Synthetic luminance creation in Photoshop

[alan4908]

Firstly, a bit of background: I've recently adopted an imaging post processing workflow for my one shot colour camera that uses the so called LLRGB technique. Basically, the idea is that you take your one shot (RGB) image and at an early stage in the processing workflow you split it into a Luminance layer (on which you perform all the sharpening to reveal the detail) and the RGB layer (on which you perform all the colour enhancement and colour noise reduction). Towards the end of the workflow you combine the image back together in two Luminosity steps, hence the term LLRGB. Apparently, all this is supposed to boost your signal to noise ratio, so you should end up with a better end result that if you did all the processing in just the RGB space.

Anyway, my question is what is the best way to convert a RGB image into a (synthetic) Luminance image ?

At the moment I'm using (with the RGB image open): IMAGE->MODE->GREYSCALE but I'm not sure if this method is maximizing the signal to noise ratio in the image.

Alan

[steppenwolf]

I have used the method you suggest and it works very well. You could also experiment with the following method:-

Duplicate your RGB image

Flatten the image

Select Image - Mode - Lab Colour

Select the 'Channels' Tab

Select the 'Lightness' channel

Copy and paste the Lightness channel into a new document

Use this as your Luminance channel

[alan4908]

Thanks Steve

I will have a go !

I was also thinking about using the channel mixer in Photoshop and setting the output to greyscale. I presume this would allow me to choose which colours I wanted to mix based on the input image. The problem is that I don't know how to measure the resultant signal to noise ratio in Photoshop - the only way I can think of doing this with my software is to create the grey scale image in Photoshop and move it back to MAXIM DL and then to measure the SNR via the information window - unfortunately this would be a bit laborious - what I really want is for someone to make a Photoshop plug-in signal to noise meter.

Alan

[ollypenrice]

Coward that I am, and unable to choose between the methods mentioned by Steve, I do it in AstroArt because it has a button for extracting luminance from RGB. My thinking is that the two PHD physicists behind AstroArt might - just conceivably - know more about it than I do...   

Pretty bad answer. I'm ashamed of it!

Olly

[alan4908]

After a bit of searching on the internet - I discovered that Photoshop does provide the means to measure the SNR in an image.  

Since SNR = mean signal/standard deviation, you can derive the information from the histogram which gives the mean and standard deviation for the image. Furthermore, if you select a particular area, the SNR relates to the area selected. So, in my latest image of M31, I decided to compare the IMAGE->MODE Greyscale to the Lab Mode Lightness suggested by Steve. I measured the SNR at 5 difference places across the image covering the core, dust lanes and background. The results indicate that the the LAB mode Lightness method gives a better SNR in the core and dust lanes but a lower result in the background areas.

core: Lightness SNR: 54.1; greyscale SNR: 37.2

dust lane#1: Lightness SNR: 11.6; greyscale SNR: 10.6

dust lane#2:Lightness SNR: 33.0; greyscale SNR: 27.9

background#1: Lightness SNR: 10.0; greyscale SNR: 11.6

background#2: Lightness SNR: 9.8; greyscale SNR: 11.7

So, for this particular image, I'd gain 10% to 18% SNR in the dust lanes from using the LAB mode Lightness method of synthetic luminescence creation, which I think would be worth having. Presumably, if I wanted to maximize the SNR throughout the entire image I'd use the IMAGE->MODE greyscale for the background and the LAB lightness for remainder of the image and merge the two together via layer masks.

Alan 

[Xiga]

Firstly, a bit of background: I've recently adopted an imaging post processing workflow for my one shot colour camera that uses the so called LLRGB technique. Basically, the idea is that you take your one shot (RGB) image and at an early stage in the processing workflow you split it into a Luminance layer (on which you perform all the sharpening to reveal the detail) and the RGB layer (on which you perform all the colour enhancement and colour noise reduction). Towards the end of the workflow you combine the image back together in two Luminosity steps, hence the term LLRGB. Apparently, all this is supposed to boost your signal to noise ratio, so you should end up with a better end result that if you did all the processing in just the RGB space.

Anyway, my question is what is the best way to convert a RGB image into a (synthetic) Luminance image ?

At the moment I'm using (with the RGB image open): IMAGE->MODE->GREYSCALE but I'm not sure if this method is maximizing the signal to noise ratio in the image.

Alan

Hmm, i like the sound of this method of separating the Luminance from the colour data.

I'm also a PhotoShop guy. Can you explain how you go about doing the re-combining at the end?

[Owmuchonomy]

You can do it like Steve says of course.  You can also do it the way Scott Rosen (see link below) does it which starts off in similar fashion to your suggestion.

http://www.astronomersdoitinthedark.com/dslr_llrgb_tutorial.php

[alan4908]

Xiga 

The recombination in Photoshop is straightforward - open your RGB and Luminance images. Copy the Luminance image and paste it as a layer

onto the top of the RGB image. Then change the blending mode of the Luminance layer to Luminosity. 

The LLRGB method actually does two recombinations (hence the two L's). The first time you recombine, you set the opacity slider on the Luminosity layer to somewhere between 0 and 100%, the second time, the opacity is set to 100%. The exact value of opacity setting at the first recombination is dependent on the colour saturation level - as a general guide I've found it is normally between 50 to 60%.  

You can see all this in action at the excellent video tutorial by Scott Rosen which Owmuchonomy has referenced (above).

Alan 

[Xiga]

Many thanks Alan & Chris. Will be sure to give this a go once the skies clear (not holding my breath!)

[fwm891]

Try opening your RGB in PS and make a duplicate layer.

Highlight the duplicate layer to activate it then add an adjustment layer to it and select B&W.

Click on the adjustment layer to activate it then right click it and from the drop down menu select clipping mask, (this makes the adjustment layer only active on the duplicate layer leaving your RGB as an RGB)

Click on the duplicate layer and change its blend mode to Luminance.

You can then add additional adjustment layers to either the luminance or the RGB for levels, curves, saturation etc etc... Just make sure you set them as clipping masks so they only act on their respective image layer.