M33

[don4l]

This is my first proper LRGB image, and I am very pleased with it.  I've always had a brain block in understanding how an "L" could bring quality to a low res RGB image.  A few weeks ago, I read that in fact the RGB just colourised the black and white "L" image.  This may be obvious to some,  but I had never thought of it that way before.

I ran a test in the Gimp.   I put a mono layer of some old "L" over a colour layer with the the mode set to "luminance".  Then I reversed the layers, and changed the blend mode to  "HSL Colour".  The results were identical!  The only difference was that I can get my head around colourising a mono image.

50m L, 25m R, 40m G, 55mB and 40m Ha.   (RGB  2x2).

Tak FSQ106, G3-16200, EQ6,

CCDCiel, PHD2, CCDStack, Gimp

Any advice would be very welcome.

 

 

[x6gas]

Lovely.

[ollypenrice]

The image is nice.

I think of LRGB this way: a colour filter blocks about 2/3 of the visible spectrum light, a bit less maybe, but about that. An L filter passes all of it. So it takes an hour per colour filter (3 hours) to get as much light as you get in 1 hour of luminance.

In practice I find I don't really need 3 hrs of RGB to match 1 hr of L, I really need 4 hrs or so. The theorists disagree but I'm not a theorist, I just measure what I actually get.

Then we find that we don't need as much colour information as luminance. The RGB can be processed for low noise (essentially it can be blurred) and high saturation. The strong L layer can be processed for sharpness (the opposite of blurring) and depth of faint signal. As a general rule we are not expecting much colour from the faintest signal anyway.  When combined, the weaker RGB signal will work well with the stronger L signal.

I think that in your M33 you have a lot of colour noise in your background sky which it would be very easy indeed to remove since there is little real colour information in the background anyway. You could lose the colour 'detail' in the background (which is an artifact anyway) without smoothing the luminance so the sky would not look artificially noise reduced.

Keep an L-RGB mentality throughout the processing!

Olly

[don4l]

18 minutes ago, ollypenrice said:

The image is nice.

I think of LRGB this way: a colour filter blocks about 2/3 of the visible spectrum light, a bit less maybe, but about that. An L filter passes all of it. So it takes an hour per colour filter (3 hours) to get as much light as you get in 1 hour of luminance.

In practice I find I don't really need 3 hrs of RGB to match 1 hr of L, I really need 4 hrs or so. The theorists disagree but I'm not a theorist, I just measure what I actually get.

Then we find that we don't need as much colour information as luminance. The RGB can be processed for low noise (essentially it can be blurred) and high saturation. The strong L layer can be processed for sharpness (the opposite of blurring) and depth of faint signal. As a general rule we are not expecting much colour from the faintest signal anyway.  When combined, the weaker RGB signal will work well with the stronger L signal.

I think that in your M33 you have a lot of colour noise in your background sky which it would be very easy indeed to remove since there is little real colour information in the background anyway. You could lose the colour 'detail' in the background (which is an artifact anyway) without smoothing the luminance so the sky would not look artificially noise reduced.

Keep an L-RGB mentality throughout the processing!

Olly

Thanks Olly.

I'll have a go at reducing the background noise in the colour.  I've tried to save my work as I went along, so I should be able to go back.

 

[don4l]

12 hours ago, ollypenrice said:

The image is nice.

I think of LRGB this way: a colour filter blocks about 2/3 of the visible spectrum light, a bit less maybe, but about that. An L filter passes all of it. So it takes an hour per colour filter (3 hours) to get as much light as you get in 1 hour of luminance.

In practice I find I don't really need 3 hrs of RGB to match 1 hr of L, I really need 4 hrs or so. The theorists disagree but I'm not a theorist, I just measure what I actually get.

Then we find that we don't need as much colour information as luminance. The RGB can be processed for low noise (essentially it can be blurred) and high saturation. The strong L layer can be processed for sharpness (the opposite of blurring) and depth of faint signal. As a general rule we are not expecting much colour from the faintest signal anyway.  When combined, the weaker RGB signal will work well with the stronger L signal.

I think that in your M33 you have a lot of colour noise in your background sky which it would be very easy indeed to remove since there is little real colour information in the background anyway. You could lose the colour 'detail' in the background (which is an artifact anyway) without smoothing the luminance so the sky would not look artificially noise reduced.

Keep an L-RGB mentality throughout the processing!

Olly

I *think* that I've managed to do as you have suggested.  I used your method of lifting the left hand side of the curve in "curves adjustment" on the RGB data.  I don't know if I have lifted it enough. 

 

 

[ollypenrice]

56 minutes ago, don4l said:

I *think* that I've managed to do as you have suggested.  I used your method of lifting the left hand side of the curve in "curves adjustment" on the RGB data.  I don't know if I have lifted it enough. 

 

 

That background sky is way, way better on my monitor. Nice one. I might just give the green a tiny lift to kill the slight magenta caste of the image but that is a great tweak you did.

Olly

[don4l]

On 27/12/2019 at 10:11, ollypenrice said:

That background sky is way, way better on my monitor. Nice one. I might just give the green a tiny lift to kill the slight magenta caste of the image but that is a great tweak you did.

Olly

Thanks Olly.

Lifting the green didn't turn out to be as simple as I thought.  Parts of the image turned an obvious green when I gave it enough to remove the magenta.  To produce this version, I lifted the green a bit on the whole image, and then duplicated it.  I applied a layer mask (greyscale of image) and adjusted the green up a bit more in that layer.  I think that I can still see some magenta,  but I am not good with colour, so I wouldn't bet on it. 

I've cropped it a bit more to save bandwidth.