Which photo Can survive? Over or Under-exposed photo?

[alireza852019]

Hi...

Which photo has the chance to be recovered more successfully? An "over-exposed" photo or an "under-exposed" one?

Thanks

[dj_paige]

The expose to the right technique indicates that you will get better recovery from an overexposed image (as long as it is not too severely overexposed).

But the great thing about Lightroom is that you can try it yourself and see which one you like better.

[Conrad_C]

Just adding that, thanks to ETTR, some say that even a very bright camera raw image is technically not "overexposed" unless important highlights are clipped. It's actually higher quality, less noisy data.

[Jao vdL · Accepted solution]

Underexposed images are far more likely to be recoverable than overexposed ones. You just get a bit more noise. With sensors from the last several generations this is actually really minor and images that are underexposed by two or even more stops are easily rescuable in post. They will look absolutely identical to the same images just taken at two stops higher ISO. Rarely a problem nowadays. In overexposed images you run a great risk of losing all detail in the highlights. In this way digital capture works differently than film capture which had a soft "rolloff" in the highlights. Digital sensors just cap when they are saturated and information is lost. Some software such as Lightroom employs algorithms to try and recover the highlights but this is very limited and only works when at least one of the color channels is not clipped. Of course if the highlights are OK clipped a bit such as an image taken straight into the sun where that area is just white, that might not be a problem. If it is a white dress where you want to see the texture it would be a real problem and you really want to make sure it is not blown out. Highlight recovery will in general not bring it back except if it is really on the border of clipping.

Bottom line: expose so that the brightest areas that you want detail in are just below blown out. Note that the highlight blinkies on most cameras trigger earlier than the real sensor clipping occurs and most of the time the histogram on camera backs clips on the right side earlier than the actual raw data does. You usually have a bit of headroom assuming you shoot raw. The camera-back histogram and the blinkies are based on the in-camera generated JPEG not on the raw data. A little blinking is not a problem. If large areas are blinking, you will not be able to recover those.

[Bob Somrak]

In general, clipped shadows are less of a problem for the aesthetics of a photo than blown highlights.  I haven't seen too many photos where blown highlights are acceptable and the ones where it is acceptable were planned to be that way.  In situations where the light and subject are consistent and there is no time constraint you can analyze your histogram and do ETTR but for subjects in changing light and subject conditions where you don't have a second chance to take another photo (during a wedding for instance) I would give headroom in the highlights so there is no chance of blowing them out and let the shadows clip.  How you expose needs to be determined for each subject.

[Todd Shaner]

Adding on to Jao vdL excellent advice. Overexposed image files that have fully clipped highlights (R,G,G2,B at maximum level) are NOT recoverable and will have no image detail in those areas. If only one or two channels are clipped LR's Exposure and Highlights controls can partially recover image detail with some color mismatch compared to the surrounding non-clipped area. Whether or not that's an issue is dependent on WHAT is clipped. If it's just bright reflections on metallic and glass surfaces or solid white areas that have no detail then it's usually not an issue.

Underexposed images are fully-recoverable by simply raising the LR Exposure control setting. However, each +1.0 EV Exposure setting increase is the equivalent of raising the camera's ISO setting by a factor of 2x. This increases shadow noise in the image by an amount dependent on the actual camera ISO setting and its sensor noise performance. When shooting at a low ISO setting (100) most cameras maintain good image quality with 1.0 EV underexposure. You may need to add a small amount of Luminance NR and/or adjust Sharpening, but the image quality should still be good.

The problem is that there is no way to determine actual raw file data clipping inside LR. It requires using an external application such as RawDigger, which allows measuring the actual raw file data to determine what pixels are overexposed or underexposed. To determine how your camera's exposure metering behaves try shooting with it set for auto exposure bracketing using -.66 and +.66 offset. Download a trial copy of RawDigger and then check the images for 'Overexposure.' Based on a sampling of subjects and lighting conditions you'll better understand when you need to add manual exposure correction and in what direction. When evaluating image files with RawDigger keep in mind what I said above. Bright reflections on metallic and glass surfaces or solid white areas that have no detail are usually not an issue. If those are the only overexposed areas in the RawDigger image the exposure is OK.

[Conrad_C]

The answer to this used to be very simple: Adding exposure only helps the image, as long as important highlights aren't clipped.

It's a little more complicated now, with the rise of ISO invariant sensors. Today, depending on your camera, it might be OK to keep the camera at its sensor's base ISO, not be concerned about filling the right end of the histogram, and boost the Exposure in Lightroom. Whether you can use this method depends on whether your camera's sensor is ISO invariant. If it isn't, ETTR would produce an image with less noise.

The following link explains it further:

ETTR, ETTL and ISO Invariance: Which is Right for You?

[TheDigitalDog]

1. Exposure only takes place at time of capture. Despite the name of controls in software (Exposure), you can't alter exposure after capture. Exposure is the attribute of how much light strikes the sensor and thus, ONLY shutter and aperture. NOT ISO! That simply affects the meter which can often lie.
2. You can't gauge ideal exposure of raw using any of LR's tools; it has no raw Histogram. You need a raw Histogram to understand the actual effect of exposure on the data, hence a product like RawDigger.
3. Ideal exposure is somewhat subjective but that usually means, NOT clipping highlights where you want detail. You may very well wish to clip specular highlights, you may not. But the only way to determine this is to examine the raw data, NOT a rendering in a raw converter.
4. IF you actual over expose highlight data you didn't want to clip, nothing will bring it back. So in that respect, slight under exposure would be more ideal; you didn't clip that highlight data you didn't wish to clip. But under exposure produces more noise in the capture.

ETTR is an old and kind of bogus term that needs to go away. EOFR (Expose Optimally For Raw) is more appropriate. Exposure is photography 101, no matter if the exposure is a raw, a JPEG, a transparency or a neg. You're not (shouldn't) be using the wrong tool ( a JPEG Histogram) to optimally expose raw then moving 'to the right'. Many of us old timers, shooting transparency film (professionally no less) that had little if any latitude for incorrect exposure ever had or needed an LCD or a Histogram to produce optimal exposure. Neither does anyone else really. The key is understanding how meters can be fooled and when, and how to treat the conditions of the media (again, film or digital; JPEG or raw) along with the behavior of the meter recommendation.

Some areas to read up on in terms of optimal exposure of raw:

https://www.fastrawviewer.com/blog/in-camera-histogram-doesn%27t-represent-exposure

http://digitaldog.net/files/ExposeForRaw.pdf

https://www.fastrawviewer.com/blog/spot-meter-exposure

https://www.rawdigger.com/howtouse/exposure-for-raw-or-for-jpegs 

[Jao vdL]

Thanks for great info everybody. Probably more than was asked for.

Completely agree with digitaldog that ETTR needs to go away in favor of

expose optimally. What has always bothered me is that the explanation

behind ETTR was always based on bogus physics and a misunderstanding of how

light detection in these sensors works. The argument of bits per stop was

always wrong as all that matters is that the size of the bitstep is smaller

than the noise at the light intensity the pixel is detecting. This is

always true with current sensors.

On Thu, Mar 28, 2019 at 12:43 PM thedigitaldog <forums_noreply@adobe.com>

[TheDigitalDog]

https://forums.adobe.com/people/Jao+vdL  wrote

What has always bothered me is that the explanation

behind ETTR was always based on bogus physics and a misunderstanding of how

light detection in these sensors works.

Not the original article that defined ETTR, from Michael Reichmann way back in 2003:

https://luminous-landscape.com/expose-right/

Nothing bogus about what he wrote based on his learnings from Thomas Knoll. It's just time for the term to go away as we have tools and ways to examine the actual raw data to understand exposure. We don't need to futz with moving a lie about the data (a JPEG Histogram) 'to the right'. We simply need to ignore that Histogram and examine one that tells us exactly about the effect of exposure on that data. When Michael wrote this piece, such tools didn't exist.

And they still DON'T exist in ACR or LR and they easily could. 

[Bob Somrak]

Rawdigger and such are fine but these are tools to analyze the photo once it is downloaded from the camera.  They don't do too much good when you are out in the field pushing the shutter button on the camera.  Using these tools to "calibrate" your exposure methods to try to eke the last 1/3 stop of exposure latitude of the sensor is a recipe for a lot of photos with unrecoverable highlights.  I prefer a "little" headroom  to prevent blown highlights and having an acceptable photo instead of a throw away. 

[TheDigitalDog]

https://forums.adobe.com/people/Bob+Somrak  wrote

Rawdigger and such are fine but these are tools to analyze the photo once it is downloaded from the camera.  They don't do too much good when you are out in the field pushing the shutter button on the camera. 

Because such tools are not necessary in the field. Never were. Not when some of us shooting professionally, on transparency film, that required nailing exposure within 1/3 stop did so every day. It's these new digital shooters who can't figure out how to expose without Histograms (worse, those that lie) or LCDs that struggle with exposure. You only need to use these tools to understand how your sensor responds to exposure with your various meters and understand how and when meters lie. It's been done for like 100 years this way Bob. It's hardly rocket science.

[Bob Somrak]

Andrew, you are not the only one here that learned on slide film.  I started shooting Kodachrome in 1970 or so with a selenium meter and learned to get the exposure "right". My point was that is is not necessary to eke out the last 1/3 stop of exposure when shooting digital.

[TheDigitalDog]

So the facts are, we and others can optimally expose, in the field and elsewhere without a Histogram! Yet you seem to suggest otherwise in post #10; which is it?

[TheDigitalDog]

https://forums.adobe.com/people/Bob+Somrak  wrote

My point was that is is not necessary to eke out the last 1/3 stop of exposure when shooting digital.

Optimal exposure is optimal and produces the best quality data. GIGO:Garbage In Garbage Out.

Old saying, it applies to photography (where again, exposure is like lesson 1):

Don't just learn the tricks of the trade. Learn the trade. -James Bennis

[Todd Shaner]

https://forums.adobe.com/people/Bob+Somrak  wrote

I started shooting Kodachrome in 1970 or so with a selenium meter and learned to get the exposure "right". My point was that is is not necessary to eke out the last 1/3 stop of exposure when shooting digital.

I started shooting Kodachrome film around 1974 and it was very unforgiving of overexposure. When taking ColorChecker Passport shots I always shoot .33 EV brackets and very often the -.33 shot is the only one without clipping in the White 95% grayscale patch. Just one of many "unforgiving" digital overexposure examples.

[JP Hess]

Here is an admittedly very dumb answer. I think exposure is a very subjective issue, highly dependent on the photographer's personal taste. I don't think there is really a right answer or a wrong answer. Given the flexibility of the Lightroom tools, I have found that any "reasonable" exposure one way or the other can be optimized and usually made good. But then perhaps I'm not as picky as some. Not very technical, I know. But that's what I have come to accept from this digital mayhem.

[TheDigitalDog]

JimHess  wrote

Here is an admittedly very dumb answer. I think exposure is a very subjective issue, highly dependent on the photographer's personal taste. I don't think there is really a right answer or a wrong answer.

Indeed, as outlined when it was stated that what is clipped (on either end of the tone curve) is up to the photographer to decide. They need to understand the limitations of (in this conversation) the sensor, its DR, where the important part of the image should be rendered, how meters, certainly reflective meters are so easily fooled (white dog on snow, black cat on coal) etc. But the subjectivity can't come about until the photographer understands how to use the tools. It isn't by using a JPEG Histogram or clipping on the LCD when shooting raw.

[Bob Somrak]

https://forums.adobe.com/people/Todd+Shaner  wrote

https://forums.adobe.com/people/Bob+Somrak   wrote

I started shooting Kodachrome in 1970 or so with a selenium meter and learned to get the exposure "right". My point was that is is not necessary to eke out the last 1/3 stop of exposure when shooting digital.

I started shooting Kodachrome film around 1974 and it was very unforgiving of overexposure. When taking ColorChecker Passport shots I always shoot .33 EV brackets and very often the -.33 shot is the only one without clipping in the White 95% grayscale patch. Just one of many "unforgiving" digital overexposure examples.

Slide film was a great learning experience.   When I was shooting slide film I always set the ASA one step higher so I shot my Kodachrome 64 at 80.  I liked the slight underexposure.

As for the Colorchecker example you give, that is EXACTLY my point.  If you are going to try to fill up the "pixel bucket" to the top by "optimizing the sensor 100%" all the time then you are going to spill over the top many times and get blown highlights, no matter how good you think your exposure technique is.  I expose for the situation at hand.  On high dynamic landscapes I will do an ETTR and also bracket.  I prefer doing all my edits using only Lightroom and using ONE exposure but will HDR if necessary or go to Photoshop.  Also, not trying to eke out the last iota of shadow detail is actually a good thing for my taste on most photos. 

[TheDigitalDog]

https://forums.adobe.com/people/Bob+Somrak  wrote

As for the Colorchecker example you give, that is EXACTLY my point.  If you are going to try to fill up the "pixel bucket" to the top by "optimizing the sensor 100%" all the time then you are going to spill over the top many times and get blown highlights, no matter how good you think your exposure technique is. 

Except with that white on the target, you will not. It's not a spectral white, it's Lstar is actually about 96. Optimal exposure would provide some data in each channel. It is also the one patch that varies the most! And there is more than one formulation of this target, the white being again, affected the most.

[Todd Shaner]

Andrew that's not the issue I was addressing. The ColorChecker Passport software and DNG Profile Editor won't run if ANY of the color or grayscale patches are detected as clipped in any of the RGB channels. This is one of many examples where even a small +.33 overexposure can cause big problems. With these subject types it's better to err on the side of under exposure or shoot an exposure bracket.

Color Channels are Clipped when Creating DNG Profile

Re: How to shoot Color Checker for DNG profile editor?

[TheDigitalDog]

https://forums.adobe.com/people/Todd+Shaner  wrote

Andrew that's not the issue I was addressing. The ColorChecker Passport software and DNG Profile Editor won't run if ANY of the color or grayscale patches are detected as clipped in any of the RGB channels.

You mean you'll be unable to build a DCP profile? Again, the white patch shouldn’t be clipped, it's not close to a specular highlight, that is if your goal is to capture it as it exists (Lstar approximately 96). Sure, you can clip it and sure, you can expose so you don't. But the white patch isn't pure white and optimal exposure wouldn't clip any of the three channels on that target. And as I think you're expressing, if your goal is to use it to build a camera profile you shouldn't. So if you end up with that white patch at Lstar 96, you didn't do anything but expose it optimally. If you clipped it, you didn't.

[Todd Shaner]

thedigitaldog  wrote

https://forums.adobe.com/people/Todd+Shaner   wrote

Andrew that's not the issue I was addressing. The ColorChecker Passport software and DNG Profile Editor won't run if ANY of the color or grayscale patches are detected as clipped in any of the RGB channels.

You mean you'll be unable to build a DCP profile? 

Yes, exactly!

Again we are going way off-topic! The OP alireza852019 hasn't posted a single reply in this thread concerning the original very simple question:

Which photo Can survive? Over or Under-exposed photo?

The simple answer is UNDEREXPOSED image files in most circumstances will survive better than overexposed image files with clipped highlights. The ColorChecker shot is a good example of why even a small amount of overexposure can cause big problems. As already mentioned even slightly overexposing a white wedding dress may have the same affect with fine detail totally clipped and unrecoverable.

Enough said guys–We've beat the subject to death. Let's wait for the OP to reply!

[TheDigitalDog]

And optimally exposed images are better than either.

[Bob Somrak]

Quote isn't working today for me

"And optimally exposed images are better than either."

Best answer in the thread thedigitaldog​