Ad
Ad
Ad
Pages: [1]   Bottom of Page
Print
Author Topic: Who's hue-locked?  (Read 3688 times)
Hening Bettermann
Sr. Member
****
Offline Offline

Posts: 554


WWW
« on: March 27, 2010, 06:12:21 PM »
ReplyReply

I remember from discussions on this forum that Photoshop in luminosity mode should be the choice if one wants to preserve hues when editing the tone curve. The same advice is found in Ben Wilmore's Photoshop Studio Techniques.

Here is my own test. I processed a raw image of the ColorChecker in Raw Developer with my home-made camera profile and output to a linear .tif. Then I adjusted the black and white patches by the numbers as close as I could get (pretty close) using a linear "curve", in 3 different ways:
a-ACR b-Photoshop adjustment layer in Normal mode and c- in luminosity mode. ACR and PS RGB are visually identical and reasonably close to the the synthetical CC, while PS lum is way out.

Is there something I overlooked? - (PSCS3, ACR 4.6)

Kind regards - Hening.

[attachment=21099:1_Synthe...cker.tif.jpeg]
[attachment=21100:2_5D2_153_ACR.jpg]
[attachment=21101:3_5D2_153_PS_RGB.jpg]
[attachment=21102:4_5D2_153_PS_lum.jpg]

---------------------------------------------------------------------
« Last Edit: August 04, 2011, 01:04:48 PM by Hening Bettermann » Logged

Jeremy Payne
Guest
« Reply #1 on: March 27, 2010, 06:37:43 PM »
ReplyReply

http://luminous-landscape.com/pdf/Curves.pdf

http://lightroom-news.com/2007/10/04/light...otoshop-curves/

Have you read these?
Logged
Hening Bettermann
Sr. Member
****
Offline Offline

Posts: 554


WWW
« Reply #2 on: March 28, 2010, 08:51:23 AM »
ReplyReply

Thank you for the reminder, JP. Yes I had read these articles before, but first now, when re-reading them, I see what I overlooked in my test setup: saturation.

What I had in mind were posts by Guillermo Luijk (which I fail to retrieve right now) where he seemed to demonstrate that curve editing in ACR is not really "hue-locked", whereas PS lum is.

Kind regards - Hening.
Logged

Marco Ugolini
Jr. Member
**
Offline Offline

Posts: 53



WWW
« Reply #3 on: March 28, 2010, 06:00:42 PM »
ReplyReply

Quote from: Hening Bettermann
Thank you for the reminder, JP. Yes I had read these articles before, but first now, when re-reading them, I see what I overlooked in my test setup: saturation.

What I had in mind were posts by Guillermo Luijk (which I fail to retrieve right now) where he seemed to demonstrate that curve editing in ACR is not really "hue-locked", whereas PS lum is.

Kind regards - Hening.
Could you please define the intended meaning of "hue-locked" in this context? If it means that a Photoshop Curve in Luminosity blend mode causes no hue shifts at all, that is incorrect, though it's also true that those are not excessively large, usually.

Relatively speaking, the hue shift is larger in Normal mode, but it occurs to some degree even in Luminosity mode — and in a manner that is not linear across the whole range of tones and colors. (For example, peak values like 255,0,0 or 255,0,255, etc., do not shift at all upon applying even the most draconian of curves, unless one moves the curve's very end points).
Logged

Marco Ugolini
Hening Bettermann
Sr. Member
****
Offline Offline

Posts: 554


WWW
« Reply #4 on: March 29, 2010, 08:08:13 AM »
ReplyReply

Hi Marco,
 
yes I thought in fact that PS lum would not cause any hue shifts.

I retrieved a post where digitaldog refers to your recipe for a luminosity mask, and I'll try it out ASAP.

Thank you for your interest! - Hening
Logged

Marco Ugolini
Jr. Member
**
Offline Offline

Posts: 53



WWW
« Reply #5 on: March 29, 2010, 02:59:54 PM »
ReplyReply

Quote from: Hening Bettermann
Hi Marco,
 
yes I thought in fact that PS lum would not cause any hue shifts.
Hi, Hening.

It causes less of a hue shift, but still some, compared to what happens in Normal blend mode.

Quote
I retrieved a post where digitaldog refers to your recipe for a luminosity mask, and I'll try it out ASAP.
What Andrew [digitaldog] is referring to is the use of an additional layer filled with white, or black, or a perfectly neutral RGB mix (R=G=B), set to "Color" blending mode, above a full-color image. From that, you can proceed to create a luminosity mask:

[blockquote]Select All [Command-A], then Edit > Copy Merged [Shift-Command-C], then paste above the full-color image the Grayscale layer which now resides in your clipboard [Command-V]; also, toss the previous black/white/neutral layer into the trash (it was only meant to be used as an intermediate tool).

Next, set the newly-created luminosity layer itself to Luminosity blend mode, then clip it to a Curves or Levels adjustment layer right above it. You may now proceed to make your tonal/color changes that way.[/blockquote]
That is a trick which allows the user to restrict the color modifications in the underlying image to changes in luminosity and saturation alone, while keeping the image's hue constant — though, strictly speaking, that is not actually completely true either. One can only go so far in lightening or darkening a color without changing its hue before it "crashes" against the "ceiling" or the "floor" of the RGB space's own boundaries. When that happens, the hue is forced to change, by necessity.

If you sample the colors before and after the changes using HSB in Photoshop, the Hue angle appears to remain constant. But HSB is device-dependent, meaning that its results vary depending on which RGB profile one is using, and thus it is not reliable as a measure of absolute color. On the other hand, if you use LCH (Lightness, Chroma, Hue — which is device independent, and therefore far more reliable), you will see that a hue shift has indeed occurred. (Unfortunately, Photoshop does not offer an LCH sampler, in spite of the fact that the math behind LCH would probably be fairly simple to implement. As things currently stand, you need professional-level tools — like MeasureTool, or ColorLab — to find out the LCH values in a file.)

One advantage of using adjustment layers clipped to a luminosity mask is that even the peak values undergo a change in appearance without having to touch the endpoints in the adjustment layer, whereas they normally remain constant unless you change the endpoints of your Curves or Levels controls.

The directions I provided above should be sufficient for you to try out this procedure, if you are interested. One warning, though: truth be told, keeping the hue constant (in terms of HSB values, anyway) creates results that, more often than not, end up looking unnatural and, frankly, kind of ugly. That is because of the way that we, as humans, perceive color — non-linearly and highly dependent on elements in the surrounding environment.

My bottom line is that, if for some reason you absolutely need to keep hue changes to a minimum (while still not perfectly constant), there are ways to do that. But tonal and color changes without some correspondingly substantial changes in hue (besides lightness and chroma/saturation) often appear unnatural. There are valid reasons for the way adjustment layers work in Photoshop in their "default" behavior, and those are often practical and sensible ones — to keep the images looking as natural as possible, according to the way our human vision system operates.
« Last Edit: March 29, 2010, 03:52:54 PM by Marco Ugolini » Logged

Marco Ugolini
Hening Bettermann
Sr. Member
****
Offline Offline

Posts: 554


WWW
« Reply #6 on: March 31, 2010, 02:47:56 PM »
ReplyReply

Thank you for your reply, Marco!

> There are valid reasons for the way adjustment layers work in Photoshop in their "default" behavior, and those are often practical and sensible ones — to keep the images looking as natural as possible, according to the way our human vision system operates.

If it was that well! How I wished for a software which I could trust really just tried to compensate for the way human vision operates as opposite to the linear sensor, and nothing else! All the time it would save me! But what I read is that the saturation, e.g., is designed to mimic the appearance of a historic medium (film), because that is what Mostpeople want because they are accustomed to it. That sounds to me like human habit rather than human vision. In my eyes, this is not a valid reason for anything. I grit my teeth and use the time it takes do what I can to work around it.

So what can I do?

I repeated the experiments as describe in my 1st post, adding a saturation adjustment layer. I adjusted the luminosity by the numbers to match the black and white patches of the synthetic CC. This required a "curve" with points 4/0 and 137/255 in PS and 2/0 and 154/255 in ACR. What is the reason for the difference?

With the saturation layer, I adjusted the Red patch to match the red value of the reference. (Triplet 161/56/57) This required the saturation slider set to +47.

This way I tried the following alternatives:

1-PS ad modum Ugolini
2-PS lum
3-PS RGB
4-ACR

1-PS Ugolini
The recipe as I spelled it out for myself. It might be useful for other beginners:

Procedure:
With the RGB image loaded in PS,
1-Create a Fill layer > Solid Color; white, black or neutral gray. I chose white.
2-Set the blend mode to Color. - This is the Luminosity Mask.
3-Select the whole canvas (cmd-a), do Edit>Copy Merged
4-With the Lum Mask in the Clipboard, delete the Fill layer
5-Paste the Lum mask above the Background (with the Background selected, press cmd-v)
6-Set blend mode to Lum
7-Create Adjustment layer Curves, and check Use Previous Layer to Create Clipping Mask.
This layer in blend mode Normal.
8-Create Adjustment Layer (Hue)/Saturation in blend mode Color

Result for PS_Ugolini:

RGB triplet of Red patch: 161/63/62
God visual match to reference for the chart overall.
Nice histogram with all levels filled!

2-PS lum
A Curves layer in Lum mode, and a Saturation layer in Color mode
160/62/62
God visual match to reference.
Loss of levels in histogram!

3-PS RGB   -- see correction in post #11
A Curves layer in Normal mode; no additional Saturation layer.
142/69/62 (!)
Visually significant inferior match to reference compared to 1 and 2. Chart looks over all muddy.
More pronounced loss of levels in histogram!
A Saturation layer in color mode set to match the Red value of the Red patch (161) requires +31. This improves the visual appearance and the histogram, which now also shows levels in the left end like ACR. Red triplet is then 161/53/47.

4-ACR
To match the black and white points, the Point Curve had to be set to 2/0 and 154/255.
With these settings, the Red showed 160/61/59.
I left the HSL Saturation sliders at 0.
Good visual match to reference.
Histogram nicely filled. In ACR, it shows some values in the left end that are not shown when the ACR-edited tif is opened in PS.

Conclusion so far: -- see correction in post #11
Colors of the CC24 are visually the same in PS_Ugolini, PS_lum and ACR. PS_ RGB is way out, comes in the ballpark with saturation added, but the numbers favor the other options.
Due to the histogram findings, ACR and PS_Ugolini have the edge.

The proof of the pudding. A real image in PS_Ugolini, PS_lum, PS_RGB and ACR. The PS_RGB version without additional saturation. The same tone curve applied to all, 5/0, 251/254.
[attachment=21187:1_198_Ugolini.jpg]  1-Ugolini
[attachment=21188:2_198_PS_lum.jpg] 2-PS_lum
[attachment=21189:3_198_PS_RGB.jpg] 3-PS_RGB
[attachment=21190:4_198_ACR.jpg] 4-ACR

Marco, are the PS_lum and Ugolini versions more ugly than the RGB or ACR versions?

In PS lum & Ugolini, the snow in the upper part has some yellowish areas. This was the 19. of march (2009), the snow was melting and growing old and may in fact have had some "dirty" parts. The snow in the foreground is bluer than in ACR/RGB, which is more trustworthy, given that the sky was blue (as seen in the water) and this part is in the shadow.

I have a feeling that I will end up using PS_Ugolini in my workflow. Thanks to Marco for having invented this, and to Andrew for bringing it to our attention!

Good light, and a good Easter! - Hening.
« Last Edit: April 21, 2010, 06:37:31 AM by Hening Bettermann » Logged

Marco Ugolini
Jr. Member
**
Offline Offline

Posts: 53



WWW
« Reply #7 on: March 31, 2010, 04:11:43 PM »
ReplyReply

Quote from: Hening Bettermann
Thank you for your reply, Marco!
You're welcome. I'm glad to help, if I can.

Quote
How I wished for a software which I could trust really just tried to compensate for the way human vision operates as opposite to the linear sensor, and nothing else! All the time it would save me! But what I read is that the saturation, e.g., is designed to mimic the appearance of a historic medium (film), because that is what Mostpeople want because they are accustomed to it. That sounds to me like human habit rather than human vision.
Well, ask yourself: is there a difference? Habit or not, if that is what we have come to expect, then that's what we want. In practical terms it makes little difference.

Also, meeting the demands of human vision is very complex business. The way Curves and Levels work is only a partial attempt to mitigate the downsides that would otherwise present themselves in the absence of those compensatory behaviors in the software.

Of course, those are not perfect remedies, only imperfect ones.

Quote
In my eyes, this is not a valid reason for anything. I grit my teeth and use the time it takes do what I can to work around it.
Add to the already complex demands of "average" human vision those (far more unpredictable and complex ones) of any specific individual, and you understand how the plot rapidly thickens.

It's always a matter of refining the results to meet one's own needs. One cannot expect "one-size-fits-all" solutions in imaging, or "one-click" magic buttons that obviate the need for critical evaluation and skillful refinements.

Quote
Marco, are the PS_lum and Ugolini versions more ugly than the RGB or ACR versions?
Before I tell you what I think, let me make sure I got it right. From left to right, the images result from: (1) Curve in Lum mode; (2) Curve in Normal mode; (3) Curve clipped to a Luminosity Mask; (4) Curves applied in CameraRaw.

Is that correct?

Quote
I have a feeling that I will end up using PS_Ugolini in my workflow. Thanks to Marco for having invented this,
Thank you for saying that. Don't forget to credit me!  
Logged

Marco Ugolini
Hening Bettermann
Sr. Member
****
Offline Offline

Posts: 554


WWW
« Reply #8 on: March 31, 2010, 07:45:37 PM »
ReplyReply

Hi again Marco!

> Before I tell you what I think, let me make sure I got it right. From left to right, the images result from: (1) Curve in Lum mode; (2) Curve in Normal mode; (3) Curve clipped to a Luminosity Mask; (4) Curves applied in CameraRaw.

Ooops it is first now that I see that my images have no titles, so I come in doubt myself if the order in fact was as intended.  -- I have now labelled and re-arranged them - because it turns out that the order was in fact wrong.

The philosophy has to wait until tomorrow or so. Good night for now! - Hening.
Logged

Marco Ugolini
Jr. Member
**
Offline Offline

Posts: 53



WWW
« Reply #9 on: March 31, 2010, 10:10:23 PM »
ReplyReply

Quote from: Hening Bettermann
Hi again Marco!

> Before I tell you what I think, let me make sure I got it right. From left to right, the images result from: (1) Curve in Lum mode; (2) Curve in Normal mode; (3) Curve clipped to a Luminosity Mask; (4) Curves applied in CameraRaw.

Ooops it is first now that I see that my images have no titles, so I come in doubt myself if the order in fact was as intended.  -- I have now labelled and re-arranged them - because it turns out that the order was in fact wrong.
Thank you for fixing that, Hening.

Based on your naming of the files, I notice a gain in saturation (the greens, in particular, look cleaner and brighter) when using either the Curves in Luminosity mode or the Curves clipped to a Luminosity Mask. (That is a bit strange, though: usually the same Curves adjustment layer produces less saturation in Luminosity mode than it does in Normal blending mode.)

The ACR image appears slightly duller and darker, and seems to have been modified with some Recovery (there seems to be more detail in the highlights compared to the other 3).

Incidentally, it would be useful to see the source image, as it appeared before the changes were applied.

Thank you for sharing this with us.
Logged

Marco Ugolini
Hening Bettermann
Sr. Member
****
Offline Offline

Posts: 554


WWW
« Reply #10 on: April 01, 2010, 09:42:13 AM »
ReplyReply

Hi again!

First a correction:

There must have been something quite wrong with my CC shot in PS_RGB on which I based my observations of post #7 . When I re-open the image today, the colors are visually OK like the rest, without any additional saturation. The Red patch reads 161/61/60. So this is to replace the text under 3-PS RGB. And the conclusion is that it is only the preservation of levels which favors the other options.

Concerning the creek image:

I have not applied highlight recovery anywhere. But adjusting the CC image in ACR to meet the black and white patches of the reference required values different from the other 3. On the real image, I used the same curve by the numbers for all 4. This may be the reason that the ACR image shows more highlight detail.

Anyway, here is the base image, a linear tif output from Raw Developer with R-L deconvolution, radius 1, 10 iterations, ProPhoto profile. Then the CA was treated in ACR   (+6/-10), Defringe all edges. No other edits. Screen shot of tif displayed in Preview, converted in Preview to jpeg of default quality (8 of 10).

[attachment=21202:5D2_198_...orr_Grab.jpg]

And here are the readings for the Green patch of the CC:
Theoretical value in ProPhoto space, according to table on Bruce Lindblooms site: 84/123/67
Synthetic CC on my monitor calibrated to native White Point (7000 K) with Gamma 2.2:          101/146/73
Ugolini   103/147/84
PS_lum   103/147/84
PS_RGB   103/148/83
ACR      102/147/82

Good light! - Hening
Logged

Hening Bettermann
Sr. Member
****
Offline Offline

Posts: 554


WWW
« Reply #11 on: April 01, 2010, 03:23:29 PM »
ReplyReply

Hi again!

Can I confirm that the Ugolini method looses detail compared to PS_lum, PS_RGB or ACR, as Andrew Rodney reported?

Below is a little crop shown at 100%. The image was shot with a Canon 5D2 and a Distagon 35 mm, ISO 400, 1.6 seconds, probably f/11.

It was processed in Raw Developer using R-L deconvolution with radius 0.6, 10 iterations. No other sharpening. Output as linear tif. Then the CA was treated in ACR, -5/0, defringe all edges.

A tone curve of 4/0, 128/96 and 255/255 was applied in the same 4 different ways as above.

[attachment=21206:1_Detail_Ugolini.jpg] 1-Detail_Ugolini
[attachment=21207:2_Detail_PS_Lum.jpg] 2-Detail_PS_Lum
[attachment=21208:3_Detail_PS_RGB.jpg] 3-Detail_PS_RGB
[attachment=21209:4_Detail_ACR.jpg] 4-Detail_ACR
 
Can you see loss of detail? This time, I, too, see that the greens are clearer in Ugolini mode and PS_lum (upper left in image).

A strange experience: This time NO saturation increase was needed in the lum modes - any such would have looked quite unnatural, with ACR and PS_RGB as the visual reference.

That leaves me with the question: How shall I adjust saturation in the future? Is there any reference (visual or by the numbers) which just tries to be as natural as possible and is not tailored to meet the appearance of film or any other artifacts, pre-exspectations or habits?

Good light! - Hening.
« Last Edit: April 01, 2010, 03:26:28 PM by Hening Bettermann » Logged

Marco Ugolini
Jr. Member
**
Offline Offline

Posts: 53



WWW
« Reply #12 on: April 01, 2010, 09:46:47 PM »
ReplyReply

Quote from: Hening Bettermann
Anyway, here is the base image, a linear tif output from Raw Developer with R-L deconvolution, radius 1, 10 iterations, ProPhoto profile.
Well, this is a screenshot of an image file.

It's difficult to estimate the potential usefulness of a procedure without using the actual image file, and not just a screenshot of it.

Any chance to get the actual TIFF?
Logged

Marco Ugolini
Marco Ugolini
Jr. Member
**
Offline Offline

Posts: 53



WWW
« Reply #13 on: April 01, 2010, 10:27:02 PM »
ReplyReply

Why don't we use an image with far more saturation and color to test the results of our workflows?

I have an image that may fit the bill:

http://tinyurl.com/yaag2jd

I am offering 3 versions of it, in PSD format (layered), all of which use exactly the same Curves adjustment layer in slightly different combinations, and with clearly different results:

1) Curve_Clipped_to_Lum_Mask.psd - the adjustment curve is clipped to a Luminosity Mask;

2) Curve_in_Lum_Mode.psd - the Curves layer is set to Luminosity blend mode;

3) Curve_in_Normal_Mode.psd - the Curves layer is used in its default Normal mode.


I look forward to your comments.
« Last Edit: April 01, 2010, 10:29:37 PM by Marco Ugolini » Logged

Marco Ugolini
Hening Bettermann
Sr. Member
****
Offline Offline

Posts: 554


WWW
« Reply #14 on: April 02, 2010, 05:00:58 PM »
ReplyReply

Hi!

Here's a download link:
http://rapidshare.com/files/371287697/5D2_...CAkorr.tif.html
MD5: FFAB4F8BC259C8103DB8F0A977AB76DF

I don't know if the the second line will be needed as a password or something, I don't have so much experience with this.

Good light! - Hening.
« Last Edit: April 02, 2010, 05:15:42 PM by Hening Bettermann » Logged

Hening Bettermann
Sr. Member
****
Offline Offline

Posts: 554


WWW
« Reply #15 on: April 02, 2010, 05:14:33 PM »
ReplyReply

Hi Marco,

your 3 file names are just names, no links...

Kind regards - Hening.
Logged

Marco Ugolini
Jr. Member
**
Offline Offline

Posts: 53



WWW
« Reply #16 on: April 02, 2010, 07:20:16 PM »
ReplyReply

Quote from: Hening Bettermann
Hi Marco,

your 3 file names are just names, no links...

Kind regards - Hening.
No, you must click on the link above those 3 names. Here it is again:

http://tinyurl.com/yaag2jd
Logged

Marco Ugolini
Hening Bettermann
Sr. Member
****
Offline Offline

Posts: 554


WWW
« Reply #17 on: April 03, 2010, 02:00:17 PM »
ReplyReply

Hi again!

Ooops! now I got the images. - I have seen this parrot before... - In the Ugolini version, the yellow feathers show highlight shades which are lost in the 2 others. I refer to the part where the feathers look almost white, and the transition to yellow. This adds to an over all less saturation, but there is no way of knowing for me which degree of saturation is most natural. In the 2 other versions, this part is washed out to homogenous yellow.  And since the bird, without a reference, does not look unnatural, I favor the Ugolini version for the highlight detail alone.

Great!

So the question left is: How can I reconstruct a NATURAL degree of saturation?
In my creek image, I added +47 based on the CC, naïvely believing/hoping that I just could use this number for all images. But in my detail image, no saturation needed/could be added at all. I see that you have not added a saturation layer either.

I have done it and played with it. As the value rises,
-the shadows in the face turn red, making the face appear reddish over all
-detail in the yellow feathers is lost, but NOT in the highlights mentioned before
-even at +100%, there is NO LOSS OF HIGHLIGT DETAIL.
-A value around +20 would be my choice IF I wanted to increase saturation. This would preserve detail in the yellow feathers and keep the reddish appearance of the white face to a minimum that could hardly be detected even in direct comparison to the RGB version.

I think Ugolini wins! - Hening.
Logged

Marco Ugolini
Jr. Member
**
Offline Offline

Posts: 53



WWW
« Reply #18 on: April 03, 2010, 07:53:25 PM »
ReplyReply

Quote from: Hening Bettermann
Hi again!

Ooops! now I got the images. - I have seen this parrot before...
Yes, it's an image from the Nikon corporation. You are likely to have seen it elsewhere.

Quote
In the Ugolini version, the yellow feathers show highlight shades which are lost in the 2 others. I refer to the part where the feathers look almost white, and the transition to yellow. This adds to an over all less saturation, but there is no way of knowing for me which degree of saturation is most natural. In the 2 other versions, this part is washed out to homogenous yellow.  And since the bird, without a reference, does not look unnatural, I favor the Ugolini version for the highlight detail alone.
Visually, I prefer the image produced just by the Curves layer in Luminosity mode ("Curve_in_Lum_Mode.psd"). The image produced by the Curves layer clipped to a Luminosity Mask ("Curve_Clipped_to_Lum_Mask.psd") seems to me a bit "washed up" in the yellows, though I agree that the finer detail looks more pronounced.

I suspect that there is no "one-size-fits-all" solution, and that different images may respond better to one technique over another. One should keep them all within reach in the available "arsenal of tools", and use those that offer the results that are closest to what is expected.

Quote
So the question left is: How can I reconstruct a NATURAL degree of saturation?
Meaning, natural-looking?

Oftentimes in our line of work one has no access to the original scene, and can only produce the results that look best based on one's own taste and preferences, or those of the designer/art director. In the end, it's most often a subjective call.

Quote
In my creek image, I added +47 based on the CC, naïvely believing/hoping that I just could use this number for all images. But in my detail image, no saturation needed/could be added at all. I see that you have not added a saturation layer either.
No. Plenty of it in the image already.

Quote
I think Ugolini wins! - Hening.
Mind you, I don't look at this as a case of "winning" or "losing".

To me, it's a matter of being aware of a whole range of different tools — any combination of which may fit the needs of a given image.
« Last Edit: April 03, 2010, 07:53:52 PM by Marco Ugolini » Logged

Marco Ugolini
Pages: [1]   Top of Page
Print
Jump to:  

Ad
Ad
Ad