Why "0" should be Zero

[TK2142]

Hi,

for those who didn't see the point of starting from neutral develop settings (in the respective LR4 discussions), watch this demonstration by Michael Frye (<- this is a link).

He nicely shows how starting from non-neutral develop settings does not suit some images at all and that these images require undoing all default develop settings before proper editing can start. I hope the demonstration shows the sceptics as to why neutral RAW develop settings can be useful.

I'll try to edit my camera profiles with the DNG profile editor once LR4 has developed to a point when it becomes usable. Hopefully, this will work as desired. However, I still think that no one should be required to jump through such hoops only to get a neutral develop setting. Setting sliders to "0", where "0" means zero, should be all that is required. Everyone else, who wants to start with some default contrast and other enhancements, should see those enhancements reflected in the slider settings.

The curve in the camera profile should just compensate any deviations of the camera from a desired norm. It should not implement someone's idea of a useful starting point for RAW editing.

[Jeff Schewe]

TK2142 wrote:

For some images it will be better to start without auto-highlights and auto-blacks so these non-optional auto-adjustments should be made optional.

You really need to understand the fundamental differences between "Auto" and "Image Adaptive"...

PV 2012 doesn't do anything "auto" (unless you click on the Auto" button). What it does do is an image adaptive ranging of the various controls' range. Does PV 2012 get more texture out of highlights and shadows? Yes, it does. Is this an "auto" function? I would dispute that characterization…What PV 2012 really does is evaluate the image and set the un-modified (zero settings) to have a range that will operate most productively, given that image's range between fully saturated (clipped) and low end blacks.

Is this a different approach? Yes, clearly…PV 2010's image adaptive controls were really only Fill Light and Clarity. Now the image adaptive control set includes everything in Basic other then Whites & Blacks (and Vibrance and Saturation).

You really need to learn how to set aside the concept or notion that anything in a raw file could EVER be "neutral" except for an image processed into a linear gamma (which as we've seen isn't really all that useful). Zero is a non sequitur...it doesn't really exist. What you THOUGHT was a "zero" setting in PV 2010 wasn't.

[TK2142]

Jeff Schewe wrote:

PV 2012 doesn't do anything "auto" (unless you click on the Auto" button).

Eric Chan himself mentioned "auto highlight recovery" and "auto black-point" functionality in LR4.

An image that shows blown highlights in LR3 will not show blown highlights in LR4 even when the same default develop settings are applied. That's because there is an automatic highlight recovery. This has nothing do with an "image adaptive" range of sliders.

You really need to learn how to set aside the concept or notion that anything in a raw file could EVER be "neutral" except for an image processed into a linear gamma (which as we've seen isn't really all that useful). 

Linear gamma does not make sense. Applying a gamma encoding that matches the output decoding (typically 2.2) makes sense. This defines "neutral". Period.

There is really no debate about the fact that the LR3 and LR4 default settings are arbitrary. Someone decided that these particular brightness and contrast boosts are useful. Maybe there are useful to many but the particular choices are *arbitrary*. Now to claim that if you did away with these boosts that the resulting settings were "just as arbitrary" and "not more neutral" just doesn't make sense.

You may be of the opinion that neutral settings are not useful for starting with image adjustments. Fine, then start with different settings. All I'm saying is that whatever you do to the RAW data (e.g., a brightness boost) should be reflected in slider settings, i.e., non-zero slider values.

And all I'm saying regarding "auto" stuff is that when someone wants "film like" roll-off of highlights or auto black-point settings that's fine, but they should ask for it. Rather than everyone getting it and then having to fight it when it doesn't fit.

[Jeff Schewe]

TK2142 wrote:

Eric Chan himself mentioned "auto highlight recovery" and "auto black-point" functionality in LR4.

An image that shows blown highlights in LR3 will not show blown highlights in LR4 even when the same default develop settings are applied. That's because there is an automatic highlight recovery. This has nothing do with an "image adaptive" range of sliders.

No, actually, it has EVERYTHING to do "image adaptive" rendering of highlight and shadow image data.

I won't contradict Eric but I think what he was trying to do was simplify the explaniation of what PV 2012 was doing. If you to rephrase the terminology to be Auto-Ranging then that's different (although I would still argue about the characterization being an Auto-Highlights and Auto-Shadows which I think summarizes your misunderstanding of PV 2012).

[TK2142]

Jeff Schewe wrote:

(although I would still argue about the characterization being an Auto-Highlights and Auto-Shadows which I think summarizes your misunderstanding of PV 2012).

*My* misunderstanding of PV2012?

Here's what Eric Chan has written:

In thread http://forums.adobe.com/message/4128876:

"Actually, the new defaults in PV 2012 are the same as the old defaults in PV 2010 rendering-wise, with the minor exception of the (now auto-calculated) black point." (emphasis is mine).

In  thread http://forums.adobe.com/thread/948951:

"highlight recovery is always enabled in PV 2012".

In thread http://forums.adobe.com/message/4166542, Eric acknowledges that the above two things have "no direct equivalent in Lr 3"

Please do not accuse someone of not knowing what is going on when you are not sure.

In any event, the non-optional "auto stuff" is a separate problem. I'm glad that we apparently made progress on the "what is neutral and what is not" front.

[Jeff Schewe]

TK2142 wrote:

In thread http://forums.adobe.com/message/4166542, Eric acknowledges that the above two things have "

Uh huh...did you actually read what he wrote? "Dorin, the automatic aspect has to do with the endpoint (i.e., white point & black point) handling.  The shape of the curve is calculated automatically to make transitions smoother as you approach clipping (to pure white or pure black).  Otherwise you'd get the traditional digital "hard clip" which can be rather rough, esp. in the shadows."

So, I ask again, do you understand what Eric is talking about? That the highlights and shadows aren't being automatically being set but the RANGE is? The "image adaptive" aspect is setting the end points for highlights and shadows? So that the range of the PV 2012 adjustments are image dependent?

Again, that's considerably different that saying highlights and shadows are being set "Automatically".

TK2142 wrote:

Please do not accuse someone of not knowing what is going on when you are not sure.

Well, I seem to be just a bit more sure of what's going on than you are...correct me if I'm wrong.

[TK2142]

Jeff Schewe wrote:

So, I ask again, do you understand what Eric is talking about? That the highlights and shadows aren't being automatically being set but the RANGE is?

If what you are saying were true then the following statement from Eric Chan wouldn't make sense:

From http://forums.adobe.com/message/4138532#4138532:

"Yes, Victoria, you can use a scene-referred starting point in PV 2010, update to PV 2012, and save that result (not just the point curve in the Tone Curve tab, but also the other resulting settings in the Basic panel) as a preset.  That will be close to a scene-referred starting point for PV 2012.  It won't be exact for a few reasons (e.g., highlight recovery is always enabled in PV 2012), but in many cases it may be close enough."

In other words, it is NOT the case that the LR4 "auto behaviour" only influences the range of the curves. The auto-calculated end shapes of the curves already influence the rendering when sliders are at default positions.

Often that can be just what you want, sometimes it isn't. And instead of making it easy to get what you want when you need it (activate recovery), LR4 makes it hard to counteract the behaviour when you don't need it. If there is consistency amongst the feedback for the LR4 basic controls I've read about is that people love it when it works (for most of the images) but hate it when it doesn't work (for a few images). With less "we know what users want"-attitude the commonplace love/hate relationship could have been a love-only relationship.

[Jeff Schewe]

TK2142 wrote:

With less "we know what users want"-attitude the commonplace love/hate relationship could have been a love-only relationship.

If you want to argue the relative merits of optimizing image adjustments with Thomas Knoll, you go right ahead.

Thomas isn't a dope...he kinda started this whole digital imaging thing ya know?

He signed off on the PV 2012 (and added to it along with Eric Chan and Mark Hamburg who started Lightroom). You go right ahead and argue with these guys if ya want...I do on very rare occasions (when I have a real point and can make the case–so far you haven't). So far your arguments haven't gotten a lot of traction...wonder why? Could it be that you might, just possibly be wrong? Think about it...

[TK2142]

Jeff Schewe wrote:

If you want to argue the relative merits of optimizing image adjustments with Thomas Knoll, you go right ahead.

I'd love to because I know a discussion with him would have logic as a common ground and he would not resort to quoting "authorities" when he runs out of arguments.

You may also acknowledge that I didn't say that the whole "image adaptive" approach is flawed. I only take issue with "0" not meaning zero and with certain behaviour being non-optional, as useful it may be in the majority of cases.

By quoting Eric Chan I demonstrated that your statements were wrong and now that you don't have to say anything anymore you appeal to authorities and claim I haven't gained any "traction". I give you that, I give up trying to achieve "traction" with you.

P.S.: FYI, I've listened to many George Jardine Adobe podcasts featuring Thomas Knoll and others, including you. What Thomas Knoll said always made sense and you may remember that he now and then disagreed with what Mark Hamburg or you said.

[Jeff Schewe]

TK2142 wrote:

I'd love to because I know a discussion with him would have logic as a common ground and he would not resort to quoting "authorities" when he runs out of arguments.

I've not run out of aurguments, I was simply addressing your statement "With less "we know what users want"-attitude the commonplace love/hate relationship could have been a love-only relationship." and letting you know who the "we" is...

[TK2142]

Jeff Schewe wrote:

I've not run out of aurguments, I was simply addressing your statement "With less "we know what users want"-attitude the commonplace love/hate relationship could have been a love-only relationship." and letting you know who the "we" is...

It does not matter who the "we" is. The only thing that matters in a discussion based on logic is who is right. I could just as well claim that there are only four elements (earth, air, fire, water) because Aristotle said so.

BTW, maybe an apology is in order. Or should we adopt the notion that there is no automatic highlight recovery in LR4 because you say so, despite all the evidence to the contrary?

[Jeff Schewe]

TK2142 wrote:

BTW, maybe an apology is in order.

I would be happy to accept your apology…the upside is that after all this time (and discussion) maybe you've learned something? Heck even Micheal Frye has said "I'm not thrilled about losing a true linear point curve. But I can live with it."...

If he can, why can't you? Do you have some sort of personal agenda?

[Bill_Janes]

Jeff Schewe wrote:

TK2142 wrote:

In thread http://forums.adobe.com/message/4166542, Eric acknowledges that the above two things have "

Uh huh...did you actually read what he wrote? "Dorin, the automatic aspect has to do with the endpoint (i.e., white point & black point) handling.  The shape of the curve is calculated automatically to make transitions smoother as you approach clipping (to pure white or pure black).  Otherwise you'd get the traditional digital "hard clip" which can be rather rough, esp. in the shadows."

So, I ask again, do you understand what Eric is talking about? That the highlights and shadows aren't being automatically being set but the RANGE is? The "image adaptive" aspect is setting the end points for highlights and shadows? So that the range of the PV 2012 adjustments are image dependent?

Again, that's considerably different that saying highlights and shadows are being set "Automatically".

TK2142 wrote:

Please do not accuse someone of not knowing what is going on when you are not sure.

Well, I seem to be just a bit more sure of what's going on than you are...correct me if I'm wrong.

The "image adaptive" adjustments that ACR7/LR4 performs are not clear to me (and apparently others). Certainly, the technical paper on LaPlacian operators is incomprehensible to me and likely most other readers of this thread. Rather than attempting to parse Eric's statements, a bit of empirical observation can shed some light on the matter.

A Stouffer wedge is a good way to evaluate the tone curve, since each step is 0.1 OD or approximately 3 steps per f/stop. Here is the wedge exposed so that step 1 is just short of clipping in the green channel and rendered into sRGB with ACR 6.6 with a linear tone curve (sliders 0 on the main tab and point curve set to linear). The image is approximately scene referred and approximates how the wedge appeared to the eye.

Step one has an OD of 0.05, but I placed it at a pixel value of 255. Steps 8 and 9 are near mid-gray (18%). Calculated values are shown in this Excel spreadsheet. The values A, M and B correspond to the steps on a Kodak Q14 chart which represents the values in many real scenes for highlight, midtone, and shadows.

Here is the ACR 7.0 PV2010 preview with the rendered sRGB values. An exposure of -0.5 EV was necessary to allow for the BaselineOffset that PV2010 uses.

And here is the PV2012 view with default settings. The highlights are too bright and scrunched together.

And here are the settings needed with PV2012 to approximate the linear curve in PV2010. I had to invoke the point curve to place the midtones as I could not get the proper values for the mid-tones with the sliders.

And now to the subject of automatic highlight recovery. It is necessary to use a special program to reveal the true status of the raw file, and I used Rawdigger. The histograms of the normally exposed image (05) and an image overexposed by 2/3 stop (image 03) are shown. In the overexposed shot the green channels are clipped and the red and blue channels are short of clipping, enabling highlight recovery.

And finally, here is the overexposed image (03) with the default PV2012 settings. It is clear that automatic highlight recovery has been performed as step 1 is now at a pixel value of 254 (I used an exposure of -0.05 EV to show that the pixel value was likely not clipped to 255). This is OK with me, but it does mean that proper ETTR exposure is diffucult to judge with PV2012 because of the automatic recovery. I anticipate that some readers will reply that they take photos of real scenes and not stepwedges and that scene referred images appear flat.

Personally, I am quite pleased with the changes in PV2012 for processing of real world images, but one does need to learn how to best use the new tools.

[TK2142]

Bill_Janes wrote:

Rather than attempting to parse Eric's statements, a bit of empirical observation can shed some light on the matter.

Great stuff, Bill, and visual proof of the mess LR4 creates in order to implement what someone else decided what "is good for me, always". It would all be fine (and perhaps I'll end up using the auto stuff for most real world images myself) if all this were optional.

There is another thread in which someone complains that LR4 auto-corrects his white product shot backgrounds for him and that he cannot blow them out with a single slider without washing out the product as well. His pain wouldn't be necessary if the auto highlight recovery were optional.

I furthermore cannot understand why there isn't a single slider anymore (was "exposure" in LR3) which allows me to negate any exposure compensation I chose as a camera setting. It seems in LR4 that one has to use a multitude of sliders to achieve that simple step.

[Lee_Jay]

Mess?  Let's talk about a mess.

I tried PV 2010 zeroed, and that is a massive mess.  It's like the ultimate "make my image look like crap" button - horrible skin tones, nasty colors on everything.  The only way that would be usable is if I also created a custom color profile designed for that tone curve.  The first thing I did was use the controls to undo most of what it did.

The profiles and the tone curves are designed to work togethger to produce a "pleasing", rather than scientifically accurate image.  The second is really only usable for science (spectroscopy, for example).

This is all a major tempest in a teapot in my opinion.  PV2010 zeroed on the top, PV2012 defaults on the bottom.

[Bill_Janes]

Lee Jay wrote:

Mess?  Let's talk about a mess.

I tried PV 2010 zeroed, and that is a massive mess.  It's like the ultimate "make my image look like crap" button - horrible skin tones, nasty colors on everything.  The only way that would be usable is if I also created a custom color profile designed for that tone curve.  The first thing I did was use the controls to undo most of what it did.

The profiles and the tone curves are designed to work togethger to produce a "pleasing", rather than scientifically accurate image.  The second is really only usable for science (spectroscopy, for example).

This is all a major tempest in a teapot in my opinion.  PV2010 zeroed on the top, PV2012 defaults on the bottom.

I don't think anyone is recommending zeroed PV2010 for fine art reproduction, but sometimes linearized output is desirable for technical analysis and specialized purposes. Of course, this "linear" output is still gamma encoded and one must remove the gamma encoding (see the link to the ICC web site for further explanation) for a scene referred image. ACR/LR is intended more for visual editing of images than for technical analysis and more specialized tools such as DCRaw or Iris can be used when needed.

If we had true high dynamic range displays, a scene referred image would look fine since it would reproduce the actual scene. However, for prints one must tone map the image so that a high dynamic range appears pleasing with a low dynamic range print. This is ably described by Karl Lang in his white paper on the Adobe web site (second link below). A simple S curve is a relatively crude method for tone mapping that is applied globally. Localized adaption is used for rendering of HDR images, but the results can appear unnatural if overdone.

While many photographers (myself included) are leery of automatic corrections, if PV2012 can give me better images with less work, I am all for it. I would be more interested in learning about the image specefic adaptions that ACR/LR performs. Of course, one can currently revert to PV2010 if he doesn't like PV2012, but that option may not be available in the future.

http://www.color.org/scene-referred.xalter

http://wwwimages.adobe.com/www.adobe.com/products/photoshop/family/prophotographer/pdfs/pscs3_render...

[Jeff Schewe]

Just to be perfectly clear, when ACR/LR designates a Process Version, it means it will stay in place for the future (assuming Adobe stays in business). That's the whole reason for that line in the sand...

With regards to PV 2012: Eric said: There are two separate ways in which PV 2012 is retrieving more highlight detail compared to PV 2010.

One is the improved highlight rendering logic, when one or two channels are clipped.  This is the "auto" part which happens even at default settings (e.g., Exposure 0, Highlights 0, etc.).  Previously in PV 2003/2010, the clipped channels would simply get clipped, so tones would be lost and colors could be shifted (e.g., skies would turn from blue to cyan).  You could get PV 2003/2010 to try to repair the partially clipped colors by using minus Exposure and/or Recovery.  In PV 2012, the new logic helps to preserve tonal distinctions and prevent color shifts, and this happens even without the user touching any controls.

The second way is the image-adaptive logic and new algorithms behind the controls (e.g., the tech driving the Highlights slider).  This part isn't automatic, because it only kicks in once the user starts adjusting things (e.g., setting Highlights to a non-zero value).

I tended to fixate on the second explanation vs. the first...if I have incorrectly described the wan PV 2012 is working, I'm sorry...(so I guess that counts as an apology, right?)

[TK2142]

Jeff Schewe wrote:

With regards to PV 2012: Eric said: ...

Thanks for sharing Eric's response. Was that from private communication or did he post that somewhere?

P.S.: I don't have an agenda. I'm just passionate about LR because it is so good in many ways. As much as PhotoDirector3, AfterShotPro, Aperture, etc. are better in some ways, LR (3.6) currently is still my first choice. Hence I want LR to be the best I can imagine.

[Jeff Schewe]

TK2142 wrote:

Thanks for sharing Eric's response. Was that from private communication or did he post that somewhere?

From an email exchange...

[Cornelia-I]

Thanks for the explanation about "auto-ranging":

Now I have an idea why PV2012 does such a marvelous job with ETTR-images (exposed to the right, overexposed shortly before blowing out highlights). And why for a +2/3EV neither full-auto-dev nor I myself choose -0.66 in exposure slider, but other values for all five basic sliders.

I have included PV2010 into my LR3-Frye-neutral-preset.

So when I think I do not achieve my vision in PV2012 it is one click away to be back at the start for setting the most usual 4 points into point curve.

But I find I never do nowadays!

I even manage to reach the same result if I start from my default settings or if I start from full-auto-development and tweak back what looks overdone for my taste.

Very valuable is the shift-doubleclick on single sliders for the "slider-only auto-dev".

I am still a dummy when it comes to use the point curve for some channels only, but this is a new avenue for me to explore and likely one limited to special causes.

TK2142,

If you stick to absolute numerical values PV2012 will be a pain for you, because with Auto-Ranging the relative scale will be moved per image. In that sense it would be *even more lying*, because "highlights -12" does not mean the same thing for two different images. Either one can make peace with this new relative concept or one should not upgrade.

I have arrived at liking it very much.

Cornelia

Cornelia

[Guest]

Do you guys ever shoot photos?

[Susan S.]

Interesting.... Has anyone who shoots panoramas, where you spend a fair amount of effort keeping exposure constant between shots had any difficulty blending panos where PV 2012 attempts to optimize differently for each image? I guess I could download the trial and find out for myself...

My playing with the beta found that the optimization of white and black points worked really well for most images with the biggest exception being those where I wanted to blow out the background deliberately. And as I shoot on white virtually never these days, it's not a big drama-I suspect tweaking the lighting to ensure more thorough blow out of the background would fix this anyway.

[CSS_Simon]

That's a very good question!  I'll experiment when I get a chance. 

My guess would be this: the auto adjustments that happen even before you touch a slider are to protect clipped highlights (and I thought blacks, though Jeff doesn't mention that above).  So they would not affect exposure and contrast of the image except right at the top (and maybe bottom) and then only to repair "damaged" parts.  Hence it would not cause problems with panoramas. 

[TK2142]

Susan S. wrote:

Has anyone who shoots panoramas, where you spend a fair amount of effort keeping exposure constant between shots had any difficulty blending panos where PV 2012 attempts to optimize differently for each image?

It is a problem with PV 2012 and there are workarounds.

[TK2142]

Michael Frye's second video on the PV 2012 controls is out.

(Partial) Summary:

• There is no ideal starting point that fits all images.
• For some images (including your's, Lee Jay), the Adobe default is a better starting point than completely flat settings.
• But often the Adobe defaults (brightness boost) compress highlight contrast and lead to a "washed out" look.
• Hence many images benefit from getting tweaked from more neutral settings (i.e., without the default brightness and contrast boosts).
• It is easier (in terms of getting to where you want to be with an image) to add contrast to an image (at the working points and with the levels it needs) then to remove contrast.
• The PV 2012 basic panel controls are better than the PV 2010 controls in the sense that they more easily allow to mimic changes you would do to the tone curve.
• But the tone curve still allows more precise control (partially because you can fine tweak the working points of your boost points).
• The tradition of setting white and black points first still makes sense. One should therefore question the suggestion to use the basic panel control sliders in the order they present themselves. Often it is better to first set "Whites" and "Blacks" and then go to "Exposure". Eric Chan argued differently in a post, referring to the fact that tweaking "Exposure" will affect the whites and thus invalidate adjustments to "Whites" but surely this argument goes both ways, as the overlap is symmetric. If you tweak "Exposure" first, you'll have to revisit its setting again if you subsequently tweak "Whites". Revisiting hence seems inevitable but makes more sense to me to get the whitepoint right before you start compressing contrast with the "Exposure" slider against a whitepoint that is too low.

The video contains some further images which demonstrate the above points. To me all this reinforces that it was a baid idea to choose "0" to mean an arbitrary Adobe default whose built-in contrast and brightness boosts are not suited for certain images.

If you can learn something from the video, great!

If you still disagree with me, you don't have to tell me again.

[Jeff Schewe]

TK2142 wrote:

If you still disagree with me, you don't have to tell me again.

The only point I would still argue is that there's a reason Eric said what he said...if you don't first set the overall tone in Exposure then you're working on an image whose overall tone is wrong. Exposure and Contrast first does set up Highlights and Shadows nicely making Whites and Blacks being relegated to fine tuning (or jumping to Curves).

Note, I don't think Curves use (point or parametric) has been eliminated, but I think the need to go to curves has been reduced...