How do you decide how much RGB to acquire?

[swag72]

This is how I see it, but if I'm wrong please do correct me.

When doing an LRGB image, the L data gives you the detail in the image, the dust lanes for example. The RGB data acts as a kind of colour wash, so that you can effectively fill the luminance data with some colour.

Assuming that I have that principle right, I have started to acquire luminance data first, get a load and then add the RGB. So, in my case I have 69x 300s of luminance data. How do I decide who much RGB to add to that? And how do I know that I have enough?

For example, I did a rough process of my LRGB data. I only had 6x300s of RGB as I had to discard some subs and so I just equalled the numbers up from the channel with the least amount. My overall image was lacking a little in colour, the galaxy didn't have the colour that I'd like to see. So, is this because I've not got enough RGB data or poor processing (again!!!)

Sorry for so many stupid and basic questions. I think I'm filling up the imaging discussion board!!!

[ollypenrice]

Firstly well done on getting those 69x5 min L subs! That's the spirit!

You'll need to get a lot more than 6 subs per channel to fill 69 L subs with colour, though. My guess is that you'll need about 90 minutes per channel. If you lose a sub or two in a particular colour there is no need to reduce the size of the stack in the other colours. It won't throw out the colour balance, which I guess you'll do in Pixinsight? It's the best routine I know.

If you are a bit short of colour and the image looks washed out when you add the L, try this;

Add L to RGB in Ps at about 30% opacity, blend mode Luminance. Flatten. Increase saturation. Blur/noise reduce/despeckle.

Re-apply the L to this semi-L/RGB image and try 60% this time. Again blur and noise reduce it and flatten, and increase saturation. Now try a third iteration of L. All the time the layer you are calling RGB is getting a little more L in it but you are noise reducing it and increasing saturation. Eventually there's a good chance that it will take a heavy application of L.

Another one (learned from Martin B and with thanks); take your RGB into Lab colour mode (Image-Mode) and split channels. Go to Image, Adjustments, Brightness and Contrast (the only time you will use this numpty tool!!) and, on the a channel and the b channel only, increase contrast by a whopping 30. Recombine the channels and reconvert to RGB. This is a low noise way of boosting saturation. I use it all the time.

Tom also says he likes it.

Olly

[RikM]

I just tried the lab colour contrast adjust with my Rosette WIP and it works a treat. I normally use the method given in one of Anna Morris's video tutorials. Convert to LAB colour, select levels and for both a and b, increase the low number by 20-30 while decreasing the high number by the same amount. This has a similar effect.

[swag72]

Cheers both - So is there a general rule Olly for a % of RGB compared to L?

Thanks for the tip for colour - Now where's my bilingual PI - CS dictionary?!!!

Looks like I may have to go back to CS afterall for a few things!

[x6gas]

I've read that you should aim to capture about the same amount of RGB as L but that you can capture it binned 2x2 and reduce the capture time for each channel by a quarter. So if you have 240 minutes of L, you grab around 60 minutes of RGB binned 2x2, with maybe a bit more for the blue channel.

I'd love to know people's thoughts on this and also if the same applies if you've used Ha as the L channel (in which case I assume you can get away with less RGB, but i've no idea what a sensible ratio is...

[martin_h]

Firstly well done on getting those 69x5 min L subs! That's the spirit!

You'll need to get a lot more than 6 subs per channel to fill 69 L subs with colour, though. My guess is that you'll need about 90 minutes per channel. If you lose a sub or two in a particular colour there is no need to reduce the size of the stack in the other colours. It won't throw out the colour balance, which I guess you'll do in Pixinsight? It's the best routine I know.

If you are a bit short of colour and the image looks washed out when you add the L, try this;

Add L to RGB in Ps at about 30% opacity, blend mode Luminance. Flatten. Increase saturation. Blur/noise reduce/despeckle.

Re-apply the L to this semi-L/RGB image and try 60% this time. Again blur and noise reduce it and flatten, and increase saturation. Now try a third iteration of L. All the time the layer you are calling RGB is getting a little more L in it but you are noise reducing it and increasing saturation. Eventually there's a good chance that it will take a heavy application of L.

Another one (learned from Martin B and with thanks); take your RGB into Lab colour mode (Image-Mode) and split channels. Go to Image, Adjustments, Brightness and Contrast (the only time you will use this numpty tool!!) and, on the a channel and the b channel only, increase contrast by a whopping 30. Recombine the channels and reconvert to RGB. This is a low noise way of boosting saturation. I use it all the time.

Tom also says he likes it.

Olly

Another way to add the Lum to RGB is to load the combined RGB and copy the layer, desaturate the layer to black and white. Load the Lum and copy it to a new layer over the top. Now rearrange the layers so that Lum is at the bottom, the desaturated RGB is in the middle, and the RGB is at the top(you will have to rename the bottom layer to move it), now set the blend mode of the top layer to "colour" the middle layer to "multiply" and the bottom layer to "normal".....adjust the middle-Multiply layer to suit....30-40% is a ball park setting.

[RobH]

In general, I would try to go for at least 20 subs each of RGB. This will give you a decent signal to noise ratio and means that you won't run into colour noise issues too quickly when pushing the data.

@Olly.....never heard of the Lab method before....I'll give that a bash

Rob

[swag72]

Do you bin the RGB? I am binning my RGB now and feel very naughty for doing so!

[Tim]

This is a low noise way of boosting saturation. I use it all the time.

Tom also says he likes it.

Olly

Olly, have you tried the softlight blend mode solution ala Ray Gabany? Its a two or three stage process for the same, very clean and easy in PS

[RobH]

Do you bin the RGB? I am binning my RGB now and feel very naughty for doing so!

I don't bin my RGB unless I'm short of time, in which case I bin, reduce sub length and still try for 20 or so for each filter.

[ollypenrice]

Tim, I don't know R Jay Gabany's method but will have a look. Thanks.

Plenty of good ideas on here.

I don't bin colour if working for myself but if guests are pushed for time we do it. There is a compromise on quality, notably in the star colour. I never bin at short FL.

However, so far as time is concerned I take the boring view that the best noise reduction is more exposure. The best saturation boost is more exposure. The best everything is...

I don't thnk you need equql time on LRGB. I'd say go for about equal RGB combined and L but it varies from image to image. I like not having a rigid workflow. I prefer it to be suggested by the individual image, as much as anything because it is more enjoyable to work that way, to me at least.

Olly

[harry page]

Hi

The true way to do LRGB is to bin your RGB , It was designed to save time

If you don't bin RGB it is said you are wasting your time doing LRGB and should spend your time just doing unbinned RGB

I have not tried a play off between the two , but would be interested if any body else has

Harry

Do you bin the RGB? I am binning my RGB now and feel very naughty for doing so!

[RobH]

Hi

The true way to do LRGB is to bin your RGB , It was designed to save time

If you don't bin RGB it is said you are wasting your time doing LRGB and should spend your time just doing unbinned RGB

I have not tried a play off between the two , but would be interested if any body else has

Harry

I've done it, and so has Olly I believe.

With binned RGB you definately lose star tightness and also resolution....it's subtle, but there.

Rob

[harry page]

I've done it, and so has Olly I believe.

With binned RGB you definately lose star tightness and also resolution....it's subtle, but there.

Rob

Then why bother with lum , why don't you just capture more RGB unbinned

Harry

[RobH]

Then why bother with lum , why don't you just capture more RGB unbinned

Harry

Time

In order to get a decent S/N ratio in each of the 3 filters, you're looking at (for me) at leat 50 full length subs each, so, 25 hours with 10 minute subs.

I go for around 60 x 5 or 6 minutes for colour in total, and then 50 or 60 full length luminance, a total of around 16 hours....still a long time but not a total marathon, plus, your individual luminance subs are seeing more light, being comprised of all the RGB photons.

If I was lucky enough to live in a country where you got night after night of clear skies, I don't think I'd bother with luminance at all, and would just shoot RGB.

Rob

[swag72]

If I was lucky enough to live in a country where you got night after night of clear skies, I don't think I'd bother with luminance at all, and would just shoot RGB.

Now that does interest me - I don't get night after night, but certainly do get a fair share of clear nights - So, why would you not bother doing luminance if that's the case?

Olly, do you shoot luminance as a rule or just RGB?

[RobH]

Check out Peter Shah's work....he generally only shoots RGB.

[x6gas]

Really interesting thread...

Any thoughts on my Ha as L question above? I.e should I still aim to shoot the same combined RGB as Ha?

[ollypenrice]

I don't have a rigid rule and am very suspicious of dogma.

If the object is really just a strong Ha emission nebula I wouldn't shoot Luminance and might shoot a derisory amount of RGB to combine with Ha. My Flying Bat nebula (one of my personal favourites) had just 30 minutes per channel RGB and about 5 hours of Ha. (You know what a nutter I am for long exposure so 30 mins is a snapshot.) Something like Simeis 147 is like that, too. Take forever in Ha but very little in RGB because there's nothing there apart fropm stars. Thanks to Tom for telling me this before I wasted too much time!

For objects with fine detail in broadband lots of Luminance is my instinct. Galaxies and intricate nebulae, for instance. You can really lean on an L layer for sharpening, as you can an Ha layer.

I can see the theory on dispensing with L and am full of admiration for Peter's stuff but RGB only doesn't work for me on many objects. It's like OSC. It can be great but sometimes it needs a kick in the rear to make shine. How you get that kick in the rear depends on the object. My own approach is 100% pragmatic. It may be sentimental or romantic but I do like to listen to (or rather look at!) the object I'm working on. I'd hate to work to a formula and, since I do this for fun, that's how I do it. I think about what the object seems to want. Galaxies, I'll shoot an eternity of L. You can never have enough.

Ha as luminance? No, you get a pink-out with blue star haloes. I add Ha to red in blend mode lighten but I process the Ha layer in a specific way with massive sharpness and contrast. In standalone mode it looks awful. I may add a little Ha as L at the end, rarely more than 15% but the FLying Bat would take a lot more. Don't know why.

Olly

[swag72]

What's your thoughts on a false luminance layer Olly?

[ollypenrice]

What's your thoughts on a false luminance layer Olly?

Well, you'd certainly use one if working from RGB only. I rarely do that but the same applies to processing OSC data. It's essential to do some things in a synthetic L layer, most notably any sharpening.

Olly

[swag72]

Cheers Olly - Next question then for you all!!

I am trying to ascertain the ratio of RGB to take to combine in an image. I have worked it out the following way.

Flats give me an average ADU of 27832: 27847: 27949 in RGB respectively with the following exposures 0.022: 0.021: 0.024

If I take 0.021 as 1 (green channel in this case) does this mean that I should be combining 5% more Red and 15% more Blue?

So for 100 exposures I should be aiming for 105:100:115 (RGB respectively)

Hope I've explained that well enough for you to understand it!!!

[martin_h]

For colour balance you could shoot a G2V star and use this method for the ratios...... Starizona's Guide to CCD Imaging

[swag72]

I will give that a go tonight and see how different it is to my flats!

[x6gas]

Thanks for your insight Olly.

I understand what you say about hard and fast rules, but for those of us starting out it really helps to have some pointers like this - doubly so when the time and trouble is taken to explain the rationale as you've done here.

Very many thanks.

Ian