillips

[vlaiv]

32 minutes ago, neil phillips said:

It's a nightmare isn't it Geof we need a better consensus. At least an understanding of all the problems we are facing and how we can learn to go forward from them. Just not exactly sure how at the moment. Anyone know anyone at NASA that might have some ideas? Just kidding

Auto white balance uses fairly simplistic algorithm. In fact there are several algorithms - and none are guaranteed to produce accurate color.

It either aligns histogram peaks - which is sort of "gray world" approach - which advocates that images are on average "gray" - meaning same amount of each color component should be present in the image.

There is actually very well defined workflow that should be used if we want to present accurate color image of the planet, but it is somewhat complex.

I can outline steps needed, and we can then discuss it.

First step is always the same and that is color calibration of sensor used. We need common baseline, and luckily we have it. It is called standard XYZ observer and we can think of it as generic sensor that we must match.

Here it is - above should be viewed as QE of some standardized sensor.

compare that to say ASI678mc

or ASI224

These curves are different - and our job is to derive transformation that will transform recorded raw R, G and B into XYZ.

Once we have that - raw XYZ components - we need to account for atmospheric extinction. This is effect of reddening of the image if object is at lower altitude (more atmosphere).

Last step depends on what we want to show.

a) do we want to show actual "color" of the planet

b) do we want to match the color of planet as it is in orbit.

To understand the difference - imagine holding a white paper - one that is white when we take it outside on a sunny day.  Now imagine you are holding the same paper in orbit. Difference being that it is not illuminated by same light. In the same way there is atmospheric extinction for the target - there is for the sun as well (it is also red when viewed close to horizon).

In our solar system - objects are illuminated by 5778K black body radiator. However - in daylight (due to all the blue light scattering on atmosphere) - we actually have 6500K illuminant.

We need to choose between the two - if we want the image of say Jupiter as it is floating in empty space and so are we as we observe it, or we want the look of "Jupiter material" if it sits in our room as we are working at our computer.

You will notice - that none of above steps is actually employed by any of photographers and consequence is - all colors are different.

[neil phillips]

2 minutes ago, vlaiv said:

Auto white balance uses fairly simplistic algorithm. In fact there are several algorithms - and none are guaranteed to produce accurate color.

It either aligns histogram peaks - which is sort of "gray world" approach - which advocates that images are on average "gray" - meaning same amount of each color component should be present in the image.

There is actually very well defined workflow that should be used if we want to present accurate color image of the planet, but it is somewhat complex.

I can outline steps needed, and we can then discuss it.

First step is always the same and that is color calibration of sensor used. We need common baseline, and luckily we have it. It is called standard XYZ observer and we can think of it as generic sensor that we must match.

Here it is - above should be viewed as QE of some standardized sensor.

compare that to say ASI678mc

or ASI224

These curves are different - and our job is to derive transformation that will transform recorded raw R, G and B into XYZ.

Once we have that - raw XYZ components - we need to account for atmospheric extinction. This is effect of reddening of the image if object is at lower altitude (more atmosphere).

Last step depends on what we want to show.

a) do we want to show actual "color" of the planet

b) do we want to match the color of planet as it is in orbit.

To understand the difference - imagine holding a white paper - one that is white when we take it outside on a sunny day.  Now imagine you are holding the same paper in orbit. Difference being that it is not illuminated by same light. In the same way there is atmospheric extinction for the target - there is for the sun as well (it is also red when viewed close to horizon).

In our solar system - objects are illuminated by 5778K black body radiator. However - in daylight (due to all the blue light scattering on atmosphere) - we actually have 6500K illuminant.

We need to choose between the two - if we want the image of say Jupiter as it is floating in empty space and so are we as we observe it, or we want the look of "Jupiter material" if it sits in our room as we are working at our computer.

You will notice - that none of above steps is actually employed by any of photographers and consequence is - all colors are different.

Great points Vlaiv. Even the answer to that might be different for some people. Some may prefer actual sky conditions at the time colour. Some may prefer actual image in space true colour. It gets more confusing by the second. I think actual true colour is a better way to think about it. As sky conditions are just changing true colour into earths distorted false colour. So, no consensus there. Just more of a myriad of different hues caused by scatter and refraction