Histogram deteriorating with conversion to Lab in PS

[Guillermo Luijk]

Since I had never heard the word 'pundit' I searched for it and found "‘Pundit‘ means an expert or knowledgeable person, an authority with opinion."

Are you talking about someone (don't remember his name) who had a loooooong discusion with Bruce Lindblom about 16 vs 8 bit image edition?

BR

[digitaldog]

Since I had never heard the word 'pundit' I searched for it and found "‘Pundit‘ means an expert or knowledgeable person, an authority with opinion."

Are you talking about someone (don't remember his name) who had a loooooong discusion with Bruce Lindblom about 16 vs 8 bit image edition?

Yes. I should have said, in this context, anti-pundit or something along those lines. That said, it would be totally irresponsible of me to say this fellow isn't an expert or very knowledgeable. We have major disagreements on certain aspects of imaging, but there's no question, there's value added opinions and techniques presented by this person. That's what makes his ideas about Raw processing, high bit, wide gamut imaging so distasteful. He'd be easier to ignore if he didn't have anything useful to say. That's not the case. So lets say, he's a sudo-pundit.

[pedro.silva]

So lets say, he's a sudo-pundit.

you mean, "sudo" as in "superuser"?  

[bjanes]

Since I had never heard the word 'pundit' I searched for it and found "‘Pundit‘ means an expert or knowledgeable person, an authority with opinion."

Are you talking about someone (don't remember his name) who had a loooooong discusion with Bruce Lindblom about 16 vs 8 bit image edition?

BR

For more background, go to Lindbloom's site and click on the info tab and then scroll down to the "Dan Margulies 16 bit challenge"
Bruce Lindbloom Site

And here is some more background:
Reconsidering 16 bits

Or it might be more profitable to do 15 iterations of the round trip with 8 bit files and repeat the analysis.

Bill

[digitaldog]

For more background, go to Lindbloom's site and click on the info tab and then scroll down to the "Dan Margulies 16 bit challenge"
Bruce Lindbloom Site
And here is some more background:
Reconsidering 16 bits

Its a tired debate. Are we supposed to believe that there's no  benefit to high bit workflows, capture devices that capture more than 8 bits, printer drives that send more than 8 bits to the printer, software that supports high bit editing but rather, its all a conspiracy and unnecessary? The "pundit" is just down on high bit (and high gamut) workflows to point a light onto himself. Its the Ken Rockwell mode of  generating attention by being controversial.

BTW, Lindbloom nailed it here (direct link):http://www.brucelindbloom.com/index.html?DanMargulis.html.

[Guillermo Luijk]

Wow, so the pundit was Margulis?!? pity it's already 3am in Spain and that means my time to go to bed to enjoy some (printed) reading is nearing. Tomorrow I'll have a look at all that old but interesting stuff.

Thanks.

[JBerardi]

"‘Pundit‘ means an expert or knowledgeable person, an authority with opinion."

Well there's a dated definition if I've ever heard one... "generating attention by being controversial" is the main goal of the modern pundit.[/thread drift]

[bjanes]

Its a tired debate. Are we supposed to believe that there's no  benefit to high bit workflows, capture devices that capture more than 8 bits, printer drives that send more than 8 bits to the printer, software that supports high bit editing but rather, its all a conspiracy and unnecessary? The "pundit" is just down on high bit (and high gamut) workflows to point a light onto himself. Its the Ken Rockwell mode of  generating attention by being controversial.

Andrew,

I repeated the experiment with the same image using 8 bit ProPhotoRGB and then 16 bit ProPhotoRGB and compared the results using the method you suggested. In my first attempt, due to unfamiliarity with the method, I applied the image to itself!

Here is the image:


Here is the difference with 16 bit processing. There are a few differences in the shadow areas.


and with 8 bit processing, there is abundant posterization.


How much these differences would show up in a print is yet to be determined, but the differences are quite striking in the analysis.

Bill

[pfigen]

"The testing in this thread is actually making me a little more interested in using messing around with LAB color. Funny how things work sometimes."

Then I highly recommend Dan's Lab book, The Canyon Connundrum. As well as leading you through the ins and outs of Lab, it will also help you learn when Lab is appropriate and when it's not - and it's definitely not always a benefit. Knowing when makes you a better retoucher.

As far as Dan being totally down on hi-bit high gamut workflows, this is somewhat disengenuous. He's not and even advocates that for most processes it won't hurt. What he objects to, and this is my opinion, is that there is a certain constituent that claims it's always beneficial to work in 16 bit large gamut color spaces. You have to actually read his books to understand what his real opinion is and understand where his complaints are based. As far as large gamut, what's larger than Lab - well, mathematically maybe ProPhoto RGB, but not in the real world.

One of the cool things The Canyon Connundrum does is show you examples of how Lab interacts with different RGB workspaces with real world images and how you can take advantage of some of Photoshop's limitations when dealing with severely out of gamut colors

The most important thing to remember is that the typical single round trip into Lab and back does no real damage to the image, and if you're adept at your Lab adjustments, your image will benefit greatly.

On a practical note, I work on hundreds of stock images a year for various clients, getting paid to make some fairly mediocre iStock and Getty Images files look like I might have actually taken them. Every one of those stock images comes to me as an 8 bit (usually) sRGB number eight compressed jpeg, and I'd say that maybe 60 percent make the trip to Lab and either back to RGB or straight to CMYK from there. These are usually printed on high end presses at 175-200 line screen and they all look great, as do the Epson proofs I make prior to going to press. Yes, every once in a while, a blue sky falls apart, but it would have broken apart regardless. So, while it might not be advised, I'm living proof that you can move 8 bit jpegs to Lab and back if you need to.

Sometimes it's not that you can't get similar results in RGB, but it's the fact that I might have to plow through a hundred images in an afternoon and I'm after the fastest most efficient workflow possible. Not everyone "gets" Lab, but for those wiling to spend the time to really learn it, it can be one of the most powerful tools in your color correction/retouching toolkit.

Peter Figen

[stamper]

"The testing in this thread is actually making me a little more interested in using messing around with LAB color. Funny how things work sometimes."

Then I highly recommend Dan's Lab book, The Canyon Connundrum. As well as leading you through the ins and outs of Lab, it will also help you learn when Lab is appropriate and when it's not - and it's definitely not always a benefit. Knowing when makes you a better retoucher.

As far as Dan being totally down on hi-bit high gamut workflows, this is somewhat disengenuous. He's not and even advocates that for most processes it won't hurt. What he objects to, and this is my opinion, is that there is a certain constituent that claims it's always beneficial to work in 16 bit large gamut color spaces. You have to actually read his books to understand what his real opinion is and understand where his complaints are based. As far as large gamut, what's larger than Lab - well, mathematically maybe ProPhoto RGB, but not in the real world.

One of the cool things The Canyon Connundrum does is show you examples of how Lab interacts with different RGB workspaces with real world images and how you can take advantage of some of Photoshop's limitations when dealing with severely out of gamut colors

The most important thing to remember is that the typical single round trip into Lab and back does no real damage to the image, and if you're adept at your Lab adjustments, your image will benefit greatly.

On a practical note, I work on hundreds of stock images a year for various clients, getting paid to make some fairly mediocre iStock and Getty Images files look like I might have actually taken them. Every one of those stock images comes to me as an 8 bit (usually) sRGB number eight compressed jpeg, and I'd say that maybe 60 percent make the trip to Lab and either back to RGB or straight to CMYK from there. These are usually printed on high end presses at 175-200 line screen and they all look great, as do the Epson proofs I make prior to going to press. Yes, every once in a while, a blue sky falls apart, but it would have broken apart regardless. So, while it might not be advised, I'm living proof that you can move 8 bit jpegs to Lab and back if you need to.

Sometimes it's not that you can't get similar results in RGB, but it's the fact that I might have to plow through a hundred images in an afternoon and I'm after the fastest most efficient workflow possible. Not everyone "gets" Lab, but for those wiling to spend the time to really learn it, it can be one of the most powerful tools in your color correction/retouching toolkit.

Peter Figen

An interesting post. For a long while I have pondered about sending some of my images to a stock library. What has put me off is that numerous magazine and internet posts state that these libraries demand the best images in a technical sense ( there was a post a couple of days ago on this subject ) Yet you state ( no criticism of you ) that you are making poor images look good? My question is why are these libraries accepting them?

[Jeremy Roussak]

For more background, go to Lindbloom's site and click on the info tab and then scroll down to the "Dan Margulies 16 bit challenge"

Ah, someone has mentioned his name. I thought I knew who was the subject of the discussion, but I had a nagging doubt that it was really Voldemort.

Jeremy

[Guillermo Luijk]

Every one of those stock images comes to me as an 8 bit (usually) sRGB number eight compressed jpeg, and I'd say that maybe 60 percent make the trip to Lab and either back to RGB or straight to CMYK from there. These are usually printed on high end presses at 175-200 line screen and they all look great, as do the Epson proofs I make prior to going to press. Yes, every once in a while, a blue sky falls apart, but it would have broken apart regardless. So, while it might not be advised, I'm living proof that you can move 8 bit jpegs to Lab and back if you need to.

In that case I think the important thing is not if your source image was 8-bit, but if the conversion process to Lab and back is done in 8 or 16 bits.
If you do that step in 16 bits (regardless the source image was 8 bit), and convert to 8-bit once in sRGB again,  the conversion process probably won't affect at all to the quality of the image. It would be only the things you do once in Lab that can affect it, like any other processing being applied to an 8-bit image.

See the third histogram is identical to the first. 16-bit rounding errors don't affect the levels of an 8-bit precision source image:

[bjanes]

In that case I think the important thing is not if your source image was 8-bit, but if the conversion process to Lab and back is done in 8 or 16 bits.
If you do that step in 16 bits (regardless the source image was 8 bit), and convert to 8-bit once in sRGB again,  the conversion process probably won't affect at all to the quality of the image. It would be only the things you do once in Lab that can affect it, like any other processing being applied to an 8-bit image.

That is especially true if the source space is relatively narrow, such as sRGB or aRGB. Dan stated that ProPhotoRGB is rarely used in professional work, but I think that is changing with the advent of inkjet printers with a wide gamut. Dan is used to working with printing presses.

Bill

[digitaldog]

How much these differences would show up in a print is yet to be determined, but the differences are quite striking in the analysis.

Might not show at all. Might show after more edits. That's the issue. Keeping your editing options open.

I suspect something like a smooth blue sky would suffer first. Then the printer would of course play a role. Now that Epson 90 series with HDR and 16-bit drivers are supported under Leopard, we're trying to see when there's an actual benefit of sending 16-bit through the print path. That's starting with high bit and going all the way to output. With 8-bit, I expect one could see banding in many images on such a device unless very careful editing is conducted.

[digitaldog]

As far as Dan being totally down on hi-bit high gamut workflows, this is somewhat disengenuous. He's not and even advocates that for most processes it won't hurt. What he objects to, and this is my opinion, is that there is a certain constituent that claims it's always beneficial to work in 16 bit large gamut color spaces. Y

Peter, I've archived all of Dan's silly quotes from his list about high bit, wide gamut, ACR is unfit for professional use etc. So you might wish to call it disingenuous but his words are clear and on the record. I find nothing close to disingenuous when his various rants over the years are placed easily within view and context. I can of course barf up all this nonsense here but much of it is archived already.

It will be interesting to poke my head in one of his PSW world sessions this month and see if he's still using CS2 as he was last show. I guess after he got kicked out of the PS beta, he decided whatever version he had was adequate for the job. Says a lot about this pundit IMHO.

Read the Lindbloom piece again. It shows how a true scientist handles a flat earth mindset.

[digitaldog]

An interesting post. For a long while I have pondered about sending some of my images to a stock library. What has put me off is that numerous magazine and internet posts state that these libraries demand the best images in a technical sense ( there was a post a couple of days ago on this subject ) Yet you state ( no criticism of you ) that you are making poor images look good? My question is why are these libraries accepting them?

You hit the nail on the head! Dan's entire life's work is about turd polishing. And he's damn good at it! If you look at his books or sessions, you see the most awful original images you can imagine. Worse, he tells his readers that many are from "Professional" photographers. Well maybe they shot them at 4X the legal alcohol limit. Really, my 85 year old mom takes better snaps.

If Dan had a clue about proper capture and Raw processing, his need for Lab or any of his other exotic fixes would be reduced to nearly zero. Lets not forget, his background (or so he says, I've yet to find anyone who can confirm what he did before becoming a "Photoshop expert" and found at least one person who dismisses his history) was supposedly fixing poor originals in a prepress house. He has a rather low regard for photographers in my opinion, but maybe my skin is thin for this group of which I belong.

If you've got crap images or your job is to make truds look better, I can think of no one else better to study than Dan. If not, he's pretty much a waste of time, especially with so many flat earth theories about high bit and wide gamut workflows.

Now if you're the shooter who under exposes JPEGs with the wrong white balance, Dan's your man. I find it far more effective to teach folks to shoot Raw and process the data appropriately which as Peter appears to desire, is far faster, less damaging and results in far better results. Not as impressive as showing a turd that on a scale from 1 to 10 is a 9, then jumping through 34 Photoshop hoops and 10 minutes later, ending up with a turd that's now only a 6. My advise, stay away from truds in the first place (don't make em)!

[digitaldog]

Dan stated that ProPhotoRGB is rarely used in professional work, but I think that is changing with the advent of inkjet printers with a wide gamut. Dan is used to working with printing presses.

Dan likes to make up stat's and history, another area that pisses me off. He would like his readers to believe he really has his pulse on the so called industry (whatever industry that might be). He's pulled the "Rush Limbaugh speaks for the Republicans" for years. And he sees everything as a file going to press. I don't know that he even has a modern ink jet printer and could recognize its qualities over a 133lpi CMYK press. Point is, one can work with high bit, wide gamut workflows as their master archive and spin off either an RGB iteration for CMYK conversions by others (I'd personally convert to ColorMatch RGB), or directly to CMYK. Oh, another of his flat earth ideas a few years ago was the U.S. Web Coated (SWOP) v2 profile in Photoshop was simply no good (I will not go into the silly details but he didn't "get" the fact that, IF you send a U.S. Web Coated (SWOP) v2 separated file to a press that really did conform to TR001, it printed really well, something Chris Murphy, Bruce Fraser and I demo'd years ago at a Seybold seminar). Its just another of his "a copy of Photoshop killed my parents" story that so many on his list believe as truth. I really believe you can lump Dan, Rush and Ken into a group. They all have useful things to say (well I'm a bit hard pressed to say this is true with Rush). What's so upsetting is the huge BS factor you have to edit through to get to the good data.

OK, back to our regularly scheduled posts on Lab conversions.....

[Peter_DL]

Peter, I've archived all of Dan's silly quotes from his list ....

So why at all ?

Fwiw – I’m sometimes using a fancy technique to repair burned skin tones which works in Lab only (with a given non-Raw image). Something about to make use of "Lab’s impossible colors" due to integer encoding. Clever technique – not more not less.

But that’s it for me.

Peter
(another Peter)

--

[digitaldog]

So why at all ?
Fwiw – I’m sometimes using a fancy technique to repair burned skin tones which works in Lab only (with a given non-Raw image). Something about to make use of "Lab’s impossible colors" due to integer encoding. Clever technique – not more not less.

What is an impossible color (yes I know, its another made up Dan term, like "false profile" and "Invisible color").

As the Chinese proverb says: The first step towards genius is calling things by their proper name. Not sure what to make of the opposite (someone who makes them up).

[Peter_DL]

What is an impossible color (yes I know, its another made up Dan term, like "false profile" and "Invisible color").

As the Chinese proverb says: The first step towards genius is calling things by their proper name. Not sure what to make of the opposite (someone who makes them up).

Pretty simple:

The integer encoding of Lab by means of three Cartesian coordinates (cuboid space), as given in Photoshop, bears a low Coding efficiency (just 35.1%). Means there are many many combinations of L, a and b which appear to be outside of the blob like Lab space and the gamut of human vision. This is illustrated in full 3D glory on Bruce Lindbloom’s website. The point is that we still can make use of these "not real colors" as Bruce likes to call it, or "impossible colors", based on the given mapping functions applied under the hood, thus turning everything to real-world colors.

For the science part, I recommend to study Bluce Lindbloom’s website (Lab Gamut display > Learn more). If you find a better term for this 3D effect just let us know.

Anyway, driven by your “anti Dan” obsession (as it seems) you missed or ignored the point that above quoted technique was the only motivation I found worth to mention for making an occasional trip to Lab.

Peter

--