RGB->LAB->RGB

[Mark McIntyre]

Someone brought this up in a photography website and I thought it a good question.
I've always read that switching from RGB to LAB and then back to RGB does not alter the color space. But it does. If you fill a doc with 0 255 0, switch to LAB, and then back to RGB, there are changes in either both RGB and CMYK values or just one of them, depending on what color profile you use.

Is this correct??

[Carl_Stawicki@adobeforums.com]

With the color values you mentioned (I'm assuming you meant 0R 255G 0B), the RGB levels are effected, but not the Lab's of those colors. The Lab values of that shade of green, while in AdobeRGB, are 83L -128a 87b before and after converting. The RGB numbers end up being 21R 254G 3B. So the levels change, but not the color.

Carl.

[Sylvain_Gingras@adobeforums.com]

When you change mode, I think it is best advise to do it through convert to profile, so that you can control to the way the conversion is done. But each time you convert a file there is some loss of data from rounding of numbers.

Can anyone confim?

[Mike_Ornellas@adobeforums.com]

There may be rounding errors during the color space conversions. To assure little or no errors, use 16 bit files.

[John_Slate@adobeforums.com]

By what version could you go to Lab in 16bit?

So the levels change, but not the color




Same color space, different numbers = different color

[Carl_Stawicki@adobeforums.com]

Different combinations of levels within the same color space can have the same color. By 'color', I mean perceptual colors, as indicated by Lab values. Levels and colors are two different things.

Carl.

[Mike_Ornellas@adobeforums.com]

John,

Photoshop CS supports 16 bit LAB.

[Mark McIntyre]

Thanks.

But I''m still confused.

Does the color degrade or not?

Many authors have encouraged switching to LAB mode to sharpen the L channel, and then back to RGB or CMYK.

Is this a problem for color shifting??

[Bruce_Fraser@adobeforums.com]

Yes, the conversion from RGB to Lab to RGB is subject to rounding errors.

If you convert using whole numbers between Fahrenheit and Celsius, 53° Fahrenheit and 54° Fahrenheit both convert to 12° Celsius.

8-bit/channel files only have 256 data points per channel. In a small RGB working space, those points are packed closer together than in a larger one. Lab, by definition, covers the entire range of visible color, so the data points are spaced further apart than in just about any RGB space. Hence when you convert, you get rounding errors.

Whether or not these errors are visually significant depends upon

the image
the rgb space
what you do to the image downstream

You can get a result meaninglessly close to sharpening the lab channel by either

1.) Sharpening RGB, then using Fade with the blend mode set to Luminosity, or

2.) duplicating/merging to a new RGB layer, sharpening it, then setting the sharpend layer's blending mode to Luminosity.

Either one will avoid the rounding errors.

If you work with 16 bits per channel, the rounding error will almost certainly be insignificant, but converting to Lab is still a lost slower than either of the above alternatives, and if your file has adjustment layers, you'll have to either flatten before conversion or suffer significant color shifts because the layers behave differently in Lab and RGB.

[JasonSmith@adobeforums.com]

I actually use #2 sharpening method (darn I thought I invented that).

[Ramón_G_Castañeda@adobeforums.com]

Many authors have encouraged switching to LAB mode to sharpen the L channel,
and then back to RGB or CMYK.




Yes, some do; but I've found it to be a dumb idea. Just loook at the color gamut in Lab and in RGB. Not hard to see that there has to be a degradation when you switch in and out of Lab.

Bruce mentions two alternative ways of doing it, but I actually prefer a different, much better route suggested in --surprise!-- Bruce Fraser's "Real World Photoshop 7". It involves getting rid of noise in the blue channel and applying different degrees of Unsharp Masking and Fading to the red and green channels separately.

Thanks, Bruce.

[Carl_Stawicki@adobeforums.com]

#2 has been my choice for quite awhile now. It's very versatile.

[Ann_Shelbourne@adobeforums.com]

Even better: Paint in the USM with the History Brush using varying opacities and Luminosity mode.

[Mark McIntyre]

Thanks for all the good feedback. I've left the same question with Dan Margulis's color theory group as well.

Have used all three methods recommended, but good to hear what others prefer.

Gaining clarity on the LAB issue.

Thanks again.

[LRK@adobeforums.com]

Where is Dan Margulis's color theory group? Can't possibly be as interesting as Bruce Fraser though. :)

[Mark McIntyre]

<http://groups.yahoo.com/group/colortheory/>

[Mark McIntyre]

Ann, great idea.

[Tim_Lookingbill@adobeforums.com]

I tried the sharpen individual red and green channel technique but always got tiny rainbow color affects (or chromatic aberations as some call it) on hard edges even in a tiff file. Jpeg was even worse.

[Ann_Shelbourne@adobeforums.com]

Tim:

I have found the same thing. I don't like the sharpen-in-color channels method -- UNLESS it is in Lab's "L" channel.

[Bruce_Fraser@adobeforums.com]

FWIW, the sharpening individual channels technique was developed when blue channels were a great deal noisier than they are today—I think the example came from a Polaroid PDC 1200, if anyone remembers those. It has been redacted from Real World Adobe Photoshop CS in favor of techniques that are less of a blunt instrument, but hey, if it works for someone...