Lab Color for B/W - Why does it work so well?

[John R Smith]

Good Morning from Cornwall, where after a chilly start the early mist burnt off, and we now have a sunny day with cloudless skies. My question –

I print only B/W photographs, never colour, and after many years of shooting film have now made the transition to a MF digital back for my Hasselblads and have spent the last two months grappling with the complexities of converting a digital colour image to monochrome. This, as you can imagine, has been a steep learning curve. After trying every monochrome strategy I could find out about, and using terrifying quantities of ink and paper in the process, I have come to the conclusion that converting the RGB image to Lab color and extracting the L channel gives me the best result as a basis for the B/W image. But exactly why this should be I don’t know. I can find no technical information on exactly how the Lab L channel is constructed, however perhaps someone here can help.

Why does the L channel from Lab color look different to a straight de-saturation, or a greyscale conversion, or a PS Channel Mixer conversion with the channels set to 33/33/33? Because it does look different – subtly so, perhaps, but nonetheless more punchy and more luminous.

Any technical information would be most welcome.

John

[Mike Arst]

I hope you receive informative replies -- I would love to hear more about this. I've recently become interested in trying black and white after a few years of shooting color alone with a digital SLR. Among the reasons I wasn't interested in black and white (though I cut my teeth on b&w printing and once loved to work in the darkroom): I have never had an inkjet printer capable of producing decent black and white prints, and I don't know if I ever will. Printers of that caliber must cost a small fortune. The thought of being able to make only sub-par b&w prints has not been appealing...

Still, it's possible to produce b&w images that display well on the web. Perhaps, I thought, I could at least please myself to that extent.

Recently I ran across a technique that uses the "L" channel -- something worked out, I hear, by a digital-imaging expert who works with John Paul Caponigro. Perhaps it's the same one you've been using. I tried it and was  surprised by how well it worked (I had never been impressed by the channel-mixer approach). My first attempt produced some posterization in the image -- probably because my test image was only an 8-bit TIFF -- but it was clear the technique has a lot of promise.

This was just after I had bought a set of plug-ins called PowerRetouche, two of which perform black and white conversions. It's clear to me that they can do a pretty good job, although it's going to take a while to master them. That aside, the "L" channel approach worked surprisingly well.

(The Photoshop plug-in called Convert to Black And White Pro was widely praised as the best of the lot, but its author stopped developing or supporting it and as far as I know it's no longer possible to get a registered [non-demo] copy.)

[crames]

... I have come to the conclusion that converting the RGB image to Lab color and extracting the L channel gives me the best result as a basis for the B/W image.

Hi John,

Lab is an implementation of CIELAB, where L is called L* (L-star) and  represents Lightness. Any technical information you find about CIELAB applies.

Lightness is defined as the perceived brightness of a color relative to the brightness of white. L or L* is perceptually uniform, meaning that differences in L correspond to perceived differences in Lightness. So the L channel (accurately) represents the relative brightness of colors in an image.
 
When you say "extracting the L channel", how do you do it exactly? If you are separating the L channel from the a and b channels, either by deleting the a & b channels or by copy/pasting the L channel into an RGB channel, then that L channel no longer represents Lightness, and is interpreted by PS according to the RGB space in effect, or its Gray: setting in Color Settings. The effect of this will in most cases result in a visible change of contrast, depending on the difference between the tone curve of L, which is about gamma=3, and the tone curve of the destination RGB space or gray-scale space. This might explain the difference in punchiness and luminousness you are seeing.

If instead you wanted actually to preserve the L channel as Lightness, then desaturate in Lab mode before converting to the desired RGB or gray-scale space.

Regards,
Cliff

[John R Smith]

Mike

I tried many printers, and like you was very underwhelmed with the results in B/W. However, for the last three years I have been printing scanned B/W film to my Epson R2400 and I am very happy indeed. I process the files entirely in 16-bit greyscale, resample to print size and 360 dpi as a final step before USM, and print using the Epson Advanced Black and White mode on Harman FB gloss. The results are indistinguishable from my darkroom prints on the old Ilford Galerie paper, except for a little bit of gloss differential. I know this, because I have scanned 25 year old MF negs for which I already had prints, printed to the same size and compared the outputs very critically.

So I would have thought that any Epson printer which has the ABW mode would do the same for you.

John

[John R Smith]

When you say "extracting the L channel", how do you do it exactly? If you are separating the L channel from the a and b channels, either by deleting the a & b channels or by copy/pasting the L channel into an RGB channel, then that L channel no longer represents Lightness, and is interpreted by PS according to the RGB space in effect, or its Gray: setting in Color Settings. The effect of this will in most cases result in a visible change of contrast, depending on the difference between the tone curve of L, which is about gamma=3, and the tone curve of the destination RGB space or gray-scale space. This might explain the difference in punchiness and luminousness you are seeing.

Cliff

In PS(7) I convert to Lab from RGB (all 16-bit), select the L-channel from the channels palette, go to Mode > Greyscale, and click OK to "Discard Other Channels?". I can observe no visible difference on screen or in the histogram in the L channel before and after this process.

John

[crames]

In PS(7) I convert to Lab from RGB (all 16-bit), select the L-channel from the channels palette, go to Mode > Greyscale, and click OK to "Discard Other Channels?". I can observe no visible difference on screen or in the histogram in the L channel before and after this process.

John,

I believe that by selecting the L channel before converting to Grayscale, the L channel is interpreted according to the Gray: setting in Color Settings. In other words, your results depend on whatever the Gray: setting is.

If you were to convert to Grayscale without selecting the L channel (convert while viewing all channels = full-color version), actual Lightness will be preserved. (I'm not saying you should preserve Lightness, just that the results will be different)

To duplicate your results, what is your Gray: setting in Color Settings?

Cliff

[Mark D Segal]

There are scads of resources on the internet and in books about converting to B&W and everyone has their favorite techniques ranging from the most simple to the most convoluted. I've been doing this for quite a while with quite a few images and would recommend the following. But for starters, there is nothing particularly complex about what you need to do to get first-class results, so the keep-it-simple philosophy should serve you well.

For medium format digital captures, if Capture 1 can read your raw files, use Capture-1 and convert to B&W non-destructively using the controls offered there. It produces excellent results.

If Capture-1 cannot handle your files but Lightroom/Camera Raw can, use the Grayscale option in the HSL panel (LR) to make a non-destructive well-controlled B&W rendition.

If you wish to work on rendered PSD or TIFF files, use the B&W adjustment layer in Photoshop, as it works very much like its counterpart in LR/ACR.

If you wish to use a third-party plug-in which provides a raft of looks, effects and localized controls over luminosity in B&W, there is nothing better on the market these days than Nik Silver EFEX Pro, which is also non-destructive because it works on its own layer.

Forget converting images to Lab. It makes for more work than necesary, is not non-destructive per se, and compared with contemporary techniques mentioned above has no value-added.

[John R Smith]

Cliff

Gray = Gray Gamma 2.2. However, in "Color Management Policies", Gray is Off. So that might make a difference, I will have to try it out.

I cannot de-saturate the file (either RGB or Lab) in 16-bit using PS7. It will only desat in 8-bit.

Mark

Thank you for your input. I use the Hasselblad back, so I have Phocus as the RAW converter. That has a B/W option. I have also used the Channel Mixer in PS Elements 6, and various PS plugins I got off the net. Then I tried using ACR and de-saturating there. I have tried all of these, done gazillions of test prints, and given my particular combination of camera/lenses/desires/software/and printer the best result is (to my eyes, and I have been shooting B/W for fifty years) the Lab 'L' channel. I don't know why, but I know what I like. My interest is how the Lab lightness algorithm works compared to a straight desat.

PS These differences are not usually very visible on-screen (or at least not on my screen), but are on my work prints. Just as in the dark room days . . .

John

[crames]

...I cannot de-saturate the file (either RGB or Lab) in 16-bit using PS7. It will only desat in 8-bit.

Hmm, PS7 does limit the things you can do in 16 bit, but I don't remember which. Maybe try setting  saturation=0 in Hue/Sat, or fill the a and b channels with zero. But  this will give you something different from what you are currently  satisfied with. I think what you're getting is Lightness interpreted in a gamma 2.2 space, which seems to give a small boost to the mid-tones.

Cliff

[Mike Arst]

John --

The technique about which I received the e-mail is discussed here: http://www.designbyfire.com/?p=17 -- with the discussion starting partway through the page. I wish it were a bit more readable (it's odd when designers make web pages with near-unreadable typography...I wonder if they propose similar book designs to their clients...fortunately, browsers' type-size controls to the rescue).

Interesting to hear that the R2400 is capable of doing such good b&w printing. I have the R1800. Close, but not close enough. I can well imagine that the Harman Gloss is a good paper for this purpose, and your finding those prints indistinguishable from darkroom prints on Galerie is a pretty encouraging recommendation. The Harman is certainly good for color. I would have been happy to use Ilford's similar inkjet-paper offering (at least for color; I found it producing somewhat stronger reds and oranges than the Harman), but it has a heavy 'reverse' curl that makes it unusable in the R1800.

Do you have any of your black and white work displayed on the web?

Mike

I tried many printers, and like you was very underwhelmed with the results in B/W. However, for the last three years I have been printing scanned B/W film to my Epson R2400 and I am very happy indeed. I process the files entirely in 16-bit greyscale, resample to print size and 360 dpi as a final step before USM, and print using the Epson Advanced Black and White mode on Harman FB gloss. The results are indistinguishable from my darkroom prints on the old Ilford Galerie paper, except for a little bit of gloss differential. I know this, because I have scanned 25 year old MF negs for which I already had prints, printed to the same size and compared the outputs very critically.

So I would have thought that any Epson printer which has the ABW mode would do the same for you.

John

[BernardLanguillier]

If you wish to use a third-party plug-in which provides a raft of looks, effects and localized controls over luminosity in B&W, there is nothing better on the market these days than Nik Silver EFEX Pro, which is also non-destructive because it works on its own layer.

I would second that.

The images in the linked page below did undergo various manipulations, but Silver efex was part of the workflow for all of them.

http://www.flickr.com/photos/bernardlangui...57623305594697/

As far as print goes, I have had execllent results on my Epson 9900 with the Epson driver B&W mode.

Cheers,
Bernard

[Mark D Segal]

Hi Bernard,

As usual - excellent photographs.

I'm using an Epson 3800 and also get clean, neutral B&W results from the combo of Photoshop with the Epson driver, on Ilford Gold Fibre Silk paper.

Cheers,

Mark

[John R Smith]

John --

Interesting to hear that the R2400 is capable of doing such good b&w printing. I have the R1800. Close, but not close enough. I can well imagine that the Harman Gloss is a good paper for this purpose, and your finding those prints indistinguishable from darkroom prints on Galerie is a pretty encouraging recommendation. The Harman is certainly good for color. I would have been happy to use Ilford's similar inkjet-paper offering (at least for color; I found it producing somewhat stronger reds and oranges than the Harman), but it has a heavy 'reverse' curl that makes it unusable in the R1800.

Do you have any of your black and white work displayed on the web?

Mike

The 1800 and the 2400 are virtually identical mechanically. However, they are very different in output. I think you would find that there would be a fairly general agreement that the 1800 is best for colour, and the 2400 is far superior for B/W but not so great for colour work. I don't have a website, or any work on the web, because - this is my hobby, not my job (I do quite a lot of photography for work, but that is now all colour digital), and I do B/W for printing on paper, not to look at on screen. And I'm not that great a photographer, anyhow. Still, I could try this attachment thing and see what happens. Hmm, looks as if I may have screwed something up . . . Oh, no, it's worked. Well, there you go, that's a quite nice fern which I have been tending for years on my window seat. Hasselblad 500 C/M, 120mm S-Planar, CFV-39 back.

John

[Mike Arst]

John -- thanks for including this sample. Ah, the legendary smoothness of the Hasselblad digital.

I have not been paying attention to the printer market for quite a while and didn't realize until yesterday that Epson has discontinued both the R1800 and R2400 (at least according to their U.S. web site). I now contemplate the likely decreasing availability of those inks (in local stores). Do I have, in effect, a printer-shaped boat anchor, worth a few dollars on eBay not long from now? I didn't find printers comparable to those two on their site. The "low end" printer now seems to be something in the 3xxx series. Perhaps I just haven't searched enough. But if I'm right...what a depressing development. They had the 1270 in their product line for quite a few years. They appear to have changed their approach to "product life." My interest in trying to print b&w is rekindled. But if Epson is going to release printers with ever-shorter life-cycles (and ink types and ink cartridges specific to each model and orphaned by its demise), I don't know if I dare even buy another one of their products...

Mike

The 1800 and the 2400 are virtually identical mechanically. However, they are very different in output. I think you would find that there would be a fairly general agreement that the 1800 is best for colour, and the 2400 is far superior for B/W but not so great for colour work. I don't have a website, or any work on the web, because - this is my hobby, not my job (I do quite a lot of photography for work, but that is now all colour digital), and I do B/W for printing on paper, not to look at on screen. And I'm not that great a photographer, anyhow. Still, I could try this attachment thing and see what happens. Hmm, looks as if I may have screwed something up . . . Oh, no, it's worked. Well, there you go, that's a quite nice fern which I have been tending for years on my window seat. Hasselblad 500 C/M, 120mm S-Planar, CFV-39 back.

John

[Mark D Segal]

John -- thanks for including this sample. Ah, the legendary smoothness of the Hasselblad digital.

I have not been paying attention to the printer market for quite a while and didn't realize until yesterday that Epson has discontinued both the R1800 and R2400 (at least according to their U.S. web site). I now contemplate the likely decreasing availability of those inks (in local stores). Do I have, in effect, a printer-shaped boat anchor, worth a few dollars on eBay not long from now? I didn't find printers comparable to those two on their site. The "low end" printer now seems to be something in the 3xxx series. Perhaps I just haven't searched enough. But if I'm right...what a depressing development. They had the 1270 in their product line for quite a few years. They appear to have changed their approach to "product life." My interest in trying to print b&w is rekindled. But if Epson is going to release printers with ever-shorter life-cycles (and ink types and ink cartridges specific to each model and orphaned by its demise), I don't know if I dare even buy another one of their products...

Using North American values for the printers and Epson inks, you may not be aware of it, but an Epson 3880 pays for itself over an Epson 2880 (replacements for the 3800 and 2400) once you've made about 400 A4-sized prints, because the ink is so much cheaper on account of the larger cartridges. Not to speak of the quality improvements in gamut, print-head design and B&W rendering. So this is anything but a depressing development - it's real progress. As for life-cycle, the Epson 3800 had a life-cycle of AT LEAST three years, and for many folks will be longer, because eventhough the 3880 is improved over the 3800, perhaps not enough to justify that upgrade. But an up-grade from a 2400 to a 3880 is DEFINITELY a HIGHLY RECOMMENDED progression for anyone who expects to make more than 400 prints over the life of the printer.

[MarkIV]

I have come to the conclusion that converting the RGB image to Lab color and extracting the L channel gives me the best result as a basis for the B/W image. But exactly why this should be I don’t know. I can find no technical information on exactly how the Lab L channel is constructed, however perhaps someone here can help. Any technical information would be most welcome.

John

I really would not come to the same conclusion.  

If you want the most powerful flexibility in B+W's (and options) I would set up an action that copies and pastes the RGB Red, Green, and Blue channels, as well as the LAB L channel all as 4 separate layers (all named and the color version trashed).  Then you can simply click on and off the eyeballs on the layers pallet to see the separate qualities of all 4.  I have not found the L channel to be the best, but every image completely different.  Sometimes it is the Red Channel, sometimes the L, sometimes the Green, sometimes even the Blue...  This is an image specific thing.  Then once you identify the particular qualities you prefer in each you can simply, and easily blend them together with different layer opacities (and you can use the Layer Blend Modes, and the Layer Style "Blend If" sliders) for an almost infinite amount of different interpretations and total control of all tones (kind of like Ansel Adam's zone system but on steroids).  Once you have mastered the opasity blend you prefer, you then can easily finish off by doing additional contrast work, or whatever...

[PeterAit]

Hi Bernard,

As usual - excellent photographs.

I'm using an Epson 3800 and also get clean, neutral B&W results from the combo of Photoshop with the Epson driver, on Ilford Gold Fibre Silk paper.

Cheers,

Mark

Are you printing a color image using the Epson B&W mode? Or, do you convert the image to B&W first and then print?

[Mark D Segal]

Are you printing a color image using the Epson B&W mode? Or, do you convert the image to B&W first and then print?

Hi Peter - first, I just visited your website and you have some excellent work there. Congratulations.

I don't use the ABW driver, because I don't have control over soft-proofing with it, though Eric Chan has produced a viable workaround for that with his ABW profiles. If you haven't done so, you may find it interesting to visit Eric's Epson 3800 page.

I haven't gotten into the ABW driver with Eric's profiles, because I've been very satisfied with the methods I'm using, and posted in post #7 above.

[PeterAit]

Hi Peter - first, I just visited your website and you have some excellent work there. Congratulations.

I don't use the ABW driver, because I don't have control over soft-proofing with it, though Eric Chan has produced a viable workaround for that with his ABW profiles. If you haven't done so, you may find it interesting to visit Eric's Epson 3800 page.

I haven't gotten into the ABW driver with Eric's profiles, because I've been very satisfied with the methods I'm using, and posted in post #7 above.

Thanks, Mark, for the suggestions and comments.

[Mike Arst]

I was wrong earlier when I described a B&W conversion technique as being provided by Caponigro. This is the one provided by Caponigro: http://www.adobe.com/digitalimag/ps_pro_primers.html -- see the link "Black and White Conversion Tutorial" (PDF file).

The technique involves copying and pasting the individual R, G and B channels into the main document, creating a new layer each time you paste. I've tried it a number of times in Photoshop CS2 and haven't been able to get it to work. With each "paste," the new layer that appears in the document contains the data only for the very first channel that was copied. That is, if I first paste the Red channel, at the point when I've copied and pasted the Blue channel, it's still only the red-channel data that appears in the new layer. The same happens when I copy and paste the green-channel data: again I get only the red-channel data in the new layer.

I've been following the instructions slowly and meticulously -- or so I thought -- but clearly I'm doing something wrong. Can anyone reading this message think what it might be? Or, is it possible there's something missing from those instructions?

The web page contains a Photoshop action that's said to automate the sequence of events described in the tutorial. That isn't quite correct. The action does considerably more than is described in the tutorial. When it is run, the channels when pasted into the main document DO look the way I would expect them to look.

(I realize that Silver Efex Pro is a mere $200. Despite the bargain pricing, I would nevertheless like to find some other workable approach, if possible. To date, Convert to B&W Pro having been discontinued, I have not yet found a decent plug-in for this purpose...)