RGB to Lab/CMYK and back to RGB

[DWThomp]

I asked a question in this thread (http://www.retouchpro.com/forums/pho...s-lab-raw.html) about going from RGB>Lab/CMYK>RGB and got some good answers from Robert & John. I usually stay in RGB and will change the color sampler readout to CMYK when doing some skin color corrections, but at times I have trouble adjusting when I'm working in RGB.

Robert had a work-around using a duplicate image converted to Lab or CMYK, flatten, convert to RGB, and drag back into the original file. This allows you to keep the previous adjustment layers in the original.

John brought up something that I've heard which is you can get shifts by converting from one to another and back. He also suggested that there are ways to minimize those impacts which should be the topic of another thread. So here it is.

Is the process that Robert uses the way to do it or are there other ways to minimize the impact. The stuff I do is probably not critical work, but I still try to do the best I can. I'd always heard the downside to going back and forth and just wonder what ways are suggested to lessen any changes.

Thanks to Robert & John for your thoughts in the other thread.

[edgework]

There's no reason to jump into CMYK if the image isn't exclusively meant to go to press. You lose all subtle detail in blues and greens and forget about pastels. If I know the image is targeted for print, I'll save some final moves for CMYK, when I know I'm not going back, (I keep an RGB layered file in any event). You can get some effective shadow moves with the black plate that aren't available in RGB, and skin tones respond well in CMYK. One reason is that many of the really hot values that RGB can come up with don't exist in CMYK and get crunched into a more realistic range. Another is that all shadow detail is in the black plate, so the moves you make with CMY curves have a more direct effect on the relationships between the different inks in skin tones.

You can still take advantage of CMYK even when you want to stay in RGB. Dupe the image, convert to CMYK and pull the black plate back to the RGB image as a selection mask. It's not quite as effective as a real black plate, but it allows you to make adjustments that RGB alone doesn't do well, separating shadows from color ranges. If you can create a good skin mask some nice refinements are available in CMYK, which you can then pull back to RGB. (I've never found skin that can't be improved by a luminosity contrast curve to the magenta channel, something that doesn't translate well in RGB).

When skin is too hot, copying the red plate, duping the image, converting to CMYK and swapping the original red for the new cyan gets around the dot gain compensation that Photoshop builds into CMYK conversions. The resulting skin is usually a bit too blue, but that can be cured with a quick curve, whereas you can try all day long to jack up the worthless cyan plate from which most information is bleached out. Then you can pull the skin back to RGB.

So there are ways to take advantage of the different spaces and what they accomplish, but moving an image wholesale into CMYK if it's not required isn't one of them.

[Tony W]

If you want to stay in RGB rather than converting back and forward then there is a plug in that allows you to do this.

I have not tried it but Curvemeister is said to give you the ability to work in RGB and edit in LAB, CMYK without leaving your preferred working environment,

[John Wheeler]

I tend to agree with edgework yet I also know that there are those that just "think" in CMYK and thinking in RGB is difficult. Here are some pointers that may be of help about Wide Gamut CMYK to help avoid some of the issues:


http://mike.russell-home.net/DigPhot...YK/Default.htm

Here is the link into CurveMiesters Tutorial section with some more specific info (same site that Tony linked in his post).

http://www.curvemeister.com/tutorial...cmyk/index.htm

Just converting to CMYK and back has a large loss of gamut as has been mentioned. Here is the visual overlay of the gamut limits:

Screenshot2011-06-08at12130PM.png


Here is a 3D image overlaying WGCMYK with Adobe RGB:

Screenshot2011-06-09at82645AM.jpg

The WGCMYK looks like it has a good overlap Adobe RGB.

However,

I tried the conversions going back and forth to the Wide Gamut CMYK Space, and while it did greatly reduce the loss of Gamut and any color shift compared to other CMYK spaces, there still was a shift that was a non-trivial amount.

Starting with an sRGB Color Space and the red color 255,0 , 0 and using a conversion to CMYK US Web Coated (SWOP) v2 the results on a round trip conversion were 207, 44, 45

Starting with 255,0,0 and converting to a wide Gamut CMYK the results on a round trip conversion were 254, 42, 11.

Less of a shift yet you end up with a slight Hue shift and reduced Saturation. This is too big a shift so I cannot recommend this as a good general approach.

I cannot vouch for the CurveMeister tool if it does a better job or not. Their website refers to this Wide Gamut CMYK space so it makes me wonder if their tools has the same issues. I cannot advise. They may have a trial version of the software to try out. Good Luck

PS - this post was leveraged from a post I did a couple months ago on NAPP

[andrewrodney]

CMYK is a highly device dependent output color space. Its sole ‘role’ is to provide CMYK values to produce some kind of print. I do hear many users say “I grew up with CMYK, its easier for me than RGB” but I don’t necessary get it. I suppose if you work with a single CMYK recipe for years, you get to know the numbers pretty well. The issue is, most every CMYK output device is vastly different from another. Even to the same device, all you have to do is alter GCR, UCR, stuff like that, and the numbers will be vastly different (think about it, UCR or GCR alter the relationships between CMY and K, ouch).

RGB when defined in what Adobe calls a working space is pretty simple in that a neutral is always an equal mix of RGB. Not the case with all RGB color spaces by a long shot. But for editing color spaces, sRGB, Adobe RGB (1998), ColorMatch, ProPhoto etc, that rule you can bank on.

For me, there are only two cases I’ll convert to CMYK. First and foremost is to produce an specific output ready document for print. And the recipe is one I’ll build or one I know is something I can bank on (Gracol 7 going to a press I KNOW is printing that exacting behavior). The other time is on a dupe to pull some channel to bring back into the RGB original for some kind of work. And I’m finding that rendering a grayscale channel as I wish, using the raw converter provides a lot more control here than trying to convert with some flavor of CMYK and hoping one of the channels is what I need. Remember, you can build grayscale appearing data from the raw and use that turned to your needs from the raw, just as you could take a single channel of a CMYK document.

[DWThomp]

Originally Posted by edgework

If you can create a good skin mask some nice refinements are available in CMYK, which you can then pull back to RGB. (I've never found skin that can't be improved by a luminosity contrast curve to the magenta channel, something that doesn't translate well in RGB).

When skin is too hot, copying the red plate, duping the image, converting to CMYK and swapping the original red for the new cyan gets around the dot gain compensation that Photoshop builds into CMYK conversions. The resulting skin is usually a bit too blue, but that can be cured with a quick curve, whereas you can try all day long to jack up the worthless cyan plate from which most information is bleached out. Then you can pull the skin back to RGB.

I've not really worked much in CMYK itself, mainly just using it to help create a mask. The luminosity contrast curve to the magenta channel sounds interesting. After duplicate the image and converting to CMYK do you ctrl-click the magenta channel to load it as a selection and apply the curve or apply a curve to magenta and change the blend mode to luminosity? After that I'm not sure how to get it back to the original image. Same with the red plate technique for the hot skin - I can get it to the point of how do I get it back to the original image.

So there are ways to take advantage of the different spaces and what they accomplish, but moving an image wholesale into CMYK if it's not required isn't one of them.

Thank you for the help and tips.

[DWThomp]

Originally Posted by andrewrodney

Remember, you can build grayscale appearing data from the raw and use that turned to your needs from the raw, just as you could take a single channel of a CMYK document.

Thank you for the good info and I'd never thought of doing this. I'll have to give it a try.

[DWThomp]

Originally Posted by John Wheeler

PS - this post was leveraged from a post I did a couple months ago on NAPP

Thanks John.

[kav]

Originally Posted by DWThomp

I asked a question in this thread (http://www.retouchpro.com/forums/pho...s-lab-raw.html) about going from RGB>Lab/CMYK>RGB and got some good answers from Robert & John. I usually stay in RGB and will change the color sampler readout to CMYK when doing some skin color corrections, but at times I have trouble adjusting when I'm working in RGB.

Be careful of your settings with this. When I process a raw file I check for detail on all channels before I do anything else. Say you start off with clipping in your red channel. Regardless of what your cmyk numbers say this will create a potential problem. The cmyk references are most useful for avoiding clipping if you're just using a generic profile. You can't really balance for while still in RGB, and it's pointless making a round trip.

[RobertAsh]

Dennis,

Thanks for the kind feedback

There are two really good reasons I've personally found to go into CMYK (besides final targeting for print):

1. Drawing out shadow detail with curves. the K channel in CMYK is mostly midtones and shadows, not much of its length is highlights. Dan Margulis' book mentioned this and one time I was stuck and had to try it. It worked. Nowadays I use Shadow/Highlights and Fill Light and get by pretty well. But that one time CMYK saved my bacon, but good (it was a wedding reception hall image shot with no flash).
2. Correcting skin tones. In situations where white balance is off it can be a bear to correct skin tones. I was in such a situation not long ago and was referred to this article. It has a nifty CMYK value chart for different ethnicities' skin tones. I had to adjust the values to be different from the ones listed in the article. The article saved my shoot by giving me a good starting piont, but its author says he works in RGB to adjust CMYK values.

So I thought -> so why not just work in CMYK then? That worked much better for me. Human skin has a lot of yellow and magenta. It doesn't have much green or blue (normally!!). Adjusting red adjusts both magenta and yellow by lock-step amounts, and you usually want to adjust them separately. So why not just adjust them directly? Edgework's note is a good summary of some similar situations he's run into.

Device dependency is a valid point. As a practical matter, however, it's good to keep in mind that all color spaces are device dependent (perhaps some more than others). Try looking at RGB images on one monitor versus another. Or in Photoshop vs. Internet Explorer. CMYK may be more so but no color space is immune from that problem. So do what meets your needs best.

It's best to stay in one color space if you can. I bought a Curvemeister license for just that purpose recently, will see if it works!

But whatever you do, don't fall into the 'color space wars' trap. Learn the strengths and weaknesses of each, both from experts and through your own application and experience. Then use the one that's best for the job at hand. Dan Margulis' book has a good summary, there are other sources as well.

[Der_W]

Did you know you can work in CMYK without leaving RGB?

It's possible to create the CMYK channels in RGB without needing to convert and without great loss (actually in most cases no loss will occur at all).
Just try this action by Sean Baker: http://sean-blog.twicebakedphoto.com...yk-in-rgb.html
And for an explanation of the steps: http://sean-blog.twicebakedphoto.com...explained.html

Hope that helps some of you :-)!

[edgework]

Originally Posted by DWThomp

I've not really worked much in CMYK itself, mainly just using it to help create a mask. The luminosity contrast curve to the magenta channel sounds interesting. After duplicate the image and converting to CMYK do you ctrl-click the magenta channel to load it as a selection and apply the curve or apply a curve to magenta and change the blend mode to luminosity? After that I'm not sure how to get it back to the original image. Same with the red plate technique for the hot skin - I can get it to the point of how do I get it back to the original image.Thank you for the help and tips.

Use a curve adjustment layer in luminosity mode. Works wonders when the tones are in a decent range but the overall shape is flat. Any further color tweaks can be done with another curve in color mode. I rely on a good skin mask to just bring the skin back to RGB. If the image is destined for print, I'd do whatever moves I need in RGB, then wait until moving into CMYK to work on skin and shadows.

As Andrew said, there's no reason to push the entire image into CMYK and lose the advantage of RGB unless it's going to press, in which case you'll lose it anyway, at some point.

One thing to keep in mind about the reduced gamut of CMYK: most items in the real world fit into it fairly well. Landscape shots, face shots and most interiors will suffer very little from the conversion. Clothing can take a heavy hit. Find an image of a dress with an intricate design built from subtle shades of greens and blues, then weep at the mud that results when you move to CMYK. At that point, you have to accept that you'll never match the color, and you try to preserve detail instead.

Also, virtually every color image you've ever seen in a book or magazine that impressed you with it's color and impact did the job with poor old CMYK inks, shoddy as they may be compared to the light from a screen. So all is not lost in the translation.

[John Wheeler]

Originally Posted by edgework

One thing to keep in mind about the reduced gamut of CMYK: most items in the real world fit into it fairly well. Landscape shots, face shots and most interiors will suffer very little from the conversion. Clothing can take a heavy hit. Find an image of a dress with an intricate design built from subtle shades of greens and blues, then weep at the mud that results when you move to CMYK. At that point, you have to accept that you'll never match the color, and you try to preserve detail instead.

Also, virtually every color image you've ever seen in a book or magazine that impressed you with it's color and impact did the job with poor old CMYK inks, shoddy as they may be compared to the light from a screen. So all is not lost in the translation.

I agree that the color of many photographs would fit within a CMYK color space such as US Web Coated (SWOP) v2 yet once you take that round-trip to a limited gamut space, the colors are gone. There are many color printers (inkjet) that have a much wider gamut than prepress CMYK as per the comparison below. The inner gamut is US Web Coated (SWOP) v2 CMYK space and the out gamut is for an ESPON 2400:

Screen shot 2011-07-28 at 6.31.54 AM.jpg

I also have a friend (landscape photographer) who prints exclusively on inkjet printers. He wants the extra punch from post processing and part of that is just tonality yet part of that is a wider color gamut. Makes a difference on his sales.

So before jumping in to truncating your color space with a round trip to CMYK, be sure you know what you lose and if that still meets your needs.

[edgework]

You can get brighter colors from a laser printer. I used to go crazy when clients would send in these ratty 11 x 17 comps with their day-glo toner colors and want to know why the Epson proof looked so dull. Even worse were the clowns who'd show up with a printed piece and say "I need to match these spot colors." I'd try to patiently explain that's why spot colors were invented... BECAUSE YOU CAN'T MATCH THEM!!! but it never did any good.

No question, keep an RGB master .psd file. But people probably should acquaint themselves with the pitfalls of CMYK; it's been around for a long time and chances are everyone will bump up against it sooner or later.

[RobertAsh]

Originally Posted by edgework

....Even worse were the clowns who'd show up with a printed piece and say "I need to match these spot colors." I'd try to patiently explain that's why spot colors were invented... BECAUSE YOU CAN'T MATCH THEM!!! but it never did any good....

What do you mean by 'that's why those colors were invented? Please explain more, that's an interesting data point.

[edgework]

The Pantone Color System. Mostly used for solid colors that can't be achieved in CMYK. Color logos for magazine covers, for example. Anything that's a bright orange, brilliant red or exotic purple, blue or green or pastel is printed with a fifth ink, specially mixed according the to Pantone formulas. Even when you can get something close with CMYK, you'd end up with halftone dots that look cheesy, whereas the unique ink is applied at 100% and so presents a smooth, continuous surface.

Sometimes you'll see a fifth ink used in photos to provide a color boost for special items: cars, clothing, etc. Usually combined with black to keep detail and shape.

[DWThomp]

Originally Posted by Der_W

It's possible to create the CMYK channels in RGB without needing to convert and without great loss (actually in most cases no loss will occur at all).

Hope that helps some of you :-)!

Thanks for those links. I installed the action, but didn't have the hsb/hsl plugin installed. I went thru the action to see what it was doing and then recorded my own to just add the CMYK & Lab channels.

Originally Posted by edgework

Use a curve adjustment layer in luminosity mode. Works wonders when the tones are in a decent range but the overall shape is flat. Any further color tweaks can be done with another curve in color mode. I rely on a good skin mask to just bring the skin back to RGB. If the image is destined for print, I'd do whatever moves I need in RGB, then wait until moving into CMYK to work on skin and shadows.

I've played around with your suggestions, but am having a time getting the changes made in the CMYK copy back into the RGB original. I'm still playing around with it, but so far I flatten the copy and bring it into the original.

I'll need to keep the end result in RGB as the lab I use requires files to be sRGB. I haven't been converting to CMYK and then converting back, but a post by Robert in the other thread gave me some questions. It seems a lot of the tutorials, etc use CMYK values and got me wondering.

Thanks again for all the help so far.

[edgework]

Originally Posted by DWThomp

I've played around with your suggestions, but am having a time getting the changes made in the CMYK copy back into the RGB original. I'm still playing around with it, but so far I flatten the copy and bring it into the original.

I'll need to keep the end result in RGB as the lab I use requires files to be sRGB. I haven't been converting to CMYK and then converting back, but a post by Robert in the other thread gave me some questions. It seems a lot of the tutorials, etc use CMYK values and got me wondering.

Thanks again for all the help so far.

You can usually use the red channel to make a good skin mask. When you dupe your file and take it into CMYK your red copy will travel with it. Use that to select the skin and just drag it back to the RGB file.

[RobertAsh]

Originally Posted by edgework

The Pantone Color System. Mostly used for solid colors that can't be achieved in CMYK. Color logos for magazine covers, for example. Anything that's a bright orange, brilliant red or exotic purple, blue or green or pastel is printed with a fifth ink, specially mixed according the to Pantone formulas. Even when you can get something close with CMYK, you'd end up with halftone dots that look cheesy, whereas the unique ink is applied at 100% and so presents a smooth, continuous surface.

Sometimes you'll see a fifth ink used in photos to provide a color boost for special items: cars, clothing, etc. Usually combined with black to keep detail and shape.

Thanks. I know what Pantone is but didn't know its origin or original purpose. Very interesting!

[RobertAsh]

Originally Posted by edgework

You can usually use the red channel to make a good skin mask. When you dupe your file and take it into CMYK your red copy will travel with it. Use that to select the skin and just drag it back to the RGB file.

This is efficient because in CMYK you usually want to either work on skin or prepare a file for printing presses.


If the skin is hard to select the way Edgework suggests, then you can

1. duplicate the image,
2. convert to profile CMYK,
3. work on the skin,
4. convert to profile <whichever flavor of > RGB,
5. drag the flattened image back into your original image and
6. mask out the parts you don't need in the original.

Edgework's idea is a lot cleaner and easier whenever you can do it his way.

[Der_W]

Originally Posted by DWThomp

Thanks for those links. I installed the action, but didn't have the hsb/hsl plugin installed. I went thru the action to see what it was doing and then recorded my own to just add the CMYK & Lab channels.

It seems you misunderstood the purpose of the action :-).
If you use this action here from the linked thread at MM: http://www.twicebakedphoto.com/download/RGBtoCMYK.atn and use the "RGB to CMYK Layers (c) 2010 Sean R Baker" action on any of your images, this will create all four CMYK channels in layers in your RGB file in a way that won't lose any data (because the image is not really converted to CMYK, it's more of a recalculation and redistribution of data).
With these layers you can then work with the advantages of CMYK without it's disadvantages and while staying in RGB.

[DWThomp]

Originally Posted by Der_W

It seems you misunderstood the purpose of the action :-).

Thank you and you're exactly right - I did . I installed the one with the hsb/hsl plugin required (I guess I could load the plugin to make it work) and got the errors when it would get to those points. I've got the actions in your link and see the difference. Thanks again.

For the action I made - Duped the original, convert to CMYK, copied the channels and moved to the original, closed copy. Duped original, converted to Lab, copied the channels and moved to original, closed copy.

[bakerser]

Originally Posted by DWThomp

For the action I made - Duped the original, convert to CMYK, copied the channels and moved to the original, closed copy. Duped original, converted to Lab, copied the channels and moved to original, closed copy.

The one Jonas linked you to does something way cooler, and won't clip color spaces on you .

[kav]

I still don't see the point of making a round trip. If I have to hand over an RGB file I do try to ensure it's cmyk friendly via a generic profile similar to the device they'll be using. I don't like to risk single channel clipping/detail loss when they make a conversion. If it's a crucial product color they have options but on most stuff it'll just end up as a mess.

[RobertAsh]

If you do Edit > Convert To Profile instead of Image > Mode > CMYK you shouldn't run into any problems like that. The point of converting to CMYK is either to work with skin tones in CMYK color space or to extract shadow detail with the convenient long shadow portion of the K curve.

It's like anything else in Photoshop, if you don't feel you need to do it, no need to bother. But if there are any times you need it or want to use it, it's there.

[kav]

There's a decent workaround for the blacks. You can duplicate an image, convert to cmyk, and duplicate the K channel back as an alpha channel. The reason you're finding skintones easier to correct there is because it doesn't produce as many irritating hue shifts. This is still an incredibly clunky solution if your file is supposed to end as an rgb file.

[RobertAsh]

Originally Posted by kav

There's a decent workaround for the blacks. You can duplicate an image, convert to cmyk, and duplicate the K channel back as an alpha channel. The reason you're finding skintones easier to correct there is because it doesn't produce as many irritating hue shifts. This is still an incredibly clunky solution if your file is supposed to end as an rgb file.

Wouldn't duplicating the K channel as an alpha channel and copying that into the RGB file subject you to the same short shadows curve that RGB has already?

What works well for me is:

• Image > Duplicate
• In the duplicate file, Edit > Convert to <your favorite CMYK>
• Do your manipulations in the duplicate file
• Edit > Convert to <your working space RGB>
• Copy the flattened file back into the RGB file

Of course it's clunky Until and unless Adobe figures out how to do color space conversions without discarding adjustment layers. That's why you only do this when you need to.

The reason for fewer irritating hue shifts in CMYK is because human skin has lots of magenta and yellow, and little or no blue and green. When you manipulate the Red curve in RGB you're manipulating both the Magenta and Yellow CMYK curves simultaneously -- with no option to manipulate Y and M individually, which is sometimes needed for good results.

[John Wheeler]

Originally Posted by RobertAsh

....What works well for me is:

• Image > Duplicate
• In the duplicate file, Edit > Convert to <your favorite CMYK>
• Do your manipulations in the duplicate file
• Edit > Convert to <your working space RGB>
• Copy the flattened file back into the RGB file

Hi Robert
I am not one that advocates jumping to a lower gamut color space and back yet I know this works for some including yourself.

Here is an alternate workflow to consider that I believe gives you the same result without producing a separate document.

• Set up your Color Settings so that the RGB working space is what you want to work in and the conversion options are also set the way you want (e.g. relative colorimetric, black point compensation etc). Set them the same you would when using Convert to Profile.
• If you have multiple Layers already in your RGB stack then create a stamped layer
• Turn that stamped layer into a Smart Object
• Open that Smart Object (if you set you Color Settings right the RGB Color Space will stay the same)
• Within the Smart Object do Edit > Convert to Profile > your favorite CMYK space
• Edit as desired in CMYK color space
• Save the Smart Object (optionally close it as well)
• Go back to the Main RGB document. The CMYK of the Smart Object will automatically be converted back to your RGB Working Space using the conversion options set in Color Settings (Identical to using Edit > Convert to Profile)
• Proceed with you Photoshop edits on top of the Smart Object

I believe this should give you exactly the same result as your step by step above yet keep everything in one document. Also given that the CMYK is in Smart Object, you can reopen the Smart Object at any point and tweak the CMYK stack as you please, save, and jump back to your Main Document with the new edits propagated into your RGB Layer Stack. Let me know if this does not produce identical results for you because it does for me.

[kav]

Originally Posted by RobertAsh

The reason for fewer irritating hue shifts in CMYK is because human skin has lots of magenta and yellow, and little or no blue and green.

This is incorrect. The skin doesn't actually contain red, green, or blue pigments. It reflects much more red than green or blue light, which really isn't ideal for digital camera sensors sadly We all work around this problem. Some of the high saturation values present in RGB are out of gamut for pretty much any cmyk profile. Since the saturation is so much more controlled it won't necessarily exacerbate hue shifts in things like skin as much. Basically the changes in saturation are more constrained.

The shadow curve thing is just that you're thinking of a gamma 1.8 based profile (at least that's how photoshop represents it) although cmyk profiles vary quite a lot. Newsprint for example has a really weird curve.

By the way what cmyk profile are you using in your tests?

[edgework]

Originally Posted by RobertAsh

When you manipulate the Red curve in RGB you're manipulating both the Magenta and Yellow CMYK curves simultaneously -- with no option to manipulate Y and M individually, which is sometimes needed for good results.

Not true at all. You don't manipulate the red curve for the magenta /yellow values, you use it for the cyan. Test it out yourself: in an RGB file set two sample points right next to each other so that they give the same numerical readout. Then set one of them to give CMYK values. You can easily see the results with curves in each channel. Darken the green curve (lower the intensity of the light) and you will see the magenta values rise accordingly. Likewise for the blue channel, which increases the yellow values. Red works on it's compliment, cyan.