This content has been marked as final. Show 91 replies
It cuts back the brightness in the Highlight region of the image to "recover" lost detail and texture in those areas.
I think that you will find Jeff Shewe's article on ACR 4.1 is a "MUST Read":
It helps to see what it's doing if you have the "show clipping" (or whatever it's called) switches in the upper corners of the histogram turned on.
and/or hold down the Option key (Alt on a PC) while moving that slider.
And I believe that highlight recovery works best -- only? -- on raw images.
In Bridge 2.1/ACR 4.1, it works on Tiffs and JPEGs too but not on PSDs.
Wolf is correct that one can recover a lot more detail from a RAW file, of course. There's not that much latitude to recover highlight detail in JPEGs and TIFFs, though something is better than nothing.
Thanks a lot, Ann.
It is quite a lengthy file. Does any know if there is a .pdf copy for downloading?
Be aware that color contamination of the highlight can occur. So far, I have not been offended by these shifts, and worse case can be tweaked out in PS.
Just open the Link and choose "Print" in Safari.
Then choose "Save as PDF" from the PDF button/drop down menu.
Name the file and save to your HD as a PDF complete with all images and text.
I would expect very few highlights to be recovered from an overexposed JPEG, because the data most likely have been clipped in all channels. Highlight recovery requires intact data in at least one channel.
To illustrate, shown below is a composite picture of a step wedge exposed at 5000K on a view box with a Nikon D200. The target is overexposed and the highlights are clipped as shown in the image second from top labeled "16 bit TIFF", which is converted in ACR to a 16 bit TIFF with no exposure compensation. Steps 1 through 4 are clipped and appear as white.
The picture labeled NEF-2 has been converted with an exposure value of -2, and the highlights are recovered and show no color cast. In the picture labeled TIFF-2, the image was converted to a 16 bit TIFF in ACR with no exposure compensation, saved, and then re-opened in ACR and highlight recovery was attempted. Some recovery is noted in step 4 (the steps are 0.3 density units, or about 1/3 stop). There is no color cast. The lower wedge values are brought down to where they should be (they were too light in the overexposed image).
The reason why recovery is possible with the raw file is that all three channels are not blown, as shown in the topmost picture, which shows the contents of the raw file (no white balance, or gamma correction). The raw pixel values have been multiplied by 8 to convert from 0..4095 in the 12 bit raw file to 0..32767 in the PS 15+1 16 bit format. The file was subsequently converted to 8 bit JEPG for illustration and may be downloaded for examination in Photoshop.
In step 2 of the raw file the green channel is blown, but the red is 221 and the blue 205 (expressed in 8 bit notation). To achieve white balance, all channels must be equal for this neutral target, and the red and blue channels are multiplied by WB coefficients and overflow occurs, clipping the values. Once clipped, they are lost and can not be recovered even when one is dealing with a 16 bit TIFF.
All you have to do is try it on overexposed images with lots of blown details.
The numbers will come down from 255 in the blown highlights, but there will be precious little detail recovered, as expected. A few, rare times I have been surprised, though.
ACR can still be a time saver for JPEGs and TIFFs for things like, white/gray color balance and other enhancements in ACR, like vibrance, saturation, clarity, straightening images in a single step, etc.
I work entirely in ACR 4.1 from scanned color negative film (16-bit) saved as Tiffs.
(An exposure on properly processed B&W or Color-negative film records a much wider range of exposure-values than does either transparency film or digital sensors.)
I then pay close attention to holding detail in both ends of the density curve when I scan; and the Fill and Recover controls in ACR 4.1 work wonderfully, and quickly (!), to provide files which retain detail in both Highlights and Shadow even in shots which were originally photographed under harsh contrasty lighting conditions.
>The numbers will come down from 255 in the blown highlights, but there will be precious little detail recovered, as expected. A few, rare times I have been surprised, though.
In my experiment, the numbers did not come down from 255 in the blown channels. They remained at 255. Please examine the illustration carefully.
But you are talking about a file which started as a deliberately over-exposed DIGITAL file, Bill.
There was no detail or texture there to hold to begin with.
Start with a real image of an actual object, and expose within the range of your recording medium, and Recover then has something that it can actually "recover".
Recovery will NOT bring total clipping down from total clipping...if all 3 channgels are clipped, they will remain clipped regardless of the Recovery setting. What Recovery _IS_doing is working on ANY channel detail (even 1 channel as Bill says) to reconstruct something...it's also doing a pretty unique curving to redistribute the extreme high end of the scale in the linear capture-something that simply no longer exists in a gamma encoded image. There is a ton of data in that brightest stop of exposure and as longs as SOMETHING is there, recovery can allow you to make use of it.
I don't think is can really be equated to "F stops of recovery" from blown highlights. The results are not the same as reducing the real exposure...
But it works, it works real good!
>(An exposure on properly processed B&W or Color-negative film records a much wider range of exposure-values than does either transparency film or digital sensors.)
Print film does have a large dynamic range, but experiments done by Roger Clark, a space scientist with a PhD in astrophysics from MIT show that the Canon EOS 1D MII has a greater dyanmic range than color print film. This is contrary to the conventional wisdom.
>But it works, it works real good!
>But you are talking about a file which started as a deliberately over-exposed DIGITAL file, Bill.
>There was no detail or texture there to hold to begin with.
>Start with a real image of an actual object, and expose within the range of your recording medium, and Recover then has something that it can actually "recover".
Ann, it doesn't matter whether the overexposure was intentional or by mistake (some of us are less than perfect). However, in the case of the raw capture, there was something to recover, whereas recovery was unsuccessful with the TIFF because there was nothing to recover. The green channel of the raw file was clipped, but could be reconstructed. I don't intentionally overexpose my pictures, but I do believe in exposure to the right and sometimes the images are overexposed and can be salvaged with highlight recovery.
If exposure is within the range of the medium, then recovery is not really necessary. The picture could be darkened with the usual methods in Photoshop or in the chemical darkroom. One can get good results from overexposed negatives, as long as the D-max of the film is not exceeded, simply by increasing the exposure time for the print.
I'm finding that the D80 actually does retain a value range I associate with color negs, at least close enough to be interchangeable in the majority of situations. It feels more neg than transparency.
An important difference is the noise(grain). In negs, the noise appears in the highlights, where it is better tolerated, and efforts to diminish it are better tolerated. An underexposed neg carries less noise penalty than an underexposed digital. But it isn't zero, and with really poor exposure, or efforts to coax more detail from even well exposed negs, garbage abounds!
I found that out the hard way.
>Start with a real image of an actual object, and expose within the range of your recording medium, and Recover then has something that it can actually "recover".
I have some additional thoughts on this topic. Let us assume that the dynamic range of my capture is 10 stops and the dynamic range of the subject is 6 stops. In this case, I have 4 stops of exposure latitude, and within this range of exposures (plus or minus 2 f/stops) I will capture the entire dynamic range of the subject. In printing the picture, it would be necessary with some exposures to use the exposure control of ACR to place the mid tones so that gray in the subject is gray in the picture. This is not recovery. Recovery would be needed if overexposure clipped the highlights.
In the above situation, exposure placement should take advantage of the characteristics of the capture medium. With digital capture, I would expose so that the highlights in the scene are just short of clipping (exposure to the right) to take advantage of the data rich upper f/stops of digital capture and to minimize noise. In other words, I would expose for the highlights. If the picture appears too light, then I would use negative exposure in ACR to place the tones where they should be.
With negative film, I would expose so that the shadows are just above clipping in the shadow area, as Ansel Adams advised for a short scale subject. This is exposing for the shadows. If the mid tones in the printed image are too dark, then I would lighten them with ACR's exposure control (actually this would move everything to the right, but I am using the mid tone as a reference point).
If the dynamic range of the subject were increased to 10 stops, then there would be no exposure latitude and only one exposure would capture the full dynamic range of the subject. If the dynamic range of the subject were 12 f/stops, then it would not be possible to capture everything and one would have to sacrifice highlights or shadows or else take two exposures and combine them.
>If the dynamic range of the subject were 12 f/stops, then it would not be possible to capture everything and one would have to sacrifice highlights or shadows or else take two exposures and combine them. >
With film, I adjust development time to compensate. (The old adage to "Expose for the shadows and develop for the highlights" long predates Ansel Adams!)
The point is that even in negatives with really dense HLs, I can get sufficient detail out of them using my scanning software to the hilt.
And the judicious use of Exposure level, Curves, Recovery and Fill in ACR 4.1 (from just a single scan) is letting me hold HL and shadow detail within printable limits in a totally remarkable way.
The zone system quantified "Expose for the shadows and develop for the highlights". As such, the lessons learned in applying the zone system still serves us well in the digital age, as Ann points out so well!
>> I would expect very few highlights to be recovered from an overexposed JPEG, because the data most likely have been clipped in all channels. Highlight recovery requires intact data in at least one channel.<<
Hi Bill, - nonetheless the Shadows/Highlights tool in Photoshop can very well recover many perceivable highlight details even from JPEGs, though this has less to do with clipping in an RGB 255 sense. Youre right that overexposure cannot be undone; its more like a sickle-shaped curve + selection which counteracts for the compression and visual merger of Levels under the upper shoulder of the notorious tone curve.
I think there are two different meanings of the term Recovery.
The Recovery slider (in LR, if comparable) was reported to do a bit of both as far as Raw files are concerned see Fig. 4 and related text. Whereas with a JPEG/TIFF in ACR if I understand your tests correctly the recovery of clipped channels is dropped but theres still a kind of effect like with the S/H tool. Makes sense, imho.
>With film, I adjust development time to compensate. (The old adage to "Expose for the shadows and develop for the highlights" long predates Ansel Adams!)
To compensate for what? If you look at the characteristic curve of T-Max 100 developed in T-Max developer for times ranging from 6 to 12 minutes, you will see that the captured dynamic range does not change much at all with increased development: it is about 3.5 logs or 11.6 f/stops, regardless of development. With increased development the slope of the characteristic curve steepens and maximum density increases, but there is very little change in the shadows. The same highlight luminance is being mapped to a higher density on the negative, and the mapped shadow luminance is unchanged. You may remember that push processing does not increase film speed (which is pegged to shadow detail).
>The point is that even in negatives with really dense HLs, I can get sufficient detail out of them using my scanning software to the hilt.
Your methods may be adequate for your needs, but they would not capture the dynamic range of many natural subjects, which can exceed 5 logs (100,000:1) or 16.6 f/stops. For that type of capture you would need HDR and 32 bit floating point representation in Photoshop. It can be done with film by combining more than one exposure, but the process is not easy because of the nonlinearities of film. See this paper by Paul Debevic (warning: large PDF) for details. It is much easier with digital because of the linear nature of the capture and can be accomplished with a few mouse clicks on merge to HDR.
When I was scanning transparencies with Silverfast, they did have some type of HDR but I never tried it and don't know how it works.
Peter Lange wrote:
>The Recovery slider (in LR, if comparable) was reported to do a bit of both as far as Raw files are concerned see Fig. 4 and related text. Whereas with a JPEG/TIFF in ACR if I understand your tests correctly the recovery of clipped channels is dropped but theres still a kind of effect like with the S/H tool. Makes sense, imho.
An interesting comment, as usual. Now that you mention it, I was doing my highlight recovery with the Exposure slider as with previous versions of Camera Raw. ACR 4.1 (and perhaps a couple of previous versions) now has a recovery slider, and I was not working with it. From what I have read and my usual procedure with an overexposed image is to use the exposure control to bring down the overall tonality of the image to where it should be. If the highlights are still blown, one can use the recovery slider to further recover highlights without making the overall image too dark.
I'll wipe the egg off my face and do a few more experiments. :)
>The same highlight luminance is being mapped to a higher density on the negative, and the mapped shadow luminance is unchanged. >
Which was my point exactly.
If you are photgraphing objects with an excessive luminance range (like the proverbial black cat in the snow, or higly reflective objects such as silver) on film, you set your exposure level so that detail will be held in the darkest areas and then use soft-working developer and/or cut development time in half so that texture still remains in the highlights.
If you scan a negative that was made like that, reproduction is fairly straight forward.
Even when the HLs in a negative are somewhat blocked-up, the ACR 4.1 Recovery slider does seem to do a fairly remarkable job of pulling some printable textural detail back into those areas.
Digital files (and transparencies) which have blown-out HLs are a different matter.
If you have landed-up with levels of 255 in all channels, there is no texture or detail there to recover. All you will succeed in doing in ACR if you lower the top of the curve or cut the exposure slider back, is to introduce an over-all tone in the HL's rather like the chemical fog you end-up with in the darkroom if you try to "burn-in" the highlights on a print when there is no discernible detail in the solid-blackness of that part of the negative.
ACR 4.1's "Recovery" can work wonders but it can't recover detail in a capture when the photographer left nothing there to recover.
>Which was my point exactly.
>If you are photgraphing objects with an excessive luminance range (like the proverbial black cat in the snow, or higly reflective objects such as silver) on film, you set your exposure level so that detail will be held in the darkest areas and then use soft-working developer and/or cut development time in half so that texture still remains in the highlights.
You are confusing contrast mapping with luminance mapping. By decreasing the development time, you are lowering the contrast but not affecting the dynamic range that is captured.
The Kodak characteristic curve of the T-Max film covers the useful range of the film and the shoulder is not shown, but it does exist. I have taken the characteristic curve for Plus-X and extended it beyond the range that Kodak considers useful so to as to show the shoulder. The actual curve may not be entirely accurate but the concept is valid, and you can see that the original graph shows a roll off of density indicative of the beginning of the shoulder and I have merely extended it.
Let's take the case of your black cat on a background of textured snow. We will further assume that the cat is an area of a shadow cast by a building, and the snow in the background receives full sunlight. You take light readings and find that the luminance of the cat is 1 lux and that of the snow in the shadow area is 1000 lux. The snow in full sunlight has a luminance of 10,000 lux. You expose so that the image of the snow on the film receives 1000 lux/seconds. The image of the cat will receive 1 lux/second. You develop normally. The cat is at the toe of the characteristic curve and the shadowed snow is at the shoulder. However, the sunlit snow will be completely blocked up, since it is on the flat part of the curve. You can adjust the exposure and decrease the development time as much as you want, but the shadowed snow will still be at the shoulder and the snow in full sunlight will still be blocked. The film simply can't handle the dynamic range of the scene.
Ansel Adams adjusted the development time to control contrast, since he liked to print everything on grade 2 paper. When you decrease your development time to allow the highlights to print on grade 2 paper, you could achieve the same effect by leaving the exposure the same and using lower contrast paper.
>If you have landed-up with levels of 255 in all channels, there is no texture or detail there to recover. All you will succeed in doing in ACR if you lower the top of the curve or cut the exposure slider back, is to introduce an over-all tone in the HL's
You have missed my point on why highlight recovery works with raw capture. The highlights may be blown in the rendered white balanced image but not in the raw file as shown below. The highlights in the tiff are completely blown, but in the raw file only the green channel is blown. ACR can reconstruct the green channel from the two that remain.
>When you decrease your development time to allow the highlights to print on grade 2 paper, you could achieve the same effect by leaving the exposure the same and using lower contrast paper. >
You could if anyone made Grade 00000 paper.
In the real world, you adjust the exposure/dev time to enable you to produce a negative that will print the maximum range of tones (preferably without burning or dodging) on your available equipment and materials.
That same low-gamma negative can also be successfully scanned to a Tiff (using appropriate curves in your scanning software) so that both Highlight AND Shadow detail can be rendered as printable by using the advanced ACR 4.1 controls in addition to its "exposure" slider.
I still maintain that you will have more success with recovering HL detail in an overexposed NEGATIVE than you will if you blow out your HL's in either a transparancy or in a digital capture.
I have not "missed your point" on why highlight recovery works with raw capture; but you are making a false comparison between a totally failed scan (your blown-out Tiff) and your over-exposed (but not quite fully burned-out) RAW capture.
The Tiff could, and should, be re-scanned from the original negative (so that it retains HL detail); while your nearly over-blown RAW (while severely pushing the limits!) is not still not totally beyond redemption via ACR's "Recovery".
(But over-expose that RAW file one tiny bit more and I suggest that it would be toast as would a Transparency if the film had been incorrectly over-exposed to the same degree.
>I still maintain that you will have more success with recovering HL detail in an overexposed NEGATIVE than you will if you blow out your HL's in either a transparancy or in a digital capture.
Probably so, since the negative film has a shoulder where the highlights are compressed before they are clipped. I think Peter Lange's suggestion may apply here. You may not actually be recovering highlights, but merely expanding the flat area of the shoulder curve so that what differences are present become more visible.
>I have not "missed your point" on why highlight recovery works with raw capture; but you are making a false comparison between a totally failed scan (your blown-out Tiff) and your over-exposed (but not quite fully burned-out) RAW capture.
I'm not making any false comparison, but merely demonstrating what might happen in the real world and demonstrating the advantage of raw capture over in camera JPEG. If you overexpose a raw file, some degree of highlight recovery is possible. According to Bruce Fraser in his Real World Camera Raw with Adobe CS2, the amount of recovery that is possible depends not only on white balance but how much head room the camera allows. He reported that recovery us not usually greater than one f/stop, depending on the camera.
If you overexpose a JPEG in camera, very little recovery is possible. Most cameras to not make TIFFs, but I used 16 bit TIFFs
The difference lies in the way that a Tiff originated: whether as a scan from an existing fully detailed image (such as correctly-processed negative film); or from an original direct (and possibly bad!) capture in a digital camera.
ACR can be invaluable optimising the final image in the first case; but powerless to do much to redeem the mess in the second one.
Capturing digitally as JPEGs generates its own problems.
I previously posted a demonstration using highlight recovery with ACR using the Exposure control to recover highlights in raw and TIFF image. This expands on the test. I used -1.15 stops of negative exposure to restore highlights, but did not use the recover slider. This reports on how Recovery of 0, 50, and 100 affects the image, which is a Stouffer step wedge. The results are plotted with Imatest and are in the same format as a characteristic curve for film. The X-axis is exposure with large values representing dense areas of the wedge and less exposure. I find this format the easiest to interpret, since pixel values follow a rather strange curve.
The recovery tool itself has little effect on the blown highlights by itself. When used in conjunction with the Exposure slider, the Recovery slider has little effect on the highlights, but acts chiefly to darken the tones extending from 1 to 4 stops below the highlights. Each triangle is 0.1 density unit or 1/3 stop.
I don't really know how this is useful, and perhaps some one can elaborate.
The fact that the shadow density increases but slightly with increased development accounts for the fact that the entire dynamic range of the subject in question is able to be recorded in such a way as to render it printable, which, after all, is the goal of all this manipulation, digital or analog.
A process that, with certain emulsions, actually increases the shadow densities (albeit at the cost of a decreased slope in those zones)can be replicated in RAW by using the parametric Curves and moving both shadow and lower value sliders towards increased brightness. You will observe a bump in the lower densities, just like compensating development.
Prints are an illusion and generating the most tantalizing illusion is what printing means to me. Whatever works. And if, by generating the illusion of a huge range while actually employing a rather restrictive range, well, sometimes the magic works!
FWIW, I think this thread has gone far beyond the question, "What is recovery?"
Sometimes they do in order to get to the real underlying factors.
You can skip the conversation if you wish, and have gotten all there is to get.
No it's just gone into the whole subject of tone-retention in far more depth than you probably expected?
The whole point of the discussions on these Forums is to help ALL readers not just the OP.
>FWIW, I think this thread has gone far beyond the question, "What is recovery?"
Perhaps so, but there is never a simple solution to a complicated problem. Sometimes a bit of background information is necessary and we can still learn from film photography or even the great masters of art painting, for example this link describes lessons from El Greco, Monet, and past masters of black and white film photography.
Still, I don't have a good idea of how to use the Recover slider. Exposure, yes, but recover?
I haven't been using ACR for that long because until CS3 shipped (with the ability to use ACR on scanned film generated 16-bit Tiffs) it was of no use to me.
The way that I am now using ACR 4.1, and the order in which I am doing it (and this is still evolving!) is something like this:
First, make sure first that all "Auto" adjustments are turned off in the Prefs.!
Set the Exposure Slider to display the optimum rendering while checking for Clipping;
Adjust the white balance sliders (which I usually do visually rather than with the eye-dropper);
Go to the Curves tab and usually make a custom one;
Use Recover and Fill to retain as much detail as possible in HLs and Shadows.
Adjust Brightness for the mid-tones;
Move the Black slider in a little to set a Black-point.
Return to the main panel and nudge Contrast and Vibrance as needed.
Go to Lens Corrections tab to deal with Fringing and Vignetting (if any);
Go to Details and set my required values there.
Return to main Panel and adjust Contrast (if neededbecause my Curve has usually taken care of that) and Clarify
Fiddle with individual sliders in HSL and Split Toning if needed.
Go back to Details and adjust the Luminosity & Color Noise sliders as needed.
Regarding Recover, I find that I often use quite high values to retain the HL detail while still allowing plenty of contrast in the mid-tones.
But, as I said, my work-flow is all still very much an on-going experiment but I already am hugely impressed by what can be accomplished using ACR 4.1.
Pretty close to my approach, Ann.
A problem with recover to be aware of is that high adjustment values may cause color contamination to the highlights.
But you probably know that!
Indeed. But you can always History-brush in some de-saturation or color-moderation in Photoshop later if you need to.
The option to DE-saturate (as well as to saturate) highlights and shadows in the Split Toning tab might be a useful addition to the next version of ACR?