Date   

Re: Carnival: what's next?

Harvey Nagai
 

I found out about this group list only a few years ago, and in mining the
messages archive had discovered the Challenges and Case Studies of yore.
Those were highly instructive in a way that just isn't possible from
presentation formats (i.e. books, articles, videos). Seeing the creative
successes and also the not-so-successful results was interesting, reading
the commentaries was invaluable.

So I would very much welcome a return of case studies, if it wouldn't be
too much trouble.


Re: Carnival: what's next?

George Machen
 

Case studies would seem to be the very raison d'être of this group! We were limited by storage restrictions on Yahoo, but I can only find in our FAQ restrictions on individual photo sizes, not a total for a given group. (100 MB per photo in Basic groups, 500MB in Premium and Enterprise groups)

Dan's time and disposition are not infinitely elastic; I would think that he doesn't have to participate in every case study (although the more the better!), if that would mean more case studies. And I suppose we could always rein it in if it gets out of hand.

-- 
George Machen
 


Re: Carnival: what's next?

bill bane
 

I would love to see and try some additional whacks with additional case studies.

 

Bill Bane

 

From: colortheory@groups.io <colortheory@groups.io> On Behalf Of ROBIN MARK D'ROZARIO
Sent: Saturday, May 23, 2020 11:10 AM
To: colortheory@groups.io
Subject: Re: [colortheory] Carnival: what's next?

 

Dear Dan,
I enjoyed the Carnival exercise and vote Yes for further case studies.
Best regards,
Robin Mark D'ROZARIO


Re: Carnival: what's next?

Thomas Hurd,MD
 

I am interested in more case studies. I also enjoyed the Carnival case study very much.

Tom Hurd

On May 23, 2020, at 12:10 PM, ROBIN MARK D'ROZARIO <rdrozario@...> wrote:

Dear Dan,
I enjoyed the Carnival exercise and vote Yes for further case studies.
Best regards,
Robin Mark D'ROZARIO


Re: Carnival: what's next?

ROBIN MARK D'ROZARIO
 

Dear Dan,
I enjoyed the Carnival exercise and vote Yes for further case studies.
Best regards,
Robin Mark D'ROZARIO


Carnival: what's next?

Dan Margulis
 

Unless someone wants further discussion this will be my last post on the Carnival image. In past case studies we’ve tried to pick winners. Here, the top entrants are so close that trying to choose between them is like splitting hairs.

Also, in previous studies there have always been some completely fouled-up entries, useful because it highlights misconceptions. Here, the only obvious mistake was that some people concentrated so much on the reds that they ignored the need for strong yellows. Even so, there were no really bad entries.

I do think it was a worthwhile exercise. It highlighted some excellent techniques. It also may have instilled some confidence, because several people in their notes describing technique said things like “I’m sure this is going to seem terrible next to the others, because I know absolutely nothing about CMYK.” And of course they all did well.

It’s true that this is a very atypical image. OTOH it represents a category that we are probably seeing more of: a job where technical limitations (in this case, a lack of gamut) make it impossible to achieve what we would like (in this case, matching an extremely intense red). People who don’t understand the process give the work to us to figure out what is and is not possible. And not just to execute, but to explain.

This brings up the question of whether to do more case studies. I’m still stuck at home and so are many of us so it could probably be done. Then again many of us are not necessarily in the mood for such diversions. I’ve got more than enough interesting images but what we would need is interested people, since it takes an effort to run these case studies. So I ask now whether the list would like to see more of them. I won’t be offended if the answer is no.

Before signing off let me discuss one special Carnival case. The original artwork is silhouetted, no background at all. In #217 Harvey Nagal had the inspiration of adding a gradated black background. In his notes, he cited Chevreul, reasoning that the red robe would appear more brilliant against black than against white. And in this he was correct. There was nothing in the instructions to prevent this move, although in real life we’d need to run it by the client first.

In real life offset printing, large black areas like this can have unintended consequences, even if technically the file complies with the ink limit. So I would recommend not using this file in real life unless you are really confident about your press conditions.

Something lighter, however, might work. Harvey said he first had tried a pure Chevreul move of making the background the complementary of the robe, which in this case would be a cyan rather than a black. He reports: “I replaced the background with the inverse color of the costume, and found that the costume did indeed look as colorful in CMYK as it did in the original RGB image. The background looked like the worst puke spewing from a sewage pipe, but neutralizing the hue showed that the luminance contrast was sufficient to make the colors look brighter.”

I would point out that there are intermediate possibilities between dead black and bright cyan. Also, that it doesn’t have to be the exact complementary: anything cool might do. And white is the most brilliant background possible, so *anything* that darkens it may have a helpful effect—it doesn’t have to be a black.

You may wish to test this for yourself by downloading and blending #217 into your own version, at say 25% opacity. Or you might try a blend that partially excludes the green channel, making the background slightly purple rather than black. Or any of a hundred other permutations. The results are interesting.

Dan


Choosing the Best RGB Space as a Source for Channel Blending

Rick Gordon
 

It might be an interesting test to see if sRGB would normally be better for extracting channels for blending. My own sense is that, since channel blending is likely to be called for when there is missing detail in a current channel (i.e., either clipped or nearly clipped to either black or white), that the most suitable channel to blend with would be one that has sufficient latitude at both extremes, and that a wide-gamut space such as ProPhoto RGB would be the most serviceable.

This is a different argument than the argument that a wide-gamut space is always preferable as a composite RGB space. I generally concur with Dan's assessment that a smaller space may be a better tool for that if the gamut required is not excessive. However, since I opt to edit in 16-bit (though not necessarily use as final), I believe that the argument — that banding is more likely to occur when the values are compressed within a smaller range of gray values — may be mitigated by the 16-bit workflow. (But I dare not open that can of worms again.)

But I do think that some real-world testing of that hypothesis might be informative, in terms of the value of generally using a high-gamut space (e.g., ProPhoto RGB) as a source to derive unclipped individual channels.

Rick Gordon

--------------------
On May 20, 2020 at 5:57:25 PM [-0700], Dan Margulis Via Groups.io wrote in an email entitled "Re: [colortheory] The group's techniques, 2":
In more normal work it wouldn’t make a difference, and if it did sRGB would likely be a better choice than ProPhoto.
___________________________________________
RICK GORDON
EMERALD VALLEY GRAPHICS AND CONSULTING
___________________________________________
WWW: http://www.shelterpub.com


Re: Topaz Gigapixel AI anyone using it.

Beat C
 

I have once tried Topaz Gigapixel for blowing up a 10 Mb image to about 4 times it’s size (so 2 times as wide and 2 times as high).
I compared to PS enlargement algorithms. The Gigapixel result seemed better, until I discovered that it sharpens as well. The PS mode ‘Keep details’ does that as well, but less.
I eventually used the PS enlargement, as I had to do a lot of processing after the enlargement (complicated image from 3 images and lots of blends and processing), so I didn’t want to start off with sharpening in the beginning.
To really compare Gigapixel and PS enlargement, the sharpening should be the same. I have the idea, that starting with a good medium to big sized image (I have no experience with small images), the PS algorithm performs not much worse than the Gigapixel. And I think the Topaz products are rather expensive and not very user friendly.

Beat Cornaz


Re: The group's techniques, 2

Dan Margulis
 



On May 19, 2020, at 10:11 AM, Roberto Tartaglione <roberto@...> wrote:

Dear Dan,
I've applied your method on more images of the same kind
and I can confirm from my point of view that is a very reliable workflow, it can easily 
be automated too. Of course, it can not solve all multiple problems of 
a conversion in CMYK, but there are many ways to finish the job.

Correct. Usually this method is a good one for brightly colored RGB files that can’t be matched in CMYK. We see, however, that it is not the best way to handle this particular job. Also, Rick’s method of using ProPhoto RGB as a channel-blending source happens to work well in an exceptionally brilliant image like this one. In more normal work it wouldn’t make a difference, and if it did sRGB would likely be a better choice than ProPhoto.

You also offered me an important food for thought:
"A CMYK fact of life: the lighter the desired color, the worse the possible gamut mismatch. If we were doing pink flowers the problem would be much more severe, and we’d see a lot more variation in the quality of our corrections.”
In my professional life one of the harder (or perhaps The Harder) object to photographers has been Wine Rosè (although I was not involved in the Pre-press), but this is another story...

If you worked in France, where the rosés tend to be a paler pink, you wouldn’t have the problem. Italian winemakers prefer their rosatos to have a distinct hue, which can indeed be tricky to portray in CMYK. Worse, when they see your CMYK result, they say “you didn’t seem to have any difficulty with our reds and our whites, qual è il tuo cazzo problema?”

Which brings up a second question, the answer to which brings up a third. The author of #209, stung by the criticism that he had made the red robe too orange, submitted a new version to prove he knew how to make it rosier if that’s what the client wanted. Granted that the original saturation can’t be matched, I don’t see why it’s such a big deal to change hue as well. Personally, I think that the orange version seems a bit more violent than the others and therefore preferable.

But, as one of you remarked, this is exactly the kind of image where you show the client multiple versions. Not just because they might or might not like the orange-red, but because they probably won’t believe you when you say you can’t match the RGB red, just as they say that when Roberto’s reproduction of rosé wine is no good. Seeing multiple versions might convince them.

That brings up a third question, posted to me offline. How can we force more detail into the yellow purse? It isn’t difficult, but there’s one catch.

It may or may not be necessary to select the purse, but one way or another you need to find the current channel with the most detail (which will certainly be the blue, in this case), make a copy of it, increase contrast so that its lightest parts are blank, and then paste into the purse in Luminosity mode at some low opacity, possibly with an MMM action later.

If this were a red purse and the robe some other color, this method would work well. But yellow offers a particular trap. If you make the red robe slightly orange, as in #209, it isn’t necessarily bad. Nor would it necessarily be bad to make it a bit more purple. With yellow, it isn’t true. Make the purse slightly warmer and nobody will object. But make any part of it seem to have a greenish feel and the job gets rejected.

It happens that as yellow gets darker, people start to perceive greenness. So here, understanding that the blend has to be subtle, I would not use Luminosity mode. Instead I would make the blend twice as strong into the green channel as into the red.

Dan


Re: Topaz Gigapixel AI anyone using it.

Frederick Yocum
 

I use it to rescue low resolution photographs, usually taken with older mobile phones. It can deliver amazing results. When it isn’t amazing, it is usually because the high resolution version looks a little plasticy. Is it worth the price. If you need increase the resolution of images frequently, yes, yes, it is.
 
regards,

Frederick Yocum
frederick@...

Mobile: +1 717 341 2226
Skype: frederickyocum
Twitter: @frederickyocum
Website:frederickyocum.com





On May 19, 2020, at 6:07 PM, John M. Henry <John@...> wrote:

 
Ok I have customers who do not want to pay for my skills or basically anything. I have taught my people (not high designers)  to  use Photolemur and On1 for quick results or the Panel. Hell I loved scanprep pro in the day of the quadra 840av and photoshop 3 to automate. 
 
Anyone tried this program my friend is using.  This not high end but I get basically customers who want to pay for meatball color work not prime steak.  If I can automate and give them above ave results it’s a win as most can’t even do that. 

 

Here is his post , sample below

Has anyone tried Topaz Gigapixel AI?  I installed the trial version a few weeks ago and so far it seems like one of the most useful little $100 apps I've seen.  We've created 100's of graduation signs over the past couple weeks and a lot of the photos were really low rez.  This does an amazing job of recoving a lot of them.  Even some low rez line art cleans up. Not trying to advertise this, but it's not often that I find a product that does better than what Photoshop can do.  Here's an photo I just used on a grad sign (after/before).  Major difference. It's hard to believe the detail it pulls out of hair when it's not really there. Same with eyes and mouth. Just thought I'd share.

 

 
 
John M. Henry
 
Speedway Press  Mitchell Printing & Mailing Inc. The Phoenix Press
1 Burkle Street 
Oswego NY 13126
315-343-3531
 
 

 



<image001.png>



Re: Topaz Gigapixel AI anyone using it.

David Dewhurst
 

I've been using Topaz Gigapixel for a month and I have been blown away with what it can do. Turning tiny 72dpi images into very usable large 300dpi images for a historical book I'm writing. Definitely worth the $79 I paid for it.

David Dewhurst
Dewhurst Photography
64 Westbrook Lane
Phillips Ranch, CA 91766 USA
(909) 319-9570
www.dewhurstphoto.com


On Tue, May 19, 2020 at 3:07 PM John M. Henry <John@...> wrote:

 

Ok I have customers who do not want to pay for my skills or basically anything. I have taught my people (not high designers)  to  use Photolemur and On1 for quick results or the Panel. Hell I loved scanprep pro in the day of the quadra 840av and photoshop 3 to automate.

 

Anyone tried this program my friend is using.  This not high end but I get basically customers who want to pay for meatball color work not prime steak.  If I can automate and give them above ave results it’s a win as most can’t even do that.

 

Here is his post , sample below

Has anyone tried Topaz Gigapixel AI?  I installed the trial version a few weeks ago and so far it seems like one of the most useful little $100 apps I've seen.  We've created 100's of graduation signs over the past couple weeks and a lot of the photos were really low rez.  This does an amazing job of recoving a lot of them.  Even some low rez line art cleans up. Not trying to advertise this, but it's not often that I find a product that does better than what Photoshop can do.  Here's an photo I just used on a grad sign (after/before).  Major difference. It's hard to believe the detail it pulls out of hair when it's not really there. Same with eyes and mouth. Just thought I'd share.

 

 

 

John M. Henry

 

Speedway Press Mitchell Printing & Mailing Inc. The Phoenix Press

1 Burkle Street

Oswego NY 13126

315-343-3531

 

 

 




Re: Topaz Gigapixel AI anyone using it.

Kenneth Harris
 

Gigapixel does Ok on the shots you'd expect it to do Ok on, camera phone shots.   I used it successfully res'ing up low-res drone shots of surfers.  It can do strange things, so you should check what it's done at 1:1.  It does not perform miracles on 200x100px thumbs and it doesn't deal with jpeging artifacts well.  It's a bit better than PS "preserve details" enlargement, but much slower.

Ken Harris


Topaz Gigapixel AI anyone using it.

John M. Henry
 

 

Ok I have customers who do not want to pay for my skills or basically anything. I have taught my people (not high designers)  to  use Photolemur and On1 for quick results or the Panel. Hell I loved scanprep pro in the day of the quadra 840av and photoshop 3 to automate.

 

Anyone tried this program my friend is using.  This not high end but I get basically customers who want to pay for meatball color work not prime steak.  If I can automate and give them above ave results it’s a win as most can’t even do that.

 

Here is his post , sample below

Has anyone tried Topaz Gigapixel AI?  I installed the trial version a few weeks ago and so far it seems like one of the most useful little $100 apps I've seen.  We've created 100's of graduation signs over the past couple weeks and a lot of the photos were really low rez.  This does an amazing job of recoving a lot of them.  Even some low rez line art cleans up. Not trying to advertise this, but it's not often that I find a product that does better than what Photoshop can do.  Here's an photo I just used on a grad sign (after/before).  Major difference. It's hard to believe the detail it pulls out of hair when it's not really there. Same with eyes and mouth. Just thought I'd share.

 

 

 

John M. Henry

 

Speedway Press Mitchell Printing & Mailing Inc. The Phoenix Press

1 Burkle Street

Oswego NY 13126

315-343-3531

 

 

 




Re: The group's techniques, 2

Rick Gordon
 

To clarify, I never left CMYK in my attempt, either. What I did — and routinely do in an action as part of my CMYK conversion workflow — is to pull off a flattened copy before conversion (I'd usually be coming from Lab), convert it to ProPhoto RGB, and copy the R, G, and B channels back into my working document, and then convert, so I have available pre-conversion R, G, and B channels to access, if desired.

I always choose ProPhoto RGB for this because there is far less likelihood of clipped or near-clipped values — and my workflow is in 16-bit, so I'm not concerned about the potential downsides of ProPhoto RGB — and the purpose is really just to get R, G, and B channels with detail throughout, for blending purposes.

Actually, as part of the same CMYK conversion action, I always pull a double conversion, one Relative and one Perceptual, which is then duplicated to Lighter Color and Darker Color. So the working result is:

  • Relative conversion on the bottom
  • Perceptual Lighter above that (disabled by default and usually not used)
  • Perceptual Darker above that, which I find almost always improves the shadows and blacks
  • A copy of the Relative in Color mode above that
I can then, if I want, do a stamp of the composite result in Normal mode below the Relative Color layer and then do channel blending actions, as if I were in Luminosity mode, which I don't need to be, because the color is protected by the Relative color layer. I could then stamp a composite of the result below the Relative Color layer, and blend another channel.

I've been very happy with the results of this action, which then also has various ready-to-engage curving actions built in above all that.

The general idea is to create a layered conversion that allows me lots of options already in place and available for use.

Rick Gordon

--------------------
On May 19, 2020 at 1:07:08 PM [-0700], Thomas Hurd,md Via Groups.io wrote in an email entitled "Re: [colortheory] The group's techniques, 2":
Although in my entries I transported back and forth between color spaces, I never thought to use ProPhoto RGB, but I will soon be trying it. The technique I described above never left CMYK after the initial conversion.

Tom Hurd
___________________________________________
RICK GORDON
EMERALD VALLEY GRAPHICS AND CONSULTING
___________________________________________
WWW: http://www.shelterpub.com


Re: The group's techniques, 2

Thomas Hurd,MD
 

Roberto, 

Thanks for offering up your image for us.

I also enjoyed and used Robin’s suggestion. In fact, I used it on the whole image, from the scratch conversion to CMYK, using fully saturated, no masking.

I duplicated the CMYK base image and then used Robin’s screen blend suggestion. Then I duplicated that layer twice to my taste for the highlights. I put those 3 stacked screen blend layers together in a group. 

Then I used a base layer duplicate on top of that and used multiply blend, opacity to my taste and more duplicated multiply layers, all for more shadows. I put the multiply layers in a group. I then decreased opacity of each group to control highlights and shadow. 
I was pretty close to what I wanted with a lot less work! So thank you Robin.

There was not enough cyan, however, so I duplicated the base layer and then blended about 20% of the yellow channel into the Cyan to give it enough weight to play around with. (Before channel blending, there was almost no cyan in the screen left arm.)

Again I changed opacity with the cyan fortified layer to get a level around C 6-8% on that red in the left arm. That was pretty close to the par image, and I am now experimenting with a curve on top of the whole stack.

It’s not all the way to the par image yet, but it’s close to it, and better than my original entries. And it just took about three minutes.

I’m looking forward to trying and experimenting with all the other techniques Dan described, as well. Although in my entries I transported back and forth between color spaces, I never thought to use ProPhoto RGB, but I will soon be trying it. The technique I described above never left CMYK after the initial conversion.

Tom Hurd


On May 19, 2020, at 10:12 AM, Roberto Tartaglione <roberto@...> wrote:


Dear Dan,
I've applied your method on more images of the same kind
and I can confirm from my point of view that is a very reliable workflow, it can easily 
be automated too. Of course, it can not solve all multiple problems of 
a conversion in CMYK, but there are many ways to finish the job.

You also offered me an important food for thought:
"A CMYK fact of life: the lighter the desired color, the worse the possible gamut mismatch. If we were doing pink flowers the problem would be much more severe, and we’d see a lot more variation in the quality of our corrections.”
In my professional life one of the harder (or perhaps The Harder) object to photographers has been Wine Rosè (although I was not involved in the Pre-press), but this is another story...

In the next days, I wish also try to experiment with another strategy: to process again my files from RAW into ProPhoto RGB instead of AdobeRGB and then to follow the same routine. It is interesting because I always thought otherwise, a few times in the past I converted the image in sRGB before CMYK to narrow the gamut before the conversion, obtaining no bad results.

I’ve also tested  the suggested  channels blending from RGB to CMYK, I think is a very successful technique that I never would have thought about, but 
I think it requires more experience, I mean it can be easier to worsen the image.
Last but not least, the tip from Robin Mark D'Rozario, can add a final touch to the image, I think
it will be useful for more than an image in my future workflow.

Once again I wish to thank Dan Margulis for focusing on this case and
to all the contributors who really helped me grow.

Roberto



Re: The group's techniques, 2

Roberto Tartaglione
 

Dear Dan,
I've applied your method on more images of the same kind
and I can confirm from my point of view that is a very reliable workflow, it can easily 
be automated too. Of course, it can not solve all multiple problems of 
a conversion in CMYK, but there are many ways to finish the job.

You also offered me an important food for thought:
"A CMYK fact of life: the lighter the desired color, the worse the possible gamut mismatch. If we were doing pink flowers the problem would be much more severe, and we’d see a lot more variation in the quality of our corrections.”
In my professional life one of the harder (or perhaps The Harder) object to photographers has been Wine Rosè (although I was not involved in the Pre-press), but this is another story...

In the next days, I wish also try to experiment with another strategy: to process again my files from RAW into ProPhoto RGB instead of AdobeRGB and then to follow the same routine. It is interesting because I always thought otherwise, a few times in the past I converted the image in sRGB before CMYK to narrow the gamut before the conversion, obtaining no bad results.

I’ve also tested  the suggested  channels blending from RGB to CMYK, I think is a very successful technique that I never would have thought about, but 
I think it requires more experience, I mean it can be easier to worsen the image.
Last but not least, the tip from Robin Mark D'Rozario, can add a final touch to the image, I think
it will be useful for more than an image in my future workflow.

Once again I wish to thank Dan Margulis for focusing on this case and
to all the contributors who really helped me grow.

Roberto



Re: The group's techniques, 2

Thomas Hurd,MD
 

Dan,

Thank you so much for managing this competition. For me, this was the very first time I presented a file to anyone in CMYK.
But it was great fun and challenging to come up with a file, even though I had no idea what to expect, just that I was trying to head as close as possible to the original impossible RGB colors as I could manage while stuck in CMYK.
Looking at all the entrants before the comments were posted, I agreed with what you, and many others, concluded. There were a lot of images that were close to one another and pretty good.
On Saturday, I just kept looking back and forth, saying “contrast, color” over and over. And I was most anxious to read everyone’s comments. And they didn’t disappoint. 48 hours later I know a lot more about color correction in CMYK. What happens when you add C, why do you add C, where does M100Y100 show up, etc. 
Before I run on (and on...), I want to say that 209 was my favorite, because I liked the color of red and there was good detail maintained along the yellow stripe. 
And for most of the last 2 days, I thought thought red was a bunch of nearly equal M and Y. Then I decided it was a little more M than Y, but C < 6. Today I think maybe it’s c16m80y70b??
It’s probably even tougher to look at a monitor for an hour, with no other colors around except a white background and a yellow bag, not enough to even give me a good excuse to bring up simultaneous color contrast. 
And now I just want to know: what is red?

Tom Hurd

On May 18, 2020, at 12:05 PM, Dan Margulis via groups.io <dmargulis@...> wrote:



When faced with multiple problems in the same image it pays to plan out the approach before striking out in all directions at once. Here, how much work should be done in RGB? How much in RGB? In CMYK?

Before seeing the image, I would have bet that, since we have a somewhat RGB-centric group, that’s where the fun might be taking place, whereas since I have more CMYK experience I might be doing more work there. It didn’t turn out that way.

Five of us determined from the get-go that we were going to make at least two versions of the image and then combine them. In four of these cases one of the versions had been desaturated, either slightly or completely. Unless I missed it in the notes, I was the only one who used Hue/Saturation limited to Reds, so as not to disturb the yellows.

Only three people applied RGB curves and only two of them were trying to affect overall color; notably in #209 there was a desire for a warmer, more orange red.

Only two people tried channel operations within RGB. Both wanted to boost contrast in the red channel, a good idea since that’s the one most responsible for shape when the interest object is bright red. One person accomplished this by blending the green channel into the red; the other by multiplying the red into itself. Neither version made it to our list of favorites.

When the image starts out this colorful, LAB doesn’t have many attractions, so few people used it. What’s going on? If people aren’t going to use much RGB, or much LAB, surely they don’t just convert into CMYK and get the disaster that is the default, #200. Or do they?

A CMYK fact of life: the lighter the desired color, the worse the possible gamut mismatch. If we were doing pink flowers the problem would be much more severe, and we’d see a lot more variation in the quality of our corrections.

Why? Since we can’t lay down solid magenta ink without creating a red rather than a pink flower, we rely on the paper itself to create lightness. This is unfortunate, since paper is not absolutely white; it reflects a certain amount of green and of blue light, both of which are red-killers.

The response of the experienced CMYK practitioner is that adding more ink allows a more colorful result, and that it is better to have the flower too dark than too gray. But exactly how dark to make it is a tough judgment call

When an RGB file is full of brilliant colors that CMYK can’t hold, it’s critically important that we get them as bright as we possibly can. Here’s my favorite trick, which I illustrated in CC2E with a picture of brightly colored cycles.

1.   Assuming approximately correct colors in RGB, make two copies and move them into LAB.

2.   Copy 1: desaturate it slightly until you are sure that nothing will be lost by moving it into CMYK.

3.   Copy 2: use Color Boost or similar action to produce colors so bright that you absolutely will lose a lot of detail on conversion.

4.   Place Copy 2 as a layer on top of Copy 1.

5.   Change layer mode to Color. No apparent change from Normal mode, since the L channels of the two layers are identical.

6.   Move into CMYK without flattening. Surprise! Suddenly the detail returns, because Color mode in CMYK tries to retain the structure of the black channel, which is responsible for a lot of the detail.

7.   Add another copy of Copy 2 to the stack and set mode to Normal, massacring detail once again.

8.   Reduce its opacity to the maximum you can stand. The higher the opacity, the more colorful, but also the more detail is sacrificed.

So by habit, I used this elegant procedure and produced a suboptimal image. Why? Because with a lot of colors in play, of various darknesses, it’s really hard to know whether we’ve gotten all the color possible. Not true in this Carnival image. Getting maximum color is a piece of cake, pretty much everybody did it. The robe is dark enough that if we push its maximum color to 0c100m100y0k nobody will care. It’s easy to do with CMYK curves. No need for all these gyrations when we can just attack the detail in the CMYK channels directly.

Three other people followed the above procedure. Meanwhile, eight people took the more direct approach of converting into CMYK followed by blends into at least the cyan and often the magenta channel. Of these, one (#210) used my LAB procedure as well. The other seven, AFAIK, just glommed the file into the terrible #200 default and then fixed it. One of the eight blenders took his blend source from within CMYK but the other eight used RGB channels as the source: at least the red into the cyan, and often the green into the magenta. That is a highly sophisticated technique that I wouldn’t have expected to be so widespread. It appeared in many of the versions we listed as favorites, including #203, #210, #211, and #214.

Some of these people also applied curves to the RGB channels to add contrast before blending with them. Particular kudos to Rick Gordon, who thought to convert the Adobe RGB file to ProPhoto RGB before using its channels as a source. ProPhoto has a much wider gamut and isn’t impressed by how red this image is. It therefore shows subtler detail in its channels than Adobe RGB does.

Beyond that there was some variation, but few surprises. Two of us experimented with different rendering intents on conversion. One person created an entire layer of 0c100m100y0k and used it, masked, for various purposes. Six people used CMYK curves, sometimes through a network of masks. Two used Selective Color to enhance the shadows.

All in all, an excellent group effort.

Dan

 

 

 

 

 

 

 

 

 

 

 


The group's techniques, 2

Dan Margulis
 

When faced with multiple problems in the same image it pays to plan out the approach before striking out in all directions at once. Here, how much work should be done in RGB? How much in RGB? In CMYK?

Before seeing the image, I would have bet that, since we have a somewhat RGB-centric group, that’s where the fun might be taking place, whereas since I have more CMYK experience I might be doing more work there. It didn’t turn out that way.

Five of us determined from the get-go that we were going to make at least two versions of the image and then combine them. In four of these cases one of the versions had been desaturated, either slightly or completely. Unless I missed it in the notes, I was the only one who used Hue/Saturation limited to Reds, so as not to disturb the yellows.

Only three people applied RGB curves and only two of them were trying to affect overall color; notably in #209 there was a desire for a warmer, more orange red.

Only two people tried channel operations within RGB. Both wanted to boost contrast in the red channel, a good idea since that’s the one most responsible for shape when the interest object is bright red. One person accomplished this by blending the green channel into the red; the other by multiplying the red into itself. Neither version made it to our list of favorites.

When the image starts out this colorful, LAB doesn’t have many attractions, so few people used it. What’s going on? If people aren’t going to use much RGB, or much LAB, surely they don’t just convert into CMYK and get the disaster that is the default, #200. Or do they?

A CMYK fact of life: the lighter the desired color, the worse the possible gamut mismatch. If we were doing pink flowers the problem would be much more severe, and we’d see a lot more variation in the quality of our corrections.

Why? Since we can’t lay down solid magenta ink without creating a red rather than a pink flower, we rely on the paper itself to create lightness. This is unfortunate, since paper is not absolutely white; it reflects a certain amount of green and of blue light, both of which are red-killers.

The response of the experienced CMYK practitioner is that adding more ink allows a more colorful result, and that it is better to have the flower too dark than too gray. But exactly how dark to make it is a tough judgment call

When an RGB file is full of brilliant colors that CMYK can’t hold, it’s critically important that we get them as bright as we possibly can. Here’s my favorite trick, which I illustrated in CC2E with a picture of brightly colored cycles.

1.   Assuming approximately correct colors in RGB, make two copies and move them into LAB.

2.   Copy 1: desaturate it slightly until you are sure that nothing will be lost by moving it into CMYK.

3.   Copy 2: use Color Boost or similar action to produce colors so bright that you absolutely will lose a lot of detail on conversion.

4.   Place Copy 2 as a layer on top of Copy 1.

5.   Change layer mode to Color. No apparent change from Normal mode, since the L channels of the two layers are identical.

6.   Move into CMYK without flattening. Surprise! Suddenly the detail returns, because Color mode in CMYK tries to retain the structure of the black channel, which is responsible for a lot of the detail.

7.   Add another copy of Copy 2 to the stack and set mode to Normal, massacring detail once again.

8.   Reduce its opacity to the maximum you can stand. The higher the opacity, the more colorful, but also the more detail is sacrificed.

So by habit, I used this elegant procedure and produced a suboptimal image. Why? Because with a lot of colors in play, of various darknesses, it’s really hard to know whether we’ve gotten all the color possible. Not true in this Carnival image. Getting maximum color is a piece of cake, pretty much everybody did it. The robe is dark enough that if we push its maximum color to 0c100m100y0k nobody will care. It’s easy to do with CMYK curves. No need for all these gyrations when we can just attack the detail in the CMYK channels directly.

Three other people followed the above procedure. Meanwhile, eight people took the more direct approach of converting into CMYK followed by blends into at least the cyan and often the magenta channel. Of these, one (#210) used my LAB procedure as well. The other seven, AFAIK, just glommed the file into the terrible #200 default and then fixed it. One of the eight blenders took his blend source from within CMYK but the other eight used RGB channels as the source: at least the red into the cyan, and often the green into the magenta. That is a highly sophisticated technique that I wouldn’t have expected to be so widespread. It appeared in many of the versions we listed as favorites, including #203, #210, #211, and #214.

Some of these people also applied curves to the RGB channels to add contrast before blending with them. Particular kudos to Rick Gordon, who thought to convert the Adobe RGB file to ProPhoto RGB before using its channels as a source. ProPhoto has a much wider gamut and isn’t impressed by how red this image is. It therefore shows subtler detail in its channels than Adobe RGB does.

Beyond that there was some variation, but few surprises. Two of us experimented with different rendering intents on conversion. One person created an entire layer of 0c100m100y0k and used it, masked, for various purposes. Six people used CMYK curves, sometimes through a network of masks. Two used Selective Color to enhance the shadows.

All in all, an excellent group effort.

Dan

 

 

 

 

 

 

 

 

 

 

 


The group's techniques, 1

Dan Margulis
 

Having read over the notes of those who participated in this case study, first I want to thank everybody for being so complete in their descriptions of what they did. This has not always been the case in past studies.

This group was very sophisticated and it showed in the choice of techniques. In evaluating them we’re helped out by the several posts yesterday by people identifying their favorites. The consensus is that there isn’t one winner, unless it is the par version. We do, however, seem to agree on which are the six or seven best. I regret to say that mine is not one of them, because I offered up a solution in search of a problem. I’ll talk about this in my next post, which is about the big-ticket item: how, generally, to assure both good detail and extremely saturated color in this unusual image.

In this post, I’ll talk about some of the techniques of lesser importance. First, the actions found in the PPW panel weren’t a big factor here. One person trotted out the Bigger Hammer. Another tried the H-K action. Their results don’t show a clear gain and possibly a loss. Two people did use the LAB Skin Desaturation action, however, and it looks like it was helpful.

With the original being as colorful as it was, only three people got into the MMM and/or CB actions, and those that commented said there wasn’t much gain.

Some of the poorer results came because the red areas are so critical as to tempt the inexperienced to ignore everything else. This shows up in weak yellows and/or colorless hands. There was a fair amount of compensation for this:

*5 people gave special attention (a mask or Blend If to isolate) to the hands.

*3 gave special attention to the shoes.

*3 gave special attention to the purse.

*2 gave special attention to all yellows and not just the purse.

*2 used masks to isolate the highlights.

Some of the variants to fine-tune the result (remember, I’m not talking about the basic question of conversion until the next post)

*Sharpening. One person used a Clarity action before moving into CMYK. One used conventional USM once in CMYK. Two others used hiraloam in CMYK and it was helpful.

*One person found the reds too splotchy and blurred the A channel while in LAB.

*Three people used some variant of Overlay blending in CMYK to try to get more shape. A particularly original technique came from Paco Márquez, who used a layer of warm yellow set to Color Burn mode to make the purse more interesting, all other areas being masked out.

In the next post, I’ll talk about our basic strategies and not just these embellishments.

Dan

 

 


Re: Carnival results

Dan Margulis
 



On May 17, 2020, at 5:49 PM, Harvey Nagai via groups.io <hnagai@...> wrote:

And I have to say I was a bit taken aback by how poorly Mr. Margulis' version 
converted back to RGB (without black point compensation). Much of the reds 
became two-toned, light and dark with the nuances of the cmyk version nearly 
gone. Whereas the par version merely got a bit flatter, retaining transitions 
from lighter reds to darker reds much better. 

I do like the way the cymk version appears to have some "glowiness" to it, 
and that seems to have been smushed out in the conversion back to rgb.

I’m not sure which version you’re referring to but I have to wonder about a profile mismatch. I looked at the par version #220 just now in both RGB and CMYK and don’t see a difference offhand.

My understanding of cmyk is limited, but it looks like the channels are mostly 
lights and darks with very little in the midtones,

Correct. That’s how CMYK produces very intense, saturated colors.

and the conversion 
exacerbated this in the rgb channels?

Two factors in play here:

1) the black channel disappears and must be merged into the red, darkening it; and to a limited extent lightening the green/blue as well.

2) Adobe RGB has a much larger gamut in reds than CMYK (which is why we could never match the RGB original). Therefore even without the influence of the black channel being added, the red channel has to be much darker than the cyan of CMYK. Otherwise, the resulting color would be half a kilometer outside of the CMYK gamut and we’d be back in the same mess we were to begin with. For the same reason the green channel of RGB must be slightly lighter than the magenta of CMYK, in this particular case. In more everyday images we rarely see such an effect.

Dan

1341 - 1360 of 33113