Got images and examples?

Rob Cole wrote:

 

sometimes, in the brightest highlights, tonal detail is lost by PV2012.

Are you speaking relatively, compared to some other conversion software or PV version, or just in an absolute sense?

If the latter, it occurs to me to point out that there certainly comes a point where the image is completely overexposed and all the camera's photosites are maxed-out.  In that case, we would expect a complete loss of tonality / detail.

Is the issue that you expect such complete overexposure to stay pure white during a conversion no matter what settings you dial in?

I'm not criticizing, just trying to fully understand your point of view and question.

-Noel

Yes, this is expected behavior. And as I can see, some details that are blown in PV2010 crop are not blown in PV2012 crop, so PV2012 is actually showing more detail, but with less contrast

Here are those images with strong level correction (range 220-235 expanded into 0-255), with PV2010 on the top, to show the amount of detail

As a side note, exposure slider functionality seems to be also changed in similar manner - it has 'soft' cutoff

Yes, I understand what you mean. I just wanted to make sure are you are talking about real loss of detail, which would mean some error in the code that would be potentially solved in next release or so, or you are not satisfied with additional compression of the highlights (they are already compressed by the tone curve in the profile), which means that we will have to wait until PV2014 or so for possible change of workflow ...

I prepared some further examples. PV2010 is on the left side, PV2012 on the right side. Photos were taken with various Canon compacts and DSLR-s (last one provided by Noel). In all cases, all controls were at default, but it was not possible to get much better result even using various sliders and local adjustments. I used my v4 camera profiles calibrated for sRGB color space + Adobe standard added in the last example

On the first example, there is a loss of contrast in central part of the crop with PV2012, caused by big white vehicle that was at the parking place at the bottom of the photo, blown by the sun, so all highlights were compressed. PV2010 crop looks better, because there are no blown parts there (or maybe a pixel or two) and I don't care about the vehicle that was there just because I took the photo from the parking place below the mountain. By raising the whites slider, it's possible to decrease the amount of highlight compression (it was also suggested by Eric), but I can't get the same look as with PV2010

On the second example in the Blue cave (daylight coming under the water surface), it's hard to actually tell which PV is better. PV2012 has more or less correct hues but is visually a bit too dark, while PV2010 has considerable hue shift of blue color towards cyan.

It's actually not a magic of PV2012, but caused by higher sensitivity of green than red and blue pixels on the sensor. In old workflow, blown red-green colors had hue 60, while blown blue colors had hue 180, due to clamping R and B channel after whitebalancing. Now, when clamping is substituted with compression of all channels, it's about 40 for red-green range and 190-200 for blue, so blown sky and other colors in blue range look visually better - no more blown cyan areas. Also, blown skin, red flowers etc also look visually better, because they don't become too yellow. Unfortunately, blown green doesn't, which is illustrated on the third example. Look the blown parts of guy's T-shirt. With PV2010 it's yellow (hue 60), which looks acceptable. On PV2012 it's brown (hue around 40) - not very good. Similar with foliage etc. I already mentioned this issue in one post, but I'm not sure someone from Adobe noticed, so maybe this example will be more convincing

On the fourth example with the sun, with PV2012 there is no strange banding as in PV2010, but whole picture looks more orange due to the reason I explained above, so again hard to tell which one is better

On the last example, PV2012 with v4 camera profile (top picture) also avoided banding around the blue spotlight on the concert, present with PV2010, hue is closer to blue, so PV2012 clearly wins. But PV2012 performed even worse with Adobe standard  profile than PV2010 (bottom picture) - there is less hue shift towards cyan, but banding much worse. And PV2012 + Adobe standard was worse than PV2010 + camera profile. It's because camera profiles are, among others, tuned to partially correct nasties caused by blown channels, while adobe standard profiles don't have this 'improvements' + Adobe standard profiles are v1 type ...

So, regarding dealing with highlights, PV2012 is a mixed bag. When I started testing ACR 7, I was quite satisfied, but now, it's visible there are also some drawbacks compared to PV2010

I remember that blue spotlight.    And what a great 40D profile that discussion led to (I use it to this day exclusively).  Thank you again for that, Vit.

-Noel

Rob, I think that tampering with raw files that you mentioned, as a kind of temporary "fix" for the problem of highlights looking a bit dull compared to PV2010, doesn't make much sense, although it would be possible (but not as a general solution, because it depends on white balance that you set in ACR)

Another "fix" would be modifying the profile, so I just made an experimental one for D300 (after some quick testing on my cameras), so you can try if you want. This also isn't an universal solution, because amount of highlight compression depends on amount of highlights, so it will blow highlights more or less than PV2010/03

It is done by simply modifying the tone curve of Camera Standard v4 from Adobe. That way, hues are preserved, but highlights are expanded back by raising the top half of the curve in different way than with whites slider (which is actually unknown to me)

http://netload.in/dateixyiaHqHaul/Nikon D300 - Nikon D300 CSv4T1.zip.htm

Hope link will work - sites I used so far for sharing files are dead, so this is the first one I found.

There is actually nothing special with this profile. Here is the difference between original v4 profile and this one:

As you can see, I just raised and clipped off upper part of the tone curve in the profile (single line from bottom left to top right), to undo compression of more input dynamic range into output color space. In some cases it will produce visually better result, in some cases worse, depending on the photo. Producing new profiles like this one isn't universal solution, because amount of compression in PV2012 depends on amount of highlights, while tone curve in the profile is fixed. However, I believe that shape of curve that is compressing highlights in PV2012 and functionality of whites slider could be tuned a bit differently. But it requires testing the PV on huge number of real life photos so I don't think it will happen before PV2014 or so

What I don't like in PV2012 is that, when Adobe finally made better emulation od camera rendering using camera profiles, at the same time highlights are different, so result is again different than from camera even with perfect camera profile, unless there are no blown channels in the raw file, which is rare on my photos, because most landscape photos have large dynamic range.

I modified a profile with short program that consist of loading profile, changing tone curve points and saving the profile into another file. Code for manipulating with profiles is public available as dng sdk, but because it's written in c++ and I'm used to program in Delphi, and also because it is very large and complex, I written my own routines for most things. However, new additions in dng workflow (v4 profiles, PV2012) are not documented yet - new dng sdk still isn't available.

I didn't use DPE for years

Another solution, which I think is actually most usable here, is, as you suggested, creating some custom curves in ACR. After some experimenting, I ended up with this curve, which returns similar result as modification of profile (without a need to use other sliders). I just added a soft rolloff at the top

Points are (part of xml file):

   <crs:ToneCurve>

    <rdf:Seq>

     <rdf:li>0, 0</rdf:li>

     <rdf:li>160, 160</rdf:li>

     <rdf:li>175, 175</rdf:li>

     <rdf:li>180, 180</rdf:li>

     <rdf:li>241, 248</rdf:li>

     <rdf:li>243, 250</rdf:li>

     <rdf:li>255, 255</rdf:li>

    </rdf:Seq>

   </crs:ToneCurve>

Unfortunately, whites slider works somewhat differently. If moved to  +10 or +20, it will also raise highlights, but results were not good (the uppest highlights looked blown, the rest still a bit dull), so it's useless for me.

Yes, it works similar way as my curve, with advantage that it won't compress near-blown highlights (or truncate them in case of curve used in that profile), but will efectively decrease the exposure of the photo, uncompressing highlights at the same time (and hues will be slightly different)

However, if you correct exposure back by moving exposure slider to the right, you will apply another curve, that will compress highlights again

That fact is - Adobe team improved the workflow of ACR, so it is possible to extract almost 1 stop more from highlights than before, due to the fact that red and blue pixels have about half sensitivity of the green channel. So raw values from sensor are in range 0-1, but after whitebalancing green is stil in range 0-1, while red and blue are in range from 0 to almost 2 (depending on sensor and color temperature). With PV2010/03, this extra range of red and blue was simply thrown off. Now it is used and effective dynamic range is larger by almost 1 stop (except in green range where it's still the same). Since it must fit into the same output color space, something has to be compressed - it's unavoidable.

Now, the question is how to do it. Besides the method used by PV2012, which retains compatibility with existing profiles, there seems to be at least one other method, used by Canon cameras and DPP before dng profiles even existed (Nikons didn't have this at that time, don't know about recent models), where this compression isn't automatic (for cameras that I have at least), but is controlled by the profile only for tones where red and greed can be above 1 after whitebalancing, so highlights are compressed only where needed. However, for implementation of this in ACR, another update of dng specification would be needed  and probably some other thing changed, so it was made this way I presume. I actually don't know various internals and new workflow isn't documented yet, so it's possible that some of my claims here are not entirely correct, but I'm quite familiar with it since I needed to know how it works to make some modifications in my program for generating camera profiles

Well, there is a big contrast between the bird and the rest of the picture, so I would say this picture is not a representative of average photo and some average solution, but requires some artistic approach, which is  up to photographer's taste, so it's not the topic we were talking about. Of course, 'average photo' is different for every photographer and I presume that average photo of Adobe team was different than mine or Rob's

In this case, it seems that my taste is different than yours. I liked warm tone caused by the sun, so I wouldn't change it. I was just fine with highlights -20 to enhance the contrast on the bird and linear curve or even curve suggested by Rob. Curve that I suggested performed worse in this case, and  Whites slider also wasn't of help. PV2010 performed worse than PV2012. And another thing, Canon sRGB camera profile tends to oversaturate brown/orange highlights and raise the hue towards yellow (probably to make sunset photos looking "better") so they look like they are blown even when that part of raw isn't blown, which also happens here to some extent

Rob Cole wrote:

 

And overall, the photo was better in PV2012... (hard to get it bright enough in PV2010 without blowing out the whites, or using extreme recovery, and/or adverse effects of large fill-light...).

 

The question is, why do you want the rest of the picture to be brighter

I personally don't like HDR look on my photos at all. Some moderate boosting of shadows is fine by me so I can see them if they are really too dark, but no more than that

As we agree, there are various artistic interpretations. I generally tend to leave things as they aproximately looked like. So, as I said, I left everything at default, just moved highlights to about -20 and maybe used the curve that you suggested with slightly decreased upper midtones to get enough contrast on the bird. It looked beautiful to me (at least on the monitor in my office, which actually isn't calibrated). Nature is beautiful by itself, so I didn't feel a need of making other "enhancements". There is nothing wrong to make it your way or like Noel, just ask yourself, why do you need coold looking photo of a warm looking scene, with bright grass that wasn't very bright in reality, because the sun was low

Noel Carboni wrote:

 

In my case my focus was on making the white bird look quite white.

 

I understand that. I personally don't believe in theory about total adaptation of our eyes, so if I use a color picker on similar photos (which I don't do very often), I usually set WB somewhere in the middle between daylight and what was measured with color picker - in this case it was also close to what camera measured.

I would add another technical note about this photo. Actually, there is no blown pixels in raw nor prior nor after whitebalancing (if WB as shot used at least). So with v4 profile (which I also sent to Noel I think, because no v4 profiles for 40D from Adobe exist), PV2012 isn't additionally compressing highlights, compared to PV2010, so at default setting, result of both PVs is almost equal and almost equal to jpeg from camera. There are some minor differences in values of some pixels, but hard to notice any visual difference.

But, it's not the case with v1 profiles (Adobe standard and original camera profiles). There are differences in both shadows (that are darker with PV2010) and highlights (that are more blown).