Announcement

Collapse
No announcement yet.

Is Lab color (L*a*b*) still relevant?

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • pixaeiro
    replied
    Re: Is Lab color (L*a*b*) still relevant?

    Originally posted by klev
    I'm pleased that someone used the correct number there (mantissa only rather than 2^32).
    Ha Ha, yep... I am a programmer. Programming in float is way easier than uint8.
    Originally posted by joe_7
    The main uses I've found for L*a*b are being able to sharpen the luminosity channel by itself, or blur the color information slightly. (The latter is great for big enlargements to make everything look smoother.)
    Yes, same me... but that is a very useful trick... and if you scale down the a and b channels from the center, you can also increase the saturation of the color...

    Leave a comment:


  • joe_7
    replied
    Re: Is Lab color (L*a*b*) still relevant?

    The main uses I've found for L*a*b are being able to sharpen the luminosity channel by itself, or blur the color information slightly. (The latter is great for big enlargements to make everything look smoother.)

    Beyond that I never got much more into it though, and use it only as a final step.

    Leave a comment:


  • andrewrodney
    replied
    Re: Is Lab color (L*a*b*) still relevant?

    Originally posted by Benny Profane View Post
    Does it have anything to do with a LUT?
    It could. But why attempt to explain this to someone again hijacking a thread and asking questions he has no intention of trying to understand. Is there a reason you come to such posts and add nothing useful to the conversation like your above question? Try using 'the Google'.

    Leave a comment:


  • Benny Profane
    replied
    Re: Is Lab color (L*a*b*) still relevant?

    Does it have anything to do with a LUT?

    Leave a comment:


  • klev
    replied
    Re: Is Lab color (L*a*b*) still relevant?

    Originally posted by Benny Profane View Post
    Wait.


    What?


    Quantization??
    It's just a term in image processing that refers to reproducing something in a set of discrete steps. It's a slightly odd term to use there, because it comes from signal reconstruction where you have to fit a finite set of points to a continuous change in voltage. In this case we were referring to errors in assigning a discrete set of values to a given mapping, even though it didn't explicitly refer to signal reconstruction. It's still similar when you convert between color spaces, as it's done most commonly via linear transformation. It sounds like something weird, but it just means that they're only dealing with multiples of the original values. The new ones aren't dependent on the squares or higher powers of the old ones, which makes it a bit simpler.

    Leave a comment:


  • Benny Profane
    replied
    Re: Is Lab color (L*a*b*) still relevant?

    Originally posted by klev View Post
    Quantization errors always exist in some way when dealing with numerical computations, assuming you permit any operations beyond addition and subtraction.
    Wait.


    What?


    Quantization??

    Leave a comment:


  • klev
    replied
    Re: Is Lab color (L*a*b*) still relevant?

    Originally posted by pixaeiro View Post
    Thanks klev... I haven't done the math yet, but I think that if you take a 32 bits per channel float RGB, and convert it to 32 bpcf Lab image, and then back to 32 bpcf RGB image, you won't loose as much as 10% to 12% of the color information because the rounding error of float math is much lower than the rounding error of 8 bit integer math (1/8388608 vs 1/256)... but yes, you are correct... there will always be a penalty to pay for going from one color mode to another.
    Well that's true, but it's pretty much what I said. There's always a penalty. I didn't say it was huge, and I've been testing it recently in octave. I don't know that a hit in perceived accuracy is that problematic. A bigger irritation is when visually smooth regions turn into piecewise smooth ones. I'm pleased that someone used the correct number there (mantissa only rather than 2^32).

    Leave a comment:


  • andrewrodney
    replied
    Re: Is Lab color (L*a*b*) still relevant?

    Originally posted by pixaeiro View Post
    I always find more intuitive to adjust the Lightness curve in Lab mode than in RGB mode, because the colors never change to unnatural values...
    The Luminosity blend mode or fade is your friend in that respect.

    Leave a comment:


  • andrewrodney
    replied
    Re: Is Lab color (L*a*b*) still relevant?

    Originally posted by pixaeiro View Post
    Andrew, are these quantization errors still a problem in applications that work in 32 bits per channel mode?
    What Klev said

    Leave a comment:


  • pixaeiro
    replied
    Re: Is Lab color (L*a*b*) still relevant?

    Originally posted by klev
    Quantization errors always exist in some way when dealing with numerical computations
    Thanks klev... I haven't done the math yet, but I think that if you take a 32 bits per channel float RGB, and convert it to 32 bpcf Lab image, and then back to 32 bpcf RGB image, you won't loose as much as 10% to 12% of the color information because the rounding error of float math is much lower than the rounding error of 8 bit integer math (1/8388608 vs 1/256)... but yes, you are correct... there will always be a penalty to pay for going from one color mode to another.

    Originally posted by skoobey
    Hm.. I still don't find it useful to work in LAB over RGB. I like the 3 channels(6 colors). And I just set adjustment layers to "color" if I want to avoid shifting perceived lightness.
    Working in Lab just gives you more tools to work with... operations like curves, gaussian blur or unsharp mask have a completely different result in Lab mode. I always find more intuitive to adjust the Lightness curve in Lab mode than in RGB mode, because the colors never change to unnatural values...

    Leave a comment:


  • skoobey
    replied
    Re: Is Lab color (L*a*b*) still relevant?

    Hm.. I still don't find it useful to work in LAB over RGB. I like the 3 channels(6 colors). And I just set adjustment layers to "color" if I want to avoid shifting perceived lightness.

    Leave a comment:


  • klev
    replied
    Re: Is Lab color (L*a*b*) still relevant?

    Originally posted by pixaeiro View Post
    Andrew, are these quantization errors still a problem in applications that work in 32 bits per channel mode?
    (not Andrew but anyway.....) Quantization errors always exist in some way when dealing with numerical computations, assuming you permit any operations beyond addition and subtraction. It's also important to note that 32 bit per channel mode uses floating point math, which is rounded by its nature to the nearest machine number after each computation. I won't bore you with more of that. An upside to 32 bits is that it creates far less haloing on gaussian blur and sharpening operations. Apart from that just remember that there is always rounding.

    Leave a comment:


  • pixaeiro
    replied
    Re: Is Lab color (L*a*b*) still relevant?

    Originally posted by andrewrodney
    Other than the issues of Lab's shortcomings (already mentioned), every time a conversion to LAB is produced, the rounding errors and severe gamut mismatch between the two spaces can account for data loss, known as quantization errors.
    Andrew, are these quantization errors still a problem in applications that work in 32 bits per channel mode?

    Leave a comment:


  • skoobey
    replied
    Re: Is Lab color (L*a*b*) still relevant?

    Originally posted by Doug.S View Post
    Another way to "work" in Lab or RGB or HSB etc., without actually converting image color profile, is to use CurveMeister plugin. It has many very useful tools...including ability to use full screen second monitor to fine tune a curve (I use second monitor in portrait mode).
    An excellent toolset for any retouching/correction work...including a "skin" tone mask and ability to "pin" colors.
    Another new toolset is Lumenzia.....for making and working with masks.
    That seems like a great tool.

    Leave a comment:


  • Doug.S
    replied
    Re: Is Lab color (L*a*b*) still relevant?

    Another way to "work" in Lab or RGB or HSB etc., without actually converting image color profile, is to use CurveMeister plugin. It has many very useful tools...including ability to use full screen second monitor to fine tune a curve (I use second monitor in portrait mode).
    An excellent toolset for any retouching/correction work...including a "skin" tone mask and ability to "pin" colors.
    Another new toolset is Lumenzia.....for making and working with masks.

    Leave a comment:

Related Topics

Collapse

  • kkamin
    Wrapping your head around the L*a*b* histogram
    by kkamin
    It's Lab day at the forum apparently. : )

    In the RGB color mode, I understand what the histogram is telling me for the 3 color channels. I can adjust it with curves or levels with comprehension.

    The Lab color channels, a & b, are confusing me greatly. All the histograms...
    09-06-2010, 06:26 PM
  • kkamin
    L*a*b* question: why can't it be simulated in RGB?
    by kkamin
    The big benefit I see of Lab is that the luminance is not tied to the color. So you can adjust saturation without clipping the colors as fast. Or darken an image without desaturating the colors as fast.

    But why can't adjustments be made in RGB with blending modes such as "color"...
    09-06-2010, 05:58 PM
  • bart_hickman
    LAB versus HSL
    by bart_hickman
    I see a lot of folks convert to the LAB color space (and back again) and you most often say you're doing it so you can edit luminance and color informaton separately.

    My question is, why not use the HSL color space instead? It seems more intuitive. Photoshop has the Hue, Saturation...
    03-10-2006, 07:30 PM
  • Toriat100
    LAB mode and retouching advantages..?
    by Toriat100
    Could any one explain LAB mode to me and its advantages for retouching/restoration?

    Many thanks
    01-24-2008, 05:53 AM
  • Chacaboy
    Questions about LAB mode
    by Chacaboy
    Hi,

    I am a new member as of yesterday. I teach Photoshop and Lightroom, college level, and I would like some practical suggestions about how to demo useful or creative methods using LAB mode. I understand how it is structured, and that the lightness or luminosity is isolated from the...
    08-24-2016, 04:05 PM
Working...
X