Question about processing NB data

[StuartT]

So far I have only done OSC imaging and when I integrate the subs, they are first aligned with each other. But if I were doing LRGB, presumably only the Ls get aligned and integrated, then only the Rs and so on. So I end up with an L master, a R master etc which I then LRGB combine in PI. 

But what if the four masters are not aligned with each other? I'd then end up with an image with the channels out of alignment (which would look like CA). I don't think LRGB combination does any aligning does it?

Sorry if I have misunderstood something

[VectorQuantity]

6 hours ago, StuartT said:

So far I have only done OSC imaging and when I integrate the subs, they are first aligned with each other. But if I were doing LRGB, presumably only the Ls get aligned and integrated, then only the Rs and so on. So I end up with an L master, a R master etc which I then LRGB combine in PI. 

But what if the four masters are not aligned with each other? I'd then end up with an image with the channels out of alignment (which would look like CA). I don't think LRGB combination does any aligning does it?

Sorry if I have misunderstood something

Hello Stuart,

Having selected your usable sub-frames and calibrated and corrected, pick one of them. Align/register all other subs irrespective of the filter type to this sub-frame.

You could of course align individually for each filter having picked a suitable sub-frame as reference for each filter from that filter set, then you'd have to align the masters.

I believe some people (not me) if using the first method also run a subsequent step to align/register the LRGB masters prior to combination, not sure why.

n.b. make sure you correct the subframes i.e. get rid of artefacts prior to alignment/registration to ensure the software you are using does not detect hot pixels as stars.

Edited June 28, 2022 by VectorQuantity
clarification

[StuartT]

18 minutes ago, VectorQuantity said:

Hello Stuart,

Having selected your usable sub-frames and calibrated and corrected, pick one of them. Align/register all other subs irrespective of the filter type to this sub-frame.

You could of course align individually for each filter having picked a suitable sub-frame as reference for each filter from that filter set.

I believe some people (not me) also run a subsequent step to align/register the LRGB masters prior to combination.

n.b. make sure you correct the subframes i.e. get rid of artefacts prior to alignment/registration to ensure the software you are using does not detect hot pixels as stars.

Thanks for explaining. That makes sense

[Elp]

I have achieved this in Siril in the RGB composite function but it's a bit picky with regard to the files being used (ie if they're different sizes it throws up errors, sometimes it just states error without explaining why when you load a file, I suspect it doesn't like tif files only fits), recently when I tried doing it it didn't work so I had to manually align to one frame (luminence) for the r, g, b manually in PS. Only started using Siril, it's good for preparing your mono files prior to doing any major editing.

Edited June 28, 2022 by Elp

[wimvb]

6 hours ago, StuartT said:

I don't think LRGB combination does any aligning does it?

Alignment as per @VectorQuantitys reply.

LRGB combination does’t do any alignment. In fact, you can combine any mono image with any RGB image as long as they have the same number of pixels (width and height).

Before LRGB combination:

create an RGB image from the three masters. Process this for optimised colour. Never mind the details, but do keep the stars in check. Make sure that these don’t lose their colour.

Process the L master for details and local contrast. Deconvolution, sharpening, local contrast enhancement, etc.

Finally combine the L with the RGB. Uncheck R, G, and B channels in LRGB combination. Choose the L master for L and apply the process to the RGB image. An Easter Egg here: the LRGB combination process can assign weights to the various channels. Set the weight for L to 0.2-0.5 if you want to apply the process in small steps.

[ollypenrice]

Although it's expensive, I find Registar to be a great boon. It is a super-accurate aligning-resizing-mosaic-making software. I use it all the time. Say I've reached the stage where I do a final crop and apply it. Then I have a bright idea for a processing refinement which involves using an earlier, uncropped layer. I just open the cropped and the uncropped in Registar and use it to align and crop the old image to the cropped. Likewise I can add new data in the same way.  It's just one of those tools I use time and time again.

Applying luminance:  Wim mentions the problem of luminance overpowering RGB if applied at 100%. This is certainly a problem. I address it in Photoshop by applying the Luminance in partial opacity, starting at about 20%, giving the result a slight blur, increasing the saturation, flattening and re-applying the luminance to that. I carry on with this several times till, in the end, I can apply the luminance at 100% and not give it any blur.  This final application prevents any residual blur.

It might also be worth resisting the temptation to take the luminance to its highest possible stretch. Leave a bit of potential stretch unused before combining with the RGB. Then go for the final, ultra-hard stretch with the LRGB. There are theoretical arguments against this technique but I find it can simplify life.

Olly

[StuartT]

23 hours ago, wimvb said:

Alignment as per @VectorQuantitys reply.

LRGB combination does’t do any alignment. In fact, you can combine any mono image with any RGB image as long as they have the same number of pixels (width and height).

Before LRGB combination:

create an RGB image from the three masters. Process this for optimised colour. Never mind the details, but do keep the stars in check. Make sure that these don’t lose their colour.

Process the L master for details and local contrast. Deconvolution, sharpening, local contrast enhancement, etc.

Finally combine the L with the RGB. Uncheck R, G, and B channels in LRGB combination. Choose the L master for L and apply the process to the RGB image. An Easter Egg here: the LRGB combination process can assign weights to the various channels. Set the weight for L to 0.2-0.5 if you want to apply the process in small steps.

interesting. That would never have occurred to me.

Now all I have to do is remember where this thread is when I finally manage to be able to afford a filter wheel, mono camera and ludicrously expensive set of filters 

Thank you

UPDATE:

So I managed to obtain some narrowband data (publicly shared) and tested out your procedure. 

On the left is RGB combination then combined the result with the L channel
On the right is LRGB combined all at once

The left hand one is nicer in my opinion (though this is obviously subjective). But what I don't understand is why they are different? As a mathematician, I would expect them to have an associative property, i.e. L + (R+G+B) should be the same as L+R+G+B 

Edited June 29, 2022 by StuartT

[wimvb]

3 hours ago, StuartT said:

L + (R+G+B) should be the same as L+R+G+B

Assuming that the channels are added as in math (ie, assuming that LRGB combination is additive). But, the range for pixel values is 0 to 1. What if the sum is large than 1? Clip? Rescale?

[wimvb]

5 hours ago, StuartT said:

interesting. That would never have occurred to me.

Now all I have to do is remember where this thread is when I finally manage to be able to afford a filter wheel, mono camera and ludicrously expensive set of filters 

Thank you

UPDATE:

So I managed to obtain some narrowband data (publicly shared) and tested out your procedure. 

On the left is RGB combination then combined the result with the L channel
On the right is LRGB combined all at once

The left hand one is nicer in my opinion (though this is obviously subjective). But what I don't understand is why they are different? As a mathematician, I would expect them to have an associative property, i.e. L + (R+G+B) should be the same as L+R+G+B 

Sorry for the short previous answer; I was in a hurry.

LRGB combination isn’t an RGB pixel operation, but rather performed in the L*a*b colour space. You can verify this by using channel combination in PI. Choose the Lab colour space, uncheck a and b, and enter the name for the L image in the L box. Then apply to the RGB image. The result is identical to LRGB combination.*

RGB combination is an RGB pixel action; it applies the R image to the R channel, G to the G channel, and B to the B channel. Neither process is simple addition, hence L+R+G+B is not equal to L+RGB.

Edit: * Actually not quite identical. I checked with PixelMath

1. created a colour image with only a red gradient (X() in PixelMath)

2. created a luminance image with another gradient (Y() in PixelMath)

3. Combined L with the red RGB image (LRGB combination) = image_3

4. Combined L with the RGB image (L*a*b channel combination) = image_4

5. subtracted image_4 from image_3 ( abs(image_3 - image_4) in PixelMath)

The resulting image wasn't completely black, so image_3 and image_4 where somewhat different, but not much.

Edited June 29, 2022 by wimvb