Pages: [1]   Go Down

Author Topic: Recovery control(s)  (Read 13922 times)

Peter_DL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 544
Recovery control(s)
« on: August 22, 2008, 03:49:05 pm »

If of interest and maybe for discussion.
It’s about an ACR feature request which I’ve placed in the respective forum over there:
http://www.adobeforums.com/webx?14@@.3bb6a85c.3bbc8060/373

>>  Hi,

This proposal refers to the Recovery slider in ACR.
I would like to have more control(s) with this feature
i.e. Amount, Tonal Width and maybe Color Correction (saturation)
such as given with the Shadow/Highlights tool in Photoshop.

Background is as follows:  ACR is capable to show ‘full’ highlight details when setting almost all sliders of main adjustment tab to zero (except white balance of course) and by choosing an appropriate value for the Exposure slider.  That’s sometimes called a ‘linear rendition’. It’s not meant to deliver an overall pleasing image, but it can be informative regarding single aspects.
Now, when we raise the main tone curve (resulting from Brightness + Contrast functions) e.g. to standard Brightness 50 + Contrast 25, data are compressed beneath the upper shoulder of this S-curve. Hence, highlight details, contrast as well as saturation are easily getting lost.

The Recovery slider seems to try to undo such damage which already happened earlier in the processing chain.  I would assume that it’s based on a gradual selection of pixel (or mask) combined with some sort of darkening curve. However, I don’t know about its mechanics, and it’s not really relevant here.
In practice, I often find that there’s a quite high amount of Recovery needed to regain more or less ‘full’ highlight details (as initially seen with said ‘linear output’) which then has a too strong influence down to the midtones. Recovery tends to darken the light midtones as well which is not necessarily desired.  So it’s easy to run into a back and forth with all other sliders effecting tonality (including the curve tab)…

With the S/H tool in Photoshop, the Tonal Width (of selected pixel, as I assume again) can be pulled down to very low values and even to zero – which isn’t actually zero as the results prove. Together with an appropriate Amount of let’s say between 20 to 80%,
highlight recovery in fact stays limited to the highlights. For example the details of white clouds, etc.
My feeling is that such choice of high Amount and low Tonal Width is often preferred to mirror and counteract the roll off and related side effects from a sigmoidal tone curve (needed to compensate for dynamic range compression).

The latter consideration is of course subject to the image content, so I’d like to keep it flexible with more control(s) as suggested above. <<

Peter

--
Logged

madmanchan

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2115
    • Web
Recovery control(s)
« Reply #1 on: August 23, 2008, 11:11:43 am »

When using Recovery, instead of just using Brightness to get the midtones right, try a combination of Brightness and Fill Light. This should get you a better balance..
Logged
Eric Chan

bjanes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3387
Recovery control(s)
« Reply #2 on: August 23, 2008, 05:26:10 pm »

Quote
When using Recovery, instead of just using Brightness to get the midtones right, try a combination of Brightness and Fill Light. This should get you a better balance..
[a href=\"index.php?act=findpost&pid=216825\"][{POST_SNAPBACK}][/a]

Eric,

In the ACR09 tutorial by Jeff Schewe and Michael, there is a clip by Thomas Knoll describing how he sets tonality. He uses the exposure slider in conjunction with alt (windows) to set the highlights just short of clipping. If this disturbs the midtones, he then adjusts them with brightness as you suggest. He did not mention Fill Light but that also makes sense. One could also use ACR curves, but I understand that the Fill Light may use a mask and the action could not be duplicated by a mere curve.

Thomas also stated that another method introduced in Lightroom is to use the exposure slider to set the midtones and then recovery to set the highlights.

In previous versions of ACR, I had always used the Exposure Slider for highlight recovery, but some people now say one should use Recovery for this purpose. I'm sure you know all about these controls, but some readers may be interested in a simple test that I made.

I exposed a Stouffer wedge normally so that the brightest patch was just short of clipping made another exposure with about +1 EV over nominal. I then rendered the raw files with ACR set to linear (black = 0, contrast = 0, brightness = 0, and point curve set to linear) and plotted the results with Imatest and Excel as shown below.

The curve on the bottom (blue) is from the normal exposure and is linear with a gamma of about 1/1.77. The curve on top (black) is from the same ACR settings but with the camera set for an overexposure of 1/stop. It shows highlight clipping as expected. A negative exposure of - 1 EV in ACR brings the result down to approximately the same as the first exposure, as expected since the control is linear.

I then set the exposure back to 0 in ACR and applied various amounts of recovery as shown. I was a bit surprised by the result with recovery set to 100. The resulting curve is sigmoidal with an inflection point at around at an exposure of about -0.55 on a log scale. I had expected that it would merely roll off the highlights.

It would seem to me that one should use negative values on the Exposure slider if one needs to darken the whole image in a linear fashion to correct simple overexposure. If the midtones are OK but the highlights are blown, one could use the recovery slider. Perhaps you can comment on the various methods of highlight recovery.

[attachment=8046:attachment]

Bill
Logged

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
Recovery control(s)
« Reply #3 on: August 23, 2008, 10:53:38 pm »

Coming back to the world of adjusting real images, in LR/ACR I start the whole process with everything in the Basic section at zero except white balance which is "As Shot" (unless it needs to be amended) and the tone curve linear. This rather flat beginning provides a useful starting point for insuring that a positive Contrast setting is not doing things one doesn't want - such as contributing to blocked shadows - (note that varying the Contrast slider primarily affects the position of the dark tones on the histogram) or highlight clipping from excessive Brightness (collateral damage can happen).

Now, if the image still displays highlight clipping, the Recovery tool has been designed specifically to target and correct this problem (provided not ALL channels are blown) and from my practical experience is the most effective way of doing so.

While Brightness and Fill can both be used as Eric suggests, I seldom use Brightness because I find its effect is often too broadcast. If I use it at all, it is very gingerly. I prefer to adjust brightness and contrast in the Parametric Tone Curve, unless I need to move the end points, which is possible in ACR using the Point Curve. Generally either curve provides a great deal of control for targeting specific luminosity effects. Enhanced separation of the dark tones, however, is very will implemented and controlled using a combination of Fill Light and Blacks.  

Now, the relevance of all this to Peter's suggestion is simply this: if the adjustment process is started with flat enough settings, the compression problem he mentions is much less likely to occur in the first place, relieving some burden on the other tools. Where the Recovery slider pushed to 100 is not adequate to reveal apparently blown image detail, negative exposure values may do so, at the risk of severely under-exposing the rest of the image. When this happens, as Jeff and Michael show in the LR2 video tutorial, the image can be rescued by implementing two sets of adjustments - one for highlights and one for the rest of the image, and blending them with a layer mask in Photoshop. In sum, it's not clear whether embellishing the behaviour of the Recovery slider would necessarily add much value relative to the program's existing potential - save for convenience in some instances, which is always welcome.
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

madmanchan

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2115
    • Web
Recovery control(s)
« Reply #4 on: August 24, 2008, 09:21:37 am »

Bill, the executive summary for highlight recovery is as you say: there are basically two ways of doing this in CR/LR: you can either use a negative value for Exposure, or use Recovery (i.e., set Recovery to a non-zero value). The main difference is that Exposure affects the entire tonal range, so this is why the entire image becomes darker as you set Exposure to more and more negative values. So highlights get pushed to midtones, midtones turn to shadows, etc. Recovery has a much more limited tonal range, restricted to the upper midtones and highlights. So, unlike with Exposure, as you use more and more Recovery, the highlights will become flatter but the existing mid & dark tones will change minimally.

My suggestion for using fill light in my first post comes from the idea that if you're using significant amount of Recovery, then you'll get some highlight detail back. If you then use Brightness to get the midtones and shadows where you want, it's likely that you'll make the highlights too bright and counteract the benefits you achieved using Recovery. So Fill Light is more appropriate for this case because you'll push the lower tones without really impacting the highlights.

Hope that makes sense.
Logged
Eric Chan

bjanes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3387
Recovery control(s)
« Reply #5 on: August 24, 2008, 06:39:11 pm »

Quote
Hope that makes sense.
[a href=\"index.php?act=findpost&pid=216936\"][{POST_SNAPBACK}][/a]

Eric,

Thanks for the explanation; it does make a lot of sense and it is certainly valuable to have a senior person from Adobe available on the forum. As Mr. Knoll stated, there may be more than one way to accomplish the same task and the choice among methods may depend on whether you work visually or numerically. I suspect that Mark Segal falls into the former group whereas I am in the latter. In any case, based on your explanation and my understanding of the issues, this is my current approach:


a. If the dynamic range of the scene would fit into that of the camera but inadvertent overexposure has resulted in clipping of the highlights, the exposure control would be the appropriate tool, since the overexposure was global and linear and the exposure control will undo the overexposure.

b. If the dynamic range of the scene is greater than that of the camera and you have exposed for the midtones and allowed the highlights to be clipped, the recovery tool would be preferable. This spring I was shooting a lot of my son's high school games in the evening with strong side or back lighting causing the white and striped uniforms to be blown out along one edge, and the recovery control worked very well. Mark's approach may be more sophisticated, but when you are processing hundreds of images, individual approaches may not be practical.

Bill
Logged

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
Recovery control(s)
« Reply #6 on: August 24, 2008, 11:34:25 pm »

Hi Bill,

Actually, I don't work either uniquely visually or numerically. I try to relate the principles behind the application to the problems in the image and see what happens, trying various permutations and combinations as needed - by visual inspection on a well colour-managed set-up. Whether one works initially from test material or real world photographs, at some point I think we all need to get a bit "visual" - no?    

Normally, (and of course there are exceptions to every guidelines), with digital capture I've come to the well-discussed conclusion after my own testing that we should expose for the highlights to prevent clipping and keep the whole tonal scale rightward rather than leftward; where either it isn't workable or one makes mistakes, the raw processor comes to the rescue. If the whole image is under-exposed or over-exposed, sure, Exposure would be the first line of attack. But where the problem is mainly highlight clipping, Recovery is made just for that. So I think we are roughly in the same ballpark on these adjustments, though we may have gotten there along different paths!

Mark
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

bjanes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3387
Recovery control(s)
« Reply #7 on: August 25, 2008, 07:24:21 am »

Quote
Actually, I don't work either uniquely visually or numerically. I try to relate the principles behind the application to the problems in the image and see what happens, trying various permutations and combinations as needed - by visual inspection on a well colour-managed set-up. Whether one works initially from test material or real world photographs, at some point I think we all need to get a bit "visual" - no?   
[{POST_SNAPBACK}][/a]

Mark,

Yes, I agree that the final results of one's adjustment must be checked visually. Rendering a high contrast scene into the limited dynamic range of a reflection print is no simple task. ACR does a default rendering and one must adapt this to the needs of the individual image. Your suggesting of starting with a linear rendering (a preset would be useful for this) and your article on [a href=\"http://luminous-landscape.com/pdf/Curves.pdf]curves[/url] is a great help to understanding the use of this control. It helps to know exactly how each of the ACR controls works, and the Adobe documentation is rather general. One can get quite a bit more information from the Fraser/Schewe Real World Camera Raw with CS3 book, from online posts such as this, and from your own testing. For the latter, I find that plotting the characteristic curve is inavluable to determine which zones are being affected and by how much, but the end adjustment is visual as you correctly point out.
 
Quote
Normally, (and of course there are exceptions to every guidelines), with digital capture I've come to the well-discussed conclusion after my own testing that we should expose for the highlights to prevent clipping and keep the whole tonal scale rightward rather than leftward; where either it isn't workable or one makes mistakes, the raw processor comes to the rescue. If the whole image is under-exposed or over-exposed, sure, Exposure would be the first line of attack. But where the problem is mainly highlight clipping, Recovery is made just for that. So I think we are roughly in the same ballpark on these adjustments, though we may have gotten there along different paths!

[a href=\"index.php?act=findpost&pid=217052\"][{POST_SNAPBACK}][/a]

I agree entirely. Leaving headroom in the right area of the histogram does not make sense unless you actually need this headroom as a safety measure in cases when a second exposure might not be possible. Once you get to know how much highlight recovery is possible with your camera, you might permit some highlight clipping in noncritical areas and use highlight recovery. I find that quite a bit of recovery is possible with the D3.

Bill
Logged

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
Recovery control(s)
« Reply #8 on: August 25, 2008, 08:13:32 am »

Quote
Mark,

Rendering a high contrast scene into the limited dynamic range of a reflection print is no simple task.

..................................
 
I find that quite a bit of recovery is possible with the D3.

Bill
[a href=\"index.php?act=findpost&pid=217093\"][{POST_SNAPBACK}][/a]

Bill, a short rejoinder here though it starts to get OT - the first statement above I think is really the heart of the matter and what keeps challenging us all the time. One thing that really helps is the choice of printer and paper. I've totally converted from primarily reliance on an Epson 4800 and Enhanced Matte paper to an Epson 3800 and Ilford Gold Fibre Silk. The increase of DR is so obvious just by looking at the picture, but even more strikingly by comparing what soft-proofing does between the profiles for these media. I have no doubt that if you were to conduct a quantitative analysis of this comparison, it would conclude similarly.

As for the recovery of images from the D3 - same thing from the 1DsMk3. In general, as long as not all three channels are blown LR/ACR has tremendous recovery capability - it never ceases to amaze me - we owe this to the photographic knowledge and mathematical ingenuity of Thomas Knoll and Co. for making this possible. Likewise, ACR/LR does a superb job of revealing hidden detail in the sub-3/4 tones with judicious use of Fill, Blacks and the tone curves - and with the newer FF sensors one isn't necessarily overwhelmed with noise provided the initial capture exposure has been technically well done.

While, as I say, I'm a believer in ETTR, I agree with you that in conditions of uncertainty it makes sense to leave some headroom. For example, we don't mind sacrificing specular highlights, but we may not want to lose the texture of bright snow (as I had to consider when photographing in Yellowstone this past February) - so for some of those scenes I did leave a bit of headroom, because the distinction between what's specular and what contributes to highlight texture is not that obvious in the field.
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

Peter_DL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 544
Recovery control(s)
« Reply #9 on: August 25, 2008, 03:57:23 pm »

Instructive discussion (as I find), particularly e.g. referring to Eric’s well versed advise about how to make optimal use of the given controls. I’ve been exercising carefully through it. Many thanks!

After all, following excerpt of Mark’s comment and the cross-reference to Blending of two differently processed versions of the same Raw seems to me spot-on. In practice, the tricky part might typically be to shape the Luminosity mask for blending i.e. by selecting the relevant highlights only (via color range selections, in the simplest case), maybe painting with black or white on the mask, or (as I like) by applying the Levels tool … to narrow & focus the tonal width for blending (and to master transitions of course). So here we are again with “tonal width”.

Quote
... Where the Recovery slider pushed to 100 is not adequate to reveal apparently blown image detail, negative exposure values may do so, at the risk of severely under-exposing the rest of the image. When this happens, as Jeff and Michael show in the LR2 video tutorial, the image can be rescued by implementing two sets of adjustments - one for highlights and one for the rest of the image, and blending them with a layer mask in Photoshop. In sum, it's not clear whether embellishing the behaviour of the Recovery slider would necessarily add much value relative to the program's existing potential - save for convenience in some instances, which is always welcome.[a href=\"index.php?act=findpost&pid=216900\"][{POST_SNAPBACK}][/a]
Peter

--
Logged
Pages: [1]   Go Up