LRGB (split from APOD: NGC 7600 (2011 Dec 23))

[zloq]

http://asterisk.apod.com/viewtopic.php?f=9&t=26274

Hi Ann-

This is most likely an LRGB exposure, which uses a long "luminance" image to capture detail, combined with separate, shorter, filter images to provide color. This is never done by professional astronomers because the only reason to do so is to make a nice "looking" image - rather than one that has accurate photometric information. Most amateur color images ("true" color - not narrow band) use this method. It is based on the fact that human color perception is not very high resolution and can tolerate a lot of smoothing, whereas the eye can see high-res detail in brightness or luminosity. So you capture a pretty good amount of color info so the colors are basically accurate in a short amount of total exposure - then you overlay a long, high-res luminance image that "multiplies" the color - to get the best of both worlds - color and resolution. Note that it isn't "bad" to use LRGB - it makes good sense to get the nicest looking result in a given amount of exposure time. But it doesn't improve the true SNR of the data - only the perceived SNR.

Note that this is a tricky variation on the theme - because if he used a "clear" filter rather than a "luminance" filter - his luminance info has IR photons that are not captured by the color filters. A clear filter has no UV/IR blocking filters that a luminance filter has. Some people will use an Ha image as the luminance and multiply over RGB - which makes a dramatic, detailed result - but it doesn't really have a physical interpretation.

When amateurs use this technique to make an image that looks nice and has pleasing colors - they all use different rules to achieve that goal. Some people try to do it accurately by calibrating the color using reference stars, while others just make it look overall about right in photoshop. Either way - nothing about the LRGB should make anything preferentially gray. Done right it should capture the "true" variation of color in the scene - but when photoshop is involved and colors are arbitrarily manipulated, it's hard to know what is what. In this case, given that the stars show a pretty good variation from red to blue, unless the galaxy was specifically masked off and processed differently (which many people do) I assume it is relatively bland and orange-gray looking.

zloq

[Chris Peterson]

This is most likely an LRGB exposure, which uses a long "luminance" image to capture detail, combined with separate, shorter, filter images to provide color. This is never done by professional astronomers because the only reason to do so is to make a nice "looking" image - rather than one that has accurate photometric information.

To be clear, the method is extremely common amongst professional astronomers. The difference is that while amateurs design their imaging around the technique, with the primary intent of producing an aesthetic image, professionals design their imaging around a particular scientific goal. In practice, that often leaves them with high S/N data collected with no filter, or with a broad photometric filter like V, and with lower S/N data made with narrowband filters. One product of the data is often an image which approximates true color (these images are common on APOD), and such an image is made using the LRGB or a related processing method.

[zloq]

To be clear, the method is extremely common amongst professional astronomers.

I certainly disagree. Professionals aint making pretty pictures at which to ogle - they are getting quantitative photometric information and high-res detail in particular wavelengths of interest. There is no improvement in photometric accuracy by this method - only an aesthetic improvement. If you can provide examples where, for example, someone got time on the Hubble telescope to do LRGB exposures and combined the results in this manner - and then used the pretty picture as a basis for publication - I would be happy to take a look. Even the Hubble heritage site is at most RGB - I don't know of any LRGB. There just wouldn't be scientific reason to do additional exposures in a separate channel that encompassed all the others - and then multiply them. The motivation for amateurs is to make a prettier picture in a given amount of time - and the benefit is real - in aesthetic terms. And if you use a clear filter, as in this case, you get to use IR flux that otherwise would not have been captured by the color filters. But scientifically that would be mixing information, and the result would become non-physical.

I think a lot of amateurs think there really is an improvement in SNR by this technique - but it's not true. It's explained incorrectly in some books - but the Handbook for Astronomical Image Processing by Berry and Burnell gets it right - describing it as something that works specifically for the human visual system. Even the tiny little wiki page on the LRGB technique describes it as an amateur technique.

Here is a pretty good description of it by NOAO - and it is explicitly referred to as a technique that is a big win *for amateurs*.

http://www.noao.edu/outreach/aop/glossary/lrgb.html

zloq

[Chris Peterson]

I certainly disagree. Professionals aint making pretty pictures at which to ogle - they are getting quantitative photometric information and high-res detail in particular wavelengths of interest. There is no improvement in photometric accuracy by this method - only an aesthetic improvement.

No, you don't disagree at all <g>. You just misunderstood me.

My point was that the intent of amateurs and professionals is usually quite different (it certainly appears we are in agreement here). But many of the "accurate" color images from professionals that we see on APOD, and generally released for public consumption, are produced using LRGB or similar techniques- because that is the only method available given the sort of data typically available. And the intent here is primarily aesthetic- even though that wasn't the intent when the data was collected.

The accurate color professional images are often "re-purposed"; the accurate color amateur images were designed that way.

[zloq]

But many of the "accurate" color images from professionals that we see on APOD, and generally released for public consumption, are produced using LRGB or similar techniques- because that is the only method available given the sort of data typically available.

I think I did hear you say this, and I do certainly disagree. How can you make an LRGB composite image if there was never time taken for an L exposure? And why would professionals using a big ground-based observatory or the HST take the time for such an exposure? How would they justify it in a request for telescope time? It may well have happened - but I don't know of a *single* example of an HST LRGB imaging session. Just R,G,B, or, more commonly, some specific mixture of distinct filters.

Certainly folks are taking exposures from the HST and big observatories and applying amateur methods to them to give them more zing - but you cannot make an LRGB image if you only have R,G,B exposures. Usually the quality of the professional R, G, B exposures is so good there is less win with a separate L exposure - whereas in amateur work you can bin the color channels to reduce read noise - etc. I guess there are opportunities to do something like HaRGB with professional data - and the results would be pretty wild - but non-physical - and not LRGB.

Here is an example instructing people on how to make an RGB image from Hubble data. They use images based on the log of intensity in 3 channels, I, V, B, and then they go nuts in photoshop to make a nice result. But no use of a separate L channel to multiply the RGB composite - which is what distinguishes LRGB. The data are so good you don't need the L.

http://hubblesource.stsci.edu/services/ ... 005-02-10/

zloq

[Chris Peterson]

How can you make an LRGB composite image if there was never time taken for an L exposure?

You can't. But it is very common for professionally acquired images to include L data.

And why would professionals using a big ground-based observatory or the HST take the time for such an exposure? How would they justify it in a request for telescope time? It may well have happened - but I don't know of a *single* example of an HST LRGB imaging session. Just R,G,B, or, more commonly, some specific mixture of distinct filters.

You obviously do miss the point I made. They don't plan for an LRGB session. But they frequently acquire LRGB or something similar because of their science goals. For example, data is often collected with no filter, and it is often collected with a photometric V filter. In both cases, that provides an L channel. RGB may be obtained with RGB filters, or more commonly synthesized from multiple narrowband filters.

Many of the "pretty" HST images that get released are essentially the result of LRGB processing. The data were used in different ways, and the pretty picture was a byproduct- often produced by someone other than the original researchers.

Here is an example instructing people on how to make an RGB image from Hubble data. They use images based on the log of intensity in 3 channels, I, V, B, and then they go nuts in photoshop to make a nice result. But no use of a separate L channel to multiply the RGB composite - which is what distinguishes LRGB. The data are so good you don't need the L.

They collect pure L all the time, because no filter is often the best choice for dim objects. In addition, a photometric V filter basically gives you an L channel. I've generated many LRG images using photometric V, Ha, and O[III] filters (with B synthesized). This is a very common approach with Hubble Heritage releases.

[zloq]

I don't mean something similar - I mean LRGB to achieve an aesthetically pleasing true color image - like the current APOD. You seem to regard all imaging techniques as equivalent to LRGB - but I am specifically talking about LRGB - not SiiFuchsiaEmeraldOiii. Furthermore - I most certainly do not consider V equivalent to L.

I do agree that HST users don't plan for LRGB sessions - which is why I haven't seen any HST LRGB images. An LRGB image is a true color RGB image obtained from a long luminance exposure coupled to relatively short RGB exposures that "color in" the gray scale L image - through multiplication. Essential to the LRGB technique is that the L filter is implicitly equal to the sum of the passbands of the R, G, B filters - so the multiplication makes sense and the result is equivalent to a long RGB image. The only reason to do LRGB is to achieve maximum aesthetic impact in a given total exposure time - which has no scientific relevance since there is no true gain in SNR of the color channels - only the visual appearance of improved SNR. That's why it's a technique that's only relevant to amateurs.

zloq

[Chris Peterson]

I don't mean something similar - I mean LRGB to achieve an aesthetically pleasing true color image - like the current APOD. You seem to regard all imaging techniques as equivalent to LRGB - but I am specifically talking about LRGB - not SiiFuchsiaEmeraldOiii. Furthermore - I most certainly do not consider V equivalent to L.

Well, a photometric V filter provides standardized luminance- data over nearly the entire visible light range, and matching the typical response of a CCD with an IR blocking filter. And I'm also talking about LRGB to achieve an aesthetically pleasing true color image. It's done all the time with Hubble data- both by amateurs going back over archived data, and by organizations like Hubble Heritage doing the same. Many of those reprocessed images have appeared as APODs.

[zloq]

You keep talking about this stuff being done all the time - so provide some examples. HST LRGB. Also - an HST VRGB image where a Johnson V filter was used for the multiplicative luminance channel (??).

Ummm - the V filter stands for Visual, and it is intended to match the central region of the visual system - not the entire bandpass of the visual system. It is GREEN because it does not represent a uniform transmission across the ENTIRE visual system - which is what an L filter does - being cutoff only in the IR and UV. So a VRGB image would be similar to GRGB and wouldn't make a whole lotta sense as a way to capture true colors, but you could make such a GRGB image with any RGB data set - if that's your idea of a good time. You can always take RGB and multiply it by X filter to make XRGB - but it would not be LRGB. It would make more sense to say V is like G, so RGB is very similar to RVB - and I believe that substitution is done by amateurs since it lets you do true(ish) color and photometry at the same time without a separate G exposure. Having V in the central region of the visual bandpass and B next to it capturing the blue lets you define color in terms of the spectral index, B-V. A V-band exposure is not a luminance exposure - it is a V-band exposure.

The Johnson UBV system has been replaced by the Sloan system, and the Johnson V has been replaced by Sloan g or g'. I own and use both - for photometry not imaging.

You said early in this thread that LRGB is extremely common among professional astronomers - and I have provided links and explanations of why it is specific to amateurs since I think that is an important point in an APOD discussion. You clearly have a very blurred view of imaging if you regard L the same as V and tri-color narrow-band the same as true color LRGB. So I am not interested in examples that are not LRGB - but examples that *are* LRGB - since you say it is "extremely common among professional astronomers." The key characteristics - *once again* are that the L exposure is long compared to the RGB, and L serves as a detailed gray scale with colors filled in by the RGB through multiplication.

Here is a plot of filter bandpass similar to the ones used for this APOD. Note that the Luminance channel encompasses the R, G, B channels - and the G channel is only the central green region - like the Johnson V filter. That's why they look a similar green color. Some amateur astro filter sets are designed to leave a gap between colors where certain light pollution lines are common - but aside from that they are intended to span the spectrum in a similar manner to the human eye.

http://www.alpineastro.com/filters/imag ... er_set.jpg

Notice how the L bandpass has sharp edges and tightly hugs the interior R, G, B. Also shown is the C filter - which does not block IR. If you do CRGB instead of LRGB, then you are mixing in IR with the color channels and it will no longer be true color, but it will still look pretty good and it will have more photons for the luminance channel. It departs from true color, though.

I'm not sure what your point is in arguing this topic since I was merely describing LRGB in response to Ann's comment - and at the same time I pointed out that it is an amateur technique. I don't argue that professional astronomers do all kinds of imaging through different filters - which you seem to regard as "equivalent" to LRGB - but my point was regarding this APOD image and its use of an L channel for luminance to generate a true color image - as is very common for amateurs and *not* among professionals.

zloq

[Chris Peterson]

You keep talking about this stuff being done all the time - so provide some examples.

I've got too much to do to track one down. But they show up all the time... I'll try to remember to point out the next one.

Ummm - the V filter stands for Visual, and it is intended to match the central region of the visual system - not the entire bandpass of the visual system. It is GREEN because it does not represent a uniform transmission across the ENTIRE visual system - which is what an L filter does - being cutoff only in the IR and UV.

A V filter chops off part of the blue and part of the red. The blue cutoff corresponds to the natural response of many CCDs. The red cutoff flattens the overall response, since CCDs tend to be most responsive to red and near IR. A V filter looks green because it is doubling up the natural response of your eye. It makes a CCD have an overall response similar to your eye. A V filter does not work for collecting green data- that is, you can't use the data collected to synthesize a green channel in order to produce a color image. It is very broadband, however, and produces an excellent luminance channel, assuming you have other filtered data to get your chrominance information from.

A "VRGB" image is very much a LRGB image, not a GRGB image.

There is no fundamental concept of a luminance filter. The best possible L filter is no filter at all, since the goal is to maximize the signal. That is the most commonly used L source with HST images. In general, with reflective optics you use no filter. With a refractor, chromatic aberration becomes an issue, so you often need to narrow the spectral range. Various "luminance filters" are marketed for this purpose, but if you already have a photometric V filter, you can easily use it to get good luminance data. Many people do so.

[zloq]

I disagree with almost every single thing you say there.

As for V not being a substitute for G because it is "does not work for collecting green" - there are many web examples that support this usage. Here is one from the European Southern Observatory:

http://www.eso.org/sci/libraries/lisa4/Cuillandre.pdf

"This leaves only the V filter for the green channel. The best combination is then the R-V-B to match the RGB channels."

They have examples of RVB color images, which you say would not work.

zloq

[Chris Peterson]

I disagree with almost every single thing you say there.

As for V not being a substitute for G because it is "does not work for collecting green" - there are many web examples that support this usage. Here is one from the European Southern Observatory:

http://www.eso.org/sci/libraries/lisa4/Cuillandre.pdf

"This leaves only the V filter for the green channel. The best combination is then the R-V-B to match the RGB channels."

They have examples of RVB color images, which you say would not work.

They are doing something different. The example is combining images made only with photometric filters. So it makes sense that a reasonable match is RVB to RGB (although it doesn't work very well). But most HST images are not made using photometric filters- and when they are, it is usually V and I filters that are used. Most of the impressive shots that have been released are made using either unfiltered or V for luminance, and then the RGB channels are synthesized from the available narrowband data.

Standard methodology is to break the image into LAB or luminance/chrominance space for processing. It is always best if you can start with a true luminance channel- and there are several wideband options frequently available with HST images.

[Jim Franklin]

Could I simply point out that you do not need need L images per se if you take enough single images of the same object with the same setup..

I could type a whole piece on it here for the process I have used, but this link provides a far better explanation than I can type up at this time..

http://www.mlunsold.com/process/LiveVie ... lanet.html

I think there is a clear misinterpretation of what is being said by both above as your both approaching it from different angles and your both right, depending on perspective...