11 replies to this topic

[Mert]

Hi all,

If I remember correctly, some time ago on CN there has
been a couple of posts with tips and tricks on generating
an artifical/synthetic blue channel.
Typically over here the seeing is not very good so the
blue channel is nearly always much worse then red and
green channel.
There were some tips or programs that did the trick but
I can't find them anymore.

Thanks for any feedback

[WayneJ]

Hi Mert,

While you can't truly create a synthetic blue channel because the blue data lies outside of the passband of the R/G filters, you can "mimic" one by using pixel match on the R and G images using the formula 2G-R=B.

Recently, I've been working on a more effective way and that is to capture a luminance instead of the blue channel using a UV/IR cut filter. From that, you can subtract the red and green data and have the blue data remaining (theoretically). The nice thing is that you can capture a LUM at a much higher framerate than the R or G and will indeed have actual blue data if the R and G are subtracted out correctly.

Regards,

Wayne

[Mitchell Duke]

Good idea Wayne, but how do you subtract the red and green from luminance?

[WayneJ]

Hi Mitch,

I'm working on that. An accurate answer may be somewhat filter dependent, but in theoretically "perfect" filters, you would simply subtract the red and green amount in each of those filters from that in the luminance (I use the image>apply image>subtract feature in Photoshop). It's somewhat more complicated because of the "crossover" of the passband of the filters, but shouldn't be too hard to work-out. I took some data the other night in lousy seeing and haven't had a chance to work on it yet.

Wayne

[Mert]

Hi Wayne,

Thanks for the info and the tip, I will follow this thread
with attention since there seems to be quit some
interesting info and experience out there!

[Sunspot]

Couldn't you get a very close approximation of blue by separating the Luminance layer into R,G,B layers and discarding the R and G layers and using the B layer with the previously captured Red and Green layers?

Paul

[Mike Phillips]

I do recall a HUGE thread on this in years past right here in the Planetary section. Someone took some Hubble shots and did an RGB combine compared to a RGsB and it was pretty imperceptible.

Mike

[WayneJ]

Couldn't you get a very close approximation of blue by separating the Luminance layer into R,G,B layers and discarding the R and G layers and using the B layer with the previously captured Red and Green layers?

Hi Paul,

If you capture the lum in monochrome through a UV/IR filter, then the resulting image is monochrome. When you convert it to RGB, it places a copy of the same image in each channel, dividing the luminance among them equally. So, if you do this you will end up substituting a copy of the lum for the blue.

Wayne

[Sunspot]

Ahhh, gotcha! That make sense. I see it now.

Thanks!!
Paul

Couldn't you get a very close approximation of blue by separating the Luminance layer into R,G,B layers and discarding the R and G layers and using the B layer with the previously captured Red and Green layers?

Hi Paul,

If you capture the lum in monochrome through a UV/IR filter, then the resulting image is monochrome. When you convert it to RGB, it places a copy of the same image in each channel, dividing the luminance among them equally. So, if you do this you will end up substituting a copy of the lum for the blue.

Wayne

[DesertRat]

I have only a limited experience in this but did perform some experiments a couple of years ago. What I noted then was:

1) It is very important for the red and green to be aligned accurately before doing the pixel math B = (2*G) - R.
The are several ways to do the alignment including image processing programs, or centering routines such as Ninox or Castrator. Resampling before centering can be useful to minimize alignment errors.

2) The red and green should be scaled approximately with the same level of histogram fill.

3) It is preferable to perform the synthetic channel math before any other processing steps such sharpening. Otherwise artifacts may occur. Even so the syn blue will appear a little noisier than the red and green as a result of the math operation.

I would only add this is really almost more of a novelty than anything else. I suppose if you forget to take a blue or as sometimes happens one takes a RGG sequence by mistake it might come in handy.

Wayne's idea seem interesting. I would point out however luminance images are often a bit softer than individual channels due to chromatic focus shift between the channels. This is more true for SCT's than Newts. Forget about achro refractors.

Cheers,
Glenn

[CPellier]

Hi guys
The orignal thread is here

Where WayneJ recalls cleverly that it's completely impossible to synthesize B from R and G. The B data can't be found outside its band. Just imagine if you were trying to synthesize R from G and B ?
You can make convicing "fakes" on a planet like Jupiter where the details are relatively the same on each channels, but at the expense of true colors. Images will be pink-greenish. However, it will be hopeless on Saturn and above all Mars, where the B details are widely different even from G and R...
First of all, the best method to go on with this is... to try to get good B images . It's not that difficult ; a first step is to improve the capturing sequence : re-focusing, lowering the frame rate, getting longer B movies.
Next, a very good tool exists to considerably improve the quality of the B image : the atmospheric dispersion corrector. All the amateurs that are using it are happy of it. Damian Peach does, just as an example. It even improves the G channel also... Beautiful RGB images have been done with it.
Finally, the true LRGB method (with the luminance coming from an IR-cut filter) is also a good way to go. But because all the colors are mixed together, you will absolutely need the atmospheric corrector, or the L will be blurred, unless the planet is very high in the sky.
My personal philosophy about processing... never use a processing that obviously tweak the "reality". Synthezised channels, IR luminances...

[DesertRat]

Actually one of the original threads is here:

http://www.cloudynig.../Number/3271558

There was an earlier dicussion as well, but the meat of it is in the the thread I'm pointing to.

CPellier made some good points. No one would argue RGsB is true color. One should make every effort to get a decent blue channel as he suggested. For SCT imaging its kind of shocking to see how far in one has to go for a good blue focus.

Look at the Meade 8 report at:

http://www.airylab.c... 2010-47001.pdf

Sorry its in French but under "Décalage des meilleurs foci sur l’axe" it shows a blue offset of a startling 140 microns at f/10!

Cheers,
Glenn