Photoshop LAB Color: Chapter 6

SystemSystem Registered Users Posts: 8,186 moderator
edited October 15, 2005 in Finishing School
Chapter 6: Dan's Myths and Dangers




What Dan wants you to know.

For those who want it quick and dirty, this is Chapter 6 as the author summarized it:

The biggest problems with LAB are using it on photos for which it is not suited, from not realizing that it can produce colors that are out of gamut, and from trying to get too perfect a result in the L channel.

LAB should be used only in Photoshop. The image should be converted before moving it out of Photoshop.

As long as there is no computer-generated graphic such as a gradient, it is okay to convert from RGB to LAB and back. The graphics should be created in the final output space—RGB or CMYK.

Using the channel mixer for converting a color image to black and white usually gives better results than using the L channel. Using the L channel is a close relative to using grayscale conversion.



The myth that LAB should be your primary workspace.
    • no myth if pressed for time
      • great for adjusting quantity of photos
      • the colorspace of choice for complicated retouching

      • its strength is its weakness-curves, etc. have big impact in LAB

        • if you have an image with bright colors and wide tonal range, you don't need LAB (see Figure 6.1)

          • short range of colors and tonal range or darkness calls for LAB (see Figure 6.2)



          The myth that using the L channel is the best way to convert to black and white.
            • the L channel is a close relative of grayscale conversion

              • the author states that the best way to convert color to grayscale “is to find areas of color contrast and convert them into luminosity contrast”
                • the channel mixer is the means of accomplishing this
                • the author shows an example with Figure 6.4B (channel mixer) compared to 6.5A (L) and 6.5B (grayscale)
                • Figure 6.6 shows channel manipulation before being converted into 6.4B, the most impressive of the three methods of black and white conversion



              The myth that you can use the L channel as a mask for RGB or CMYK without using LAB.
                • Dan said that most could probably ignore this



                The danger that layers will not work the same way in LAB.
                  • certain blending modes aren't supported in LAB

                    • the wise choice is to flatten the image before converting to LAB


                    The myth that the Luminosity mode on a separate layer in RGB or CMYK works as well as LAB.
                      • the best example is in sharpening (see Figure 5.11)



                      The myth that the A channel is red-green versus magenta-green.
                        • the author shows the numbers of converting RGB red and CMYK magenta into LAB, proving by numbers that it is magenta



                        The danger of not understanding what extreme AB values can do.
                          • with out of gamut colors on your screen that you want to convert to CMYK, you need to convert to RGB first



                          The myth that converting to LAB hurts the image.
                            • No
                              • but LAB files can be altered to include colors that RGB can't keep
                              • colorspace conversions, particularly in LAB, are a bad idea for computer-generated graphics, like Photoshop gradients---convert first, then construct gradient in CMYK



                            The myth that LAB is the only way to specify known colors.
                              • LAB numbers one of safest, but may not be able to match in print

                                • at the least, you would need custom inks



                                The danger that other programs do not support LAB.
                                  • cannot save in JPEG—convert to RGB first
                                  • can use TIFF, PDF, EPS formats, but better off with the Photoshop format (.psd)



                                  The technical half.
                                    • the author goes into detail about how converting to LAB and back will not hurt the image
                                    • he also discusses using an 8 bit vs. 16 bit file

                                    Comments

                                    • ruttrutt Registered Users Posts: 6,511 Major grins
                                      edited October 14, 2005
                                      Thanks, George. This is a good summary of a difficult and technical chapter. Really this chapter is aimed at people who have worked as prepress professionals for long enough to have formed (mis)conceptions about LAB. More than 1/2 of it is a highly technical and numerical argument about why 8 bit color depth is good enough for all practical purposes and in particular can survive many conversions between LAB and RGB. Dan has been arguing this issue in various forms for years. It's interesting in a sort of nerdly way, but my take away is that Dan thinks you can get away with 8 bit color almost all the time and he should know.

                                      George has summarized the other points well. I want to add one thing. This chapter is like Ch. 2. Don't panic. You can skip it and not lose much. Ch. 7 will revisit the issue of how LAB should fit in your workflow in a much more practical and applied way.

                                      I don't want to shut down discussion of this chapter. By all means, let's nerd out. But there really isn't much technique to learn here, so don't feel too bad for if you don't play.
                                      If not now, when?
                                    • SystemSystem Registered Users Posts: 8,186 moderator
                                      edited October 14, 2005
                                      b/w conversion
                                      if you have any questions as to which direction to convert to b/w, please study Dan's little section on this subject (figures 6.4 thru 6.6)-
                                      I believe he's convincingly shown that the channel mixer gives you superior results-
                                      if you have doubts, try it out (and if you get different results, by all means show it)-
                                    • ruttrutt Registered Users Posts: 6,511 Major grins
                                      edited October 14, 2005
                                      If you really want the nuts and bolts of B&W conversion as Dan sees it, you need a different book. Professional Photoshop has a chapter called Friend and Foe in Black and White which is a must read for anyone who is serious about B&W conversion. It goes way beyond the channel mixer.
                                      If not now, when?
                                    • BenA2BenA2 Registered Users Posts: 364 Major grins
                                      edited October 15, 2005
                                      Question on using luminosity mode in RGB
                                      I haven't had a chance to buy Dan's new book, but I've been reading these threads with much interest. I'd be very interested in hearing more about why Dan says using the luminosity mode in RGB is inferior to using the L channel.

                                      For instance, I frequently use luminosity mode layers for curve adjustments. I have been under the impression that adjusting an RGB curve in luminosity layer mode was exactly the same as converting to LAB, adjusting L, and converting back to RGB, but without the hassle.

                                      Can someone expand upon why this is not the case?

                                      Thanks in advance,
                                      Ben
                                    • DavidTODavidTO Registered Users, Retired Mod Posts: 19,160 Major grins
                                      edited October 15, 2005
                                      BenA2 wrote:
                                      I haven't had a chance to buy Dan's new book, but I've been reading these threads with much interest. I'd be very interested in hearing more about why Dan says using the luminosity mode in RGB is inferior to using the L channel.

                                      For instance, I frequently use luminosity mode layers for curve adjustments. I have been under the impression that adjusting an RGB curve in luminosity layer mode was exactly the same as converting to LAB, adjusting L, and converting back to RGB, but without the hassle.

                                      Can someone expand upon why this is not the case?

                                      Thanks in advance,
                                      Ben

                                      It's not a quick answer, that I know of. Maybe Rutt can condense it better than me. In fact, I'm not going to condense the answer so much as side-step it.

                                      There are many ways to use L/Luminosity and to compare LAB to RGB.

                                      If you're sharpening, LAB and RGB will perform the same 1/2 the time, 1/4 of the time LAB will be a little better, and 1/4 of the time LAB will be much better, according to Margulis.

                                      It has to do with how RGB handles light areas. The lightest areas are by definition in RGB nearly colorless. The result is that in many instances RGB with a Luminosity sharpen will produce white halos where LAB will not.

                                      The specifics of your workflow and whether there's any real difference is far too complicated and you haven't provided enough information about what it is that you're doing. There may be no real difference between the two the way you're using RGB/Luminosity.

                                      Another thing that he says is that while in many instances L sharpening is a little better than RGB sharpening, blurring is a different matter. Blurring in RGB is much worse than LAB because RGB can't blend colors as well as LAB.

                                      Here's an example from the book (which you should definitely get a copy of). Say you want to blend a color halfway between bright green and bright red. What should you get? It would have to be some kind of yellow, since they share a yellow component. So you would end up with some kind of subdued yellow. If you average 255R 0G 0B with 0R 255G 0B the way that RGB works, you'd end up with 128R 128G 0B. In LAByou would start with 54L 81A 70B for red and 88L (79A) 81B. Average those and you get 71L 1A 75B. Test it out with the Color Palette in PS, you can change it from RGB to LAB and put in the numbers. The difference is amazing.

                                      There's tons more explaining the differences. I don't pretend to understand it all, and certainly RGB does a great job in a lot of instances and the RGB/Luminosity is powerful as well.

                                      You've got to have a pretty nerdly demeanor to want to understand why it's better. I'm just cracking the surface, but it's fascinating.

                                      You can just be slightly nerdly to enjoy applying the basics without understanding all the theory behind it.

                                      But if you want to understand the theory, you've got to read the book, I think, since it's too large a subject to say "This is why it's always better to use L than RGB/Luminosity." First of all it's not always better, although it's never worse, and secondly the reasons that it is better where it is are best explained by the master, Mr. Margulis.
                                      Moderator Emeritus
                                      Dgrin FAQ | Me | Workshops
                                    • ruttrutt Registered Users Posts: 6,511 Major grins
                                      edited October 15, 2005
                                      No need to sidestep, David. You were on the right track; almost there.

                                      The big difference between LAB and a lumonsity mode layer in RGB is that LAB can represent those impossie colors -- strongly colored but also very light or dark. These colors are displayed within the RGB gamut (they have to be) and when the image is converted to RGB they will be converted to possible RGB colors, but cannot be created with RGB curves even in a lumonsity blend layer.

                                      Practically, sunsets with LAB steepening are the most striking example I know of where LAB gives very different and much better results than you can get in RGB. In this chapter, Dan gives an example of an interior with some lamps that are very bright but also yellow.

                                      In my experience this strength of LAB is also dangerous. It's easy for novices (and even for me) to end up with undesirable color casts in the deep shadows. For this reason, I often avoid working on shadow detail in LAB and take a final trip into CMYK to get deep rich shadows without "plugging" (Dan's excellent word) the shadow details. In CMYK, I just work in the K channel, steepening the curve and sometimes sharpening, and then, if necessary, use Selective Color: Black to reduce the C,M,&Y components of the shadows. Does that make sense?
                                      If not now, when?
                                    Sign In or Register to comment.