Recently had an sms asking this question, might be beneficial to share some of my thoughts here:
Dating back to the late 1980s and early 1990s, astrophotographers would use the method of tri-colour imaging to capture deep sky objects. For each object, each primary colour channel would be taken (R, G and B). However, one of the drawbacks of doing imaging this way especially for faint DSOs (where SNR is low), is the need to go really long exposures for each colour channel. This seemed rather straight forward to do, but the process of ensuring good consistent guiding, low or almost null tracking errors, focus shifts, made this process laboriously challenging. In the early 2000s, a new method called LRGB was devised to cope with this issue. Besides capturing data from each of the colour channels, an additional luminance frame (L) was captured of the same object using a clear filter. This clear filter would usually be carefully selected to have UV- and IR-cutoffs and covers about the same bandpass as the RGB filters added together. With the addition of the L component, deep and detailed structures could be recorded in faint DSOs, without the need to go for really long exposures for each colour channel. Nevertheless, ultimately, in order to get best SNR for the object in colour, if you could, go deep for the colour (R,G,B) channels as well, where your system is reliable enough to do so.
However, the actual concept of LRGB has adapted various meanings as different astrophotographers experimented with various methods of capture and combination. Where clear sky time was limited, some have experimented with deriving a synthetic (or fake) L channel by combining the existing R,G,B data together. This "combined" L channel is then added to the RGB component. This has the effect of both increasing the SNR slightly as a result of lowering the noise. Surprisingly, the effects was notable and you could use this method where clear sky time is limited. But of course, a "real" L channel that is captured independently will always get a better image. This is because the effective read noise levels in a L image is lower than the combined RGB with each having similar read noise levels. Unless you are striving for sky-limited exposures for each colour channel, this will always be the case - hence the real LRGB is always a better deal.
So ultimately, having an LRGB composite will provide caveats in terms of bringing out faint data like the faint lobes of galactic arms, or the outer regions of a nebula, etc. However, if one has a very deep and long L channel captured independently and combined with relatively shorter R,G,B data exposures, there will be an apparent loss in colour data. The faint regions will appear "grey", and the entire image will loose its colour richness. As such, one will still have to take longer exposures for the colour channels, in order to yield richer tones. Personally, I have encountered this in the worse possible case. Taking a rich and deep luminance channel from Malaysia's dark skies, and combining it with shorter R, G, B exposed channels taken from Singapore! The resulting image looked rich in details but it was almost greyed out. A silly endeavour...
Is there a need for LRGB?
Hehe, I think now the cow can Moo.
So we learn from Remus that when you want to combine LRGB, the SNR of L and RGB should be pretty close. I think thats why it is recommended that the RGB be taken with 2x2 binning to increase sensitivity and yet the exposure can be shorter.
According to Mr Peach, planetary LRGB pics are harder to colour balance especially Mars. and Mars looks better in RGB than LRGB.
So we learn from Remus that when you want to combine LRGB, the SNR of L and RGB should be pretty close. I think thats why it is recommended that the RGB be taken with 2x2 binning to increase sensitivity and yet the exposure can be shorter.
According to Mr Peach, planetary LRGB pics are harder to colour balance especially Mars. and Mars looks better in RGB than LRGB.
Photo Album:
http://www.flickr.com/photos/14113965@N03/
http://www.flickr.com/photos/14113965@N03/
yes, Meng Lee, there is always the option of binning to increase sensitivity per area.
However, the concept of LRGB and RGB in my post was pertaining to deep sky objects only. For planets, the concept slightly different. Some would use other components as their L channels depending on what they would like to bring out. Saturn for example, uses L (unchanged) typically. Jupiter, L=IR. Mars L=R, and so on. But there is no fixed rule on these, it all depends on what you want to enhance or better the image. Also for planetary, exposures are far shorter.
However, the concept of LRGB and RGB in my post was pertaining to deep sky objects only. For planets, the concept slightly different. Some would use other components as their L channels depending on what they would like to bring out. Saturn for example, uses L (unchanged) typically. Jupiter, L=IR. Mars L=R, and so on. But there is no fixed rule on these, it all depends on what you want to enhance or better the image. Also for planetary, exposures are far shorter.
I guess the biggest drawback of LRGB will be colour balancing (white balancing) especially for nebula whose emission is very dominant in R. For L = Clear, its not so bad, but for L = Ha the white balancing should be terrible. And when Ha is used, the star sizes would be different from RGB, so I think that's why Gendler said, L=Ha+R for nebulae will be better for colour balancing and star sizes matching.
Photo Album:
http://www.flickr.com/photos/14113965@N03/
http://www.flickr.com/photos/14113965@N03/
hmm..so many things to learn. any gd books or website to recommend?
one qns, what abt the amt of time required for each colour channel? how does one determine the amt of time required for each colour?
another qns, wat abt h-alpha and binning, does one do binning for colour info, or does one use 1x1?
concerning mosaics, does one stitch the pieces together first, or does one combine the colour and then stitch it together?
~MooEy~
one qns, what abt the amt of time required for each colour channel? how does one determine the amt of time required for each colour?
another qns, wat abt h-alpha and binning, does one do binning for colour info, or does one use 1x1?
concerning mosaics, does one stitch the pieces together first, or does one combine the colour and then stitch it together?
~MooEy~
actually i do not see it as a drawback. Here, L is treated more as a symbol. The channel can be an L, or it could very well be Ha, or Ha+R. Infact, this could be likewise for the colour channels as well. It is highly unlikely that the top imagers are using completely pure R, G, or B in their LRGB composites. Why? Let us take Eta Carinae for an example. It is dominant in Ha very much. In the past, very often we see apparently beautiful Eta Carinae in deep red. This is the consequence of using pure Ha in the L domain, or possibly a combination of Ha+R. Infact, whenever one is imaging Ha, and intending to "mix" into the composite, you must have R together, otherwise you will see stars with central spot-like brightening. The R component is mixed together to "bloat" up the stars and have it "on par" with the other colour components. On the other hand, you could lessen the effect of Ha otherwise, but that may not go down well if you have taken great effort yielding a good Ha component. There are alternative hybrid combinations that I have experimented that works even better. If you study Eta Carinae more carefully from those taken with professional observatories, the Key Hole central region is not just Ha dominant, but has an array of emissivity responses going from Sulphur to Oxygen radicals. Now, if you take Ha alone, you are not going to get a balanced Eta Carinae which it should be. One would still need to go longer exposures in blue and green to "balance up". Alternatively, there are short-cuts to this which I have experimented. Use Ha on blue channel. Just a bit. Maybe 10% opacity. Try it, you will notice that overall response is more balanced. Of course, this Ha little component on blue channel must be Ha+B. This is one golden rule in Ha combination for any of the colour channels. Have been doing this for all the ST2KXM images obtained so far. They are never pure HaRGB. It would not work well. Even for NGC7000, you will notice it is not pure Ha dominant. Good images of the North American have slight central pinkish and biege hues. However, this is not the case for the Pelican.
But nowadays, even HaRGB is losing out in effectiveness to (Ha+R)RGB or even (Ha+L)RGB. As we go deeper into exploring the best ways to bring out details in faint regions or strive for ever increasing SNR, hybrid combination methods have to be used. The conventional LRGB is just a stepping stone. And do take note, that LRGB originated from the days of film. The purpose of LRGB in CCD imaging was to mimic the logarithmic response of film, which had very good saturation and rich tonality that CCD imaging lacked then. Now, we have composites and hybrid methods that we can work with in CCD imaging, giving it a boost till today.
But nowadays, even HaRGB is losing out in effectiveness to (Ha+R)RGB or even (Ha+L)RGB. As we go deeper into exploring the best ways to bring out details in faint regions or strive for ever increasing SNR, hybrid combination methods have to be used. The conventional LRGB is just a stepping stone. And do take note, that LRGB originated from the days of film. The purpose of LRGB in CCD imaging was to mimic the logarithmic response of film, which had very good saturation and rich tonality that CCD imaging lacked then. Now, we have composites and hybrid methods that we can work with in CCD imaging, giving it a boost till today.
MooEy wrote:hmm..so many things to learn. any gd books or website to recommend?
most of my time is spent in experimenting various methods. image processing is all about maths mainly. lolx. but you can start from rod wodaski's books. i think u have mine??!
one qns, what abt the amt of time required for each colour channel? how does one determine the amt of time required for each colour?
ok, here let's talk about CCD imaging only. this is where the flexibility lies. If you want to take a good image of an object, you must know the character of the object first. Read up on it. Know its emission character. If possible, look at all the good images of it from professional observatories, not just from the amateurs. They are after amateurs, so there are room for improvement. Let's take M45 for example. Should we take 1:1:1 exposures for R,G,B. Well, we could, but perhaps 1:1:2 or 1:1:1.5 ratios may be better to bring out Merope's nebula for example. But if you look at excellent images of M45, you will notice that there are not just blue reflective nebulosities present, but only so faint Ha component at the "bottom" of the cluster. There you are, you probably would want to include some Ha. Now let's take NGC7000 for example. This is a completely different ball-game. It has lots of Ha. So you must have a Ha filtered exposure added to the usual 1:1:1 RGB ratio. However, some may want to take 2:1:1. But I believe the latter ratio will result in blood-bleeding stars. If you note in greater detail as per my earlier post, you will notice that NGC7000 has other central hues. So it is probably not a very good idea to have too dominant reds in the image. NGC7000 is actually slightly pinkish. We can see the same thing for M8 as well. There are too many blood red M8s floating around the net that gives you the impression that this should be the right balance. Sorry, this is not the case. We have one fine example taken by Frank Sky Ghost. Look at it. Observe the bottom part of M8, it has reflective components. And the overall hue for M8 is probably skewed more towards magenta. So you must have some blues there.
On another point, the exposure ratios will also depend on what filters you are using. I am referring to Goldman's filters for the above argument, since the transmission throughput per colour filter is balanced already. If you are using CS, or Astronomik filters, then you have to do some experimentation first to note what is the actual exposure ratios in order to get 1:1:1 at the end of the day. You may be getting 1:1:0.7 instead for these filter sets. So it means to get 1:1:1, you need to calculate and extend the B channel more in order to "balance up". But if you are taking M45 with this filter set, it also means if you want to go deeper for B, you need to expose even longer for this filter set in B than using the Goldman B filter.
another qns, wat abt h-alpha and binning, does one do binning for colour info, or does one use 1x1?
Actually, i am a fond believer of 1x1 only even for colour channels. Why? In order to have an excellent image with rich and DETAILED colour structures, i would rather have 1x1 and go long exposures or take in dark skies. But if you are fine with an ok image and battling with limited clear sky time, you can opt for binning methods, and yes, if you were to use binning for each colour channel, you must employ this method for all the rest of the colour channels. After which, if you take another 1x1 L channel, you must scale up the 2x2 coloured channels by resampling during your processing.
concerning mosaics, does one stitch the pieces together first, or does one combine the colour and then stitch it together?
right now, i am still working on the best way to do colour mosaics. the main problem is the varying sky gradient profile. Also, if you take one component mosaic from another on different nights, this would be an added challenge, since seeing might vary too, which results in different star bloat sizes. once you have tackled the sky gradient issues and nullify it for each mosaic component, I do not think the sequence matters anymore. At least from my understanding now from a recent colour E.Carinae mosaic.
~MooEy~
hmm..so many things to learn. any gd books or website to recommend?
most of my time is spent in experimenting various methods. image processing is all about maths mainly. lolx. but you can start from rod wodaski's books. i think u have mine??!
one qns, what abt the amt of time required for each colour channel? how does one determine the amt of time required for each colour?
ok, here let's talk about CCD imaging only. this is where the flexibility lies. If you want to take a good image of an object, you must know the character of the object first. Read up on it. Know its emission character. If possible, look at all the good images of it from professional observatories, not just from the amateurs. They are after amateurs, so there are room for improvement. Let's take M45 for example. Should we take 1:1:1 exposures for R,G,B. Well, we could, but perhaps 1:1:2 or 1:1:1.5 ratios may be better to bring out Merope's nebula for example. But if you look at excellent images of M45, you will notice that there are not just blue reflective nebulosities present, but only so faint Ha component at the "bottom" of the cluster. There you are, you probably would want to include some Ha. Now let's take NGC7000 for example. This is a completely different ball-game. It has lots of Ha. So you must have a Ha filtered exposure added to the usual 1:1:1 RGB ratio. However, some may want to take 2:1:1. But I believe the latter ratio will result in blood-bleeding stars. If you note in greater detail as per my earlier post, you will notice that NGC7000 has other central hues. So it is probably not a very good idea to have too dominant reds in the image. NGC7000 is actually slightly pinkish. We can see the same thing for M8 as well. There are too many blood red M8s floating around the net that gives you the impression that this should be the right balance. Sorry, this is not the case. We have one fine example taken by Frank Sky Ghost. Look at it. Observe the bottom part of M8, it has reflective components. And the overall hue for M8 is probably skewed more towards magenta. So you must have some blues there.
On another point, the exposure ratios will also depend on what filters you are using. I am referring to Goldman's filters for the above argument, since the transmission throughput per colour filter is balanced already. If you are using CS, or Astronomik filters, then you have to do some experimentation first to note what is the actual exposure ratios in order to get 1:1:1 at the end of the day. You may be getting 1:1:0.7 instead for these filter sets. So it means to get 1:1:1, you need to calculate and extend the B channel more in order to "balance up". But if you are taking M45 with this filter set, it also means if you want to go deeper for B, you need to expose even longer for this filter set in B than using the Goldman B filter.
another qns, wat abt h-alpha and binning, does one do binning for colour info, or does one use 1x1?
Actually, i am a fond believer of 1x1 only even for colour channels. Why? In order to have an excellent image with rich and DETAILED colour structures, i would rather have 1x1 and go long exposures or take in dark skies. But if you are fine with an ok image and battling with limited clear sky time, you can opt for binning methods, and yes, if you were to use binning for each colour channel, you must employ this method for all the rest of the colour channels. After which, if you take another 1x1 L channel, you must scale up the 2x2 coloured channels by resampling during your processing.
concerning mosaics, does one stitch the pieces together first, or does one combine the colour and then stitch it together?
right now, i am still working on the best way to do colour mosaics. the main problem is the varying sky gradient profile. Also, if you take one component mosaic from another on different nights, this would be an added challenge, since seeing might vary too, which results in different star bloat sizes. once you have tackled the sky gradient issues and nullify it for each mosaic component, I do not think the sequence matters anymore. At least from my understanding now from a recent colour E.Carinae mosaic.
most of my time is spent in experimenting various methods. image processing is all about maths mainly. lolx. but you can start from rod wodaski's books. i think u have mine??!
one qns, what abt the amt of time required for each colour channel? how does one determine the amt of time required for each colour?
ok, here let's talk about CCD imaging only. this is where the flexibility lies. If you want to take a good image of an object, you must know the character of the object first. Read up on it. Know its emission character. If possible, look at all the good images of it from professional observatories, not just from the amateurs. They are after amateurs, so there are room for improvement. Let's take M45 for example. Should we take 1:1:1 exposures for R,G,B. Well, we could, but perhaps 1:1:2 or 1:1:1.5 ratios may be better to bring out Merope's nebula for example. But if you look at excellent images of M45, you will notice that there are not just blue reflective nebulosities present, but only so faint Ha component at the "bottom" of the cluster. There you are, you probably would want to include some Ha. Now let's take NGC7000 for example. This is a completely different ball-game. It has lots of Ha. So you must have a Ha filtered exposure added to the usual 1:1:1 RGB ratio. However, some may want to take 2:1:1. But I believe the latter ratio will result in blood-bleeding stars. If you note in greater detail as per my earlier post, you will notice that NGC7000 has other central hues. So it is probably not a very good idea to have too dominant reds in the image. NGC7000 is actually slightly pinkish. We can see the same thing for M8 as well. There are too many blood red M8s floating around the net that gives you the impression that this should be the right balance. Sorry, this is not the case. We have one fine example taken by Frank Sky Ghost. Look at it. Observe the bottom part of M8, it has reflective components. And the overall hue for M8 is probably skewed more towards magenta. So you must have some blues there.
On another point, the exposure ratios will also depend on what filters you are using. I am referring to Goldman's filters for the above argument, since the transmission throughput per colour filter is balanced already. If you are using CS, or Astronomik filters, then you have to do some experimentation first to note what is the actual exposure ratios in order to get 1:1:1 at the end of the day. You may be getting 1:1:0.7 instead for these filter sets. So it means to get 1:1:1, you need to calculate and extend the B channel more in order to "balance up". But if you are taking M45 with this filter set, it also means if you want to go deeper for B, you need to expose even longer for this filter set in B than using the Goldman B filter.
another qns, wat abt h-alpha and binning, does one do binning for colour info, or does one use 1x1?
Actually, i am a fond believer of 1x1 only even for colour channels. Why? In order to have an excellent image with rich and DETAILED colour structures, i would rather have 1x1 and go long exposures or take in dark skies. But if you are fine with an ok image and battling with limited clear sky time, you can opt for binning methods, and yes, if you were to use binning for each colour channel, you must employ this method for all the rest of the colour channels. After which, if you take another 1x1 L channel, you must scale up the 2x2 coloured channels by resampling during your processing.
concerning mosaics, does one stitch the pieces together first, or does one combine the colour and then stitch it together?
right now, i am still working on the best way to do colour mosaics. the main problem is the varying sky gradient profile. Also, if you take one component mosaic from another on different nights, this would be an added challenge, since seeing might vary too, which results in different star bloat sizes. once you have tackled the sky gradient issues and nullify it for each mosaic component, I do not think the sequence matters anymore. At least from my understanding now from a recent colour E.Carinae mosaic.
You might want to see this website:
http://starizona.com/acb/ccd/advtheorycolor.aspx
I think most of our discussion boils down to : What is the correct colour balancing? Are you willing to trade off a bit of resolution for imaging time (the binning part) ?
With atmospheric factor taken into account, I think probably the dirtiest way to achieve colour balance is to store a library of 1:1:1 images of G2V stars at various altitude and white balance using the library images. So you can actually image RGB in 1:1:1 and the library of G2V star images will allow you to compensate for filter-ccd mismatch and altitude correction, of course given that RGB all not underexposed nor overexposed.
I agree with Remus that gradient is headache in mosaics, especially when nebulae is very extensive in the image. Field curvature will be another big headache as u start piecing things up.
http://starizona.com/acb/ccd/advtheorycolor.aspx
I think most of our discussion boils down to : What is the correct colour balancing? Are you willing to trade off a bit of resolution for imaging time (the binning part) ?
With atmospheric factor taken into account, I think probably the dirtiest way to achieve colour balance is to store a library of 1:1:1 images of G2V stars at various altitude and white balance using the library images. So you can actually image RGB in 1:1:1 and the library of G2V star images will allow you to compensate for filter-ccd mismatch and altitude correction, of course given that RGB all not underexposed nor overexposed.
I agree with Remus that gradient is headache in mosaics, especially when nebulae is very extensive in the image. Field curvature will be another big headache as u start piecing things up.
Photo Album:
http://www.flickr.com/photos/14113965@N03/
http://www.flickr.com/photos/14113965@N03/