What does 'hi' means?

[fjrabon]

I used an example of 3 stops underexposed. The example Destin gave was a whopping 6 stops underexposed. You would not be able to restore enough detail from an image 6 stops underexposed. You are mistaken.

What would the shadows be in an image that's 6 stops underexposed? Clipped? Plugged? Unusable? All of the previous.

look at the actual example I gave above. Same picture, same settings, one shot at ISO 100, one shot at ISO 3200, that's 5 stops of ISO. The details were completely there, and the image wasn't even awfully noisy. Then I also reduced the shutter time by almost 300% and still, all the details were there, though noisy. That was shooting a black guitar amp. That's because Destin's camera creates a lot of noise, which makes the images unusable faster. Since the OP was about a D7000, the difference would even be more extreme in how many stops you can recover.

Again, in those images, no detail was lost, noise just started to be introduced. That's the difference between the left side and the right side. Highlights are 100% GONE when they clip. They can't even sort of come back. On the left side things just start to get noisy, how noisy depends on 4 factors 1) how much noise your camera creates through its physical operation. 2) How much noise is in the environment (an area with lots of electronics will produce a lot more extraneous environmental noise, as stray electrons hit the sensor) 3) how good your noise reduction algorithms are. and 4) how much you had to amplify the signal.

You don't clip on the left side. That doesn't even make sense. Look up what clipping means. The only way to clip on the shadows is to have your black point set to create purposeful clipping, or if your sensor just doesn't pick up photons below a certain intensity (which, for modern DSLRs, is an EXTREMELY LOW threshold, for all intents and purposes, it's nonexistent on a modern DSLR sensor). What Destin has been talking about is getting to the point where his noise levels are so high that the signal begins to get lost, which is an entirely different concept.

 

[o hey tyler]

I used an example of 3 stops underexposed. The example Destin gave was a whopping 6 stops underexposed. You would not be able to restore enough detail from an image 6 stops underexposed. You are mistaken.

What would the shadows be in an image that's 6 stops underexposed? Clipped? Plugged? Unusable? All of the previous.

look at the actual example I gave above. Same picture, same settings, one shot at ISO 100, one shot at ISO 3200, that's 5 stops of ISO. The details were completely there, and the image wasn't even awfully noisy. Then I also reduced the shutter time by almost 300% and still, all the details were there, though noisy. That was shooting a black guitar amp. That's because Destin's camera creates a lot of noise, which makes the images unusable faster. Since the OP was about a D7000, the difference would even be more extreme in how many stops you can recover.

Again, in those images, no detail was lost, noise just started to be introduced. That's the difference between the left side and the right side. Highlights are 100% GONE when they clip. They can't even sort of come back. On the left side things just start to get noisy, how noisy depends on 4 factors 1) how much noise your camera creates through its physical operation. 2) How much noise is in the environment (an area with lots of electronics will produce a lot more extraneous environmental noise, as stray electrons hit the sensor) 3) how good your noise reduction algorithms are. and 4) how much you had to amplify the signal.

You don't clip on the left side. That doesn't even make sense. Look up what clipping means. The only way to clip on the shadows is to have your black point set to create purposeful clipping, or if your sensor just doesn't pick up photons below a certain intensity (which, for modern DSLRs, is an EXTREMELY LOW threshold, for all intents and purposes, it's nonexistent on a modern DSLR sensor). What Destin has been talking about is getting to the point where his noise levels are so high that the signal begins to get lost, which is an entirely different concept.

Are we looking at the same images? There is absolutely no-way-I-would-ever touch the underexposed shot at ISO100. Maybe to you, the "details" are there. To me, it's a right click-delete.

It's interesting that Adobe calls it highlight and shadow clipping. Regardless of what anyone else defines "clipped" as, I see it as having an area of the image too dark, or too bright to warrant an exposure adjustment in an effort to fix it. Like the one you posted, where all of the shadows were clipped at ISO 100 and then "brought back" in post.

 

[fjrabon]

Are we looking at the same images? There is absolutely no-way-I-would-ever touch the underexposed shot at ISO100. Maybe to you, the "details" are there. To me, it's a right click-delete.

It's interesting that Adobe calls it highlight and shadow clipping. Regardless of what anyone else defines "clipped" as, I see it as having an area of the image too dark, or too bright to warrant an exposure adjustment in an effort to fix it. Like the one you posted, where all of the shadows were clipped at ISO 100 and then "brought back" in post.

I never once said they were pictures you'd want to use. What I said is that the details weren't lost in the same way you lose highlights, they're not GONE. You can clip shadows, but that involves setting your black point to clip them at that point ON PURPOSE. What Adobe is referring to is values below the defined black point, which is the point that the software turns any close to black value into pure black. If you totally remove the black point, to where any non black value point, no matter how dark, is still non black, you can 'bring it back (along with a boatload of noise, yes). If you define the black point all the way over, like magic, your entire image is shadow clipped. Thus, in camera, you don't really have shadow clipping, except in a really extreme case.

 

[Destin]

Are we looking at the same images? There is absolutely no-way-I-would-ever touch the underexposed shot at ISO100. Maybe to you, the "details" are there. To me, it's a right click-delete.

It's interesting that Adobe calls it highlight and shadow clipping. Regardless of what anyone else defines "clipped" as, I see it as having an area of the image too dark, or too bright to warrant an exposure adjustment in an effort to fix it. Like the one you posted, where all of the shadows were clipped at ISO 100 and then "brought back" in post.

I never once said they were pictures you'd want to use. What I said is that the details weren't lost in the same way you lose highlights, they're not GONE. You can clip shadows, but that involves setting your black point to clip them at that point ON PURPOSE. What Adobe is referring to is values below the defined black point, which is the point that the software turns any close to black value into pure black. If you totally remove the black point, to where any non black value point, no matter how dark, is still non black, you can 'bring it back (along with a boatload of noise, yes). If you define the black point all the way over, like magic, your entire image is shadow clipped. Thus, in camera, you don't really have shadow clipping, except in a really extreme case.

If it can't be applied to real world photography then why does it matter. If the shadows are too dark to be saved and TURNED INTO A USABLE IMAGE, then in my book they are clipped, because they are dark enough to force to to delete the image.


What you're saying is like saying that I baked a birthday cake for my mom, and burned it to a crisp. It's still a cake sure, but it's COMPLETELY useless and the fact that it is a cake doesn't matter anymore.

Or I get in a car accident and total my car. Hey, all the parts are still there. Can't drive it but it's all still there. USELESS.

 

[Dao]

fjrabon, maybe take a look at this Optimizing Exposure (Which is a revised article of "Exposure to the right")

 

[fjrabon]

If it can't be applied to real world photography then why does it matter. If the shadows are too dark to be saved and TURNED INTO A USABLE IMAGE, then in my book they are clipped, because they are dark enough to force to to delete the image.


What you're saying is like saying that I baked a birthday cake for my mom, and burned it to a crisp. It's still a cake sure, but it's COMPLETELY useless and the fact that it is a cake doesn't matter anymore.

Or I get in a car accident and total my car. Hey, all the parts are still there. Can't drive it but it's all still there. USELESS.

no, what I'm saying is that there's a difference between raw cake, which is still there, though possibly not practically usable, and completely evaporated cake, that isn't even there any longer. On the left side, while the information may be noisy to the point of being unpleasant, the information is still there. On the right side, when it clips, it's purely gone. You can make a black sky blue and noisy, but a white sky is only ever going to be white, it's gone.

And yes, there is a practical "real world application" of this concept, it's just less extreme. It's the idea of shooting purposefully underexposed. Pros do it all the time in certain situations. Sure, you don't always do it, there's no reason to put yourself through exposure correction when you don't have to, but in certain situations, they like the enhanced dynamic range you get when underexposed. Sure, they would never underexpose by 5 stops (or even 2 stops), they'd use HDR in those situations, because it would provide the same effect, but with less noise.

Also, as technology gets better, the amount we can restore from the left side keeps growing and growing. 10 years ago, one full stop underexposed was COMPLETELY unusable. Whereas, today professional photographers will often shoot a full stop underexposed on purpose, to avoid clipping highlights. THey usually won't think about needing HDR unless they get past a full stop of exposure. The difference in number of stops you can recover on a D7000 and your camera is pretty substantial.

My whole point, this entire time was at the right side you run into clipping, on the left side you run into noise. They're simply different concepts.

 

[fjrabon]

fjrabon, maybe take a look at this Optimizing Exposure (Which is a revised article of "Exposure to the right")

yeah, I've read that before. If you're not blowing out highlights, you'd want to overexpose and fix in post, for an ideal exposure. The technical best way to expose is to expose so that the brightest spot is just short of blowing out, and then fix it later. THough in most cases, going to that extreme is not really necessary, and just causes you to waste time in post when you wouldn't really need to.

Essentially, that article is making the same point I've been making. the left side of the histogram means you run into noise. The signal to noise ratio on the right side of the histogram is higher, because the signal is higher, while the noise (denominator) is basically the same. The problem with the right side though is you hit a hard stop when you max out the brightness, or clip (blow out) the highlights in the camera. If you have a small dynamic range, you could technically get a better picture by overexposing it so that the highlights fall just short of clipping and then adjusting in post so that it looks correct in post, you'll have some unmeasurably lower amount of noise in your picture that way. Practically that's probably more of a pain than it's worth, in most cases.

The problem comes when you have a high dynamic range. In that case you're better off underexposing (unless you're fine with clipping the highlights) and then dealing with the noise. You can gain some additional dynamic range, with a trade off of additional noise, by underexposing. Which is what I've been saying all along. Obviously, at this point in technology, you're probably best off dealing with the issue by using HDR if the amount you have to underexpose is a full stop or more.

 

[Destin]

If it can't be applied to real world photography then why does it matter. If the shadows are too dark to be saved and TURNED INTO A USABLE IMAGE, then in my book they are clipped, because they are dark enough to force to to delete the image.


What you're saying is like saying that I baked a birthday cake for my mom, and burned it to a crisp. It's still a cake sure, but it's COMPLETELY useless and the fact that it is a cake doesn't matter anymore.

Or I get in a car accident and total my car. Hey, all the parts are still there. Can't drive it but it's all still there. USELESS.

no, what I'm saying is that there's a difference between raw cake, which is still there, though possibly not practically usable, and completely evaporated cake, that isn't even there any longer. On the left side, while the information may be noisy to the point of being unpleasant, the information is still there. On the right side, when it clips, it's purely gone. You can make a black sky blue and noisy, but a white sky is only ever going to be white, it's gone.

And yes, there is a practical "real world application" of this concept, it's just less extreme. It's the idea of shooting purposefully underexposed. Pros do it all the time in certain situations. Sure, you don't always do it, there's no reason to put yourself through exposure correction when you don't have to, but in certain situations, they like the enhanced dynamic range you get when underexposed. Sure, they would never underexpose by 5 stops (or even 2 stops), they'd use HDR in those situations, because it would provide the same effect, but with less noise.

Also, as technology gets better, the amount we can restore from the left side keeps growing and growing. 10 years ago, one full stop underexposed was COMPLETELY unusable. Whereas, today professional photographers will often shoot a full stop underexposed on purpose, to avoid clipping highlights. THey usually won't think about needing HDR unless they get past a full stop of exposure. The difference in number of stops you can recover on a D7000 and your camera is pretty substantial.

My whole point, this entire time was at the right side you run into clipping, on the left side you run into noise. They're simply different concepts.

And It's still bad advice you're dishing out. Underexposing in camera causes issues with white balance, color cast, etc. Underexposing by a half stop to avoid highlight clipping is NOT the same thing as you've been arguing. Also, exposing the the left is NOT all that common of a thing to do. ETTR is MUCH more common, because it reduces the amount of visible noise in a photo.

I'm done with this thread because we're arguing something that is completely pointless. There is simply no reason to underexpose by 6 stops (that was the initial arguement in this thread, since you apparently didn't bother reading the first two pages) in camera by using a lower iso. It's better to get your exposure where you want it in camera. Always. Every time. No question.

 

[fjrabon]

And It's still bad advice you're dishing out. Underexposing in camera causes issues with white balance, color cast, etc. Underexposing by a half stop to avoid highlight clipping is NOT the same thing as you've been arguing. Also, exposing the the left is NOT all that common of a thing to do. ETTR is MUCH more common, because it reduces the amount of visible noise in a photo.

I'm done with this thread because we're arguing something that is completely pointless. There is simply no reason to underexpose by 6 stops (that was the initial arguement in this thread, since you apparently didn't bother reading the first two pages) in camera by using a lower iso. It's better to get your exposure where you want it in camera. Always. Every time. No question.

well, you're actually making two different arguments there, because ETTR is about getting a 'wrong looking' exposure in camera and then normalizing it in post. If you're using ETTR and taking a picture of a black sock, you'd expose it to be really light greay (just short of pure white), and then normalize it back to black in post. You're calling making a black sock really light grey "getting the exposure you want in the camera"?

I never debated, or argued for the original argument. My point has always been that if you have to choose between clipping highlights and noise, noise is easier to fix, because clipped highlights are impossible to fix. I never once stated that you'd WANT to underexpose by 6 stops, I just said that you can recover the image (and every single time I said, WITH A LOT OF NOISE) by several stops.

There are in fact times when ETTR would even underexpose compared to what would be considered a 'correct' exposure. If you've got one really bright highlight, and an otherwise dark picture, ETTR would actually underexpose compared to what 'normal' exposing would do, because it would set the brightest point just short of clipping. Whereas normal exposure would probably just clip the highlight so everything else was neutral grey.

 

[fjrabon]

Not to get into a technical discussion, but increasing the ISO does not make the sensor more sensitive to light. It doesn't change anything about the sensor itself (aperture and shutter speed control how much light the sensor gets). The ISO setting tells the camera how much to amplify the signal when it processes it to a digital file. Depending on the camera you are using, it may be better to shoot with a higher iso, or on some new models (nikon d7000 for example) it may be better to shoot at iso 100 or 200 and then adjust the image on the computer instead (so that highlights aren't clipped like they would have been had the camera done the processing).

also, I think you've been attacking a straw man, Destin. As far as I can tell, from the very beginning, shooting underexposed was only ever about avoiding clipped highlights. The above is a quote from the first post about this. I didn't see anybody ever advocating shooting 5 stops under exposed just for the heck of it. It was always about avoiding clipped highlights. I was never arguing that you should always shoot underexposed for no reason, because the left side is better. I use ETTR in a lot of situations, basically every time the dynamic range is at all an issue. Essentially, part of what I've been arguing is pro-ETTR, because ETTR is about not having clipped highlights as well as bumping up things when they aren't coming close to clipping on the right. If you use ETTR, sometimes you'll actually get an image that is considered classically underexposed using the traditional matrix middle grey method, if the highlights are really bright.

 

[EchoingWhisper]

BS. YOU CAN ABSOLUTELY CLIP SHADOWS. If you take a photo that is severely underexposed the shadow areas will be SO DARK THAT THE CAMERA DOES NOT RECORD DETAIL IN THEM.

The left side only gets blocked out when the amount of photons hitting the sensor falls to an EXTREMELY low number. You can bring back a shadowed out area most of the time, if you're willing to deal with the noise. And with today's digital noise removal improvements and sensor improvements, that's going away faster and faster. Whereas a highlight, when it's gone, it's gone.

You mean like shooting a scene at f/4, 1/30th, and iso 100, when the proper exposure is f/4, 1/30th, and iso6400. I'm pretty sure you just proved the point that i've been trying to prove throughout this entire thread for me.

Please re-read what rsbones wrote - you capture the same amount of shadow detail no matter what ISO you set your camera on as long as the exposure (aperture and shuttere speed) and available light is the same.

 

[EchoingWhisper]

You mean like shooting a scene at f/4, 1/30th, and iso 100, when the proper exposure is f/4, 1/30th, and iso6400. I'm pretty sure you just proved the point that i've been trying to prove throughout this entire thread for me.

no, you'd be able to pull most of the shadows up under those settings with a raw image.

no, you wouldn't. I've tried it several times and you can't.

You could.

 

[Destin]

The left side only gets blocked out when the amount of photons hitting the sensor falls to an EXTREMELY low number. You can bring back a shadowed out area most of the time, if you're willing to deal with the noise. And with today's digital noise removal improvements and sensor improvements, that's going away faster and faster. Whereas a highlight, when it's gone, it's gone.

You mean like shooting a scene at f/4, 1/30th, and iso 100, when the proper exposure is f/4, 1/30th, and iso6400. I'm pretty sure you just proved the point that i've been trying to prove throughout this entire thread for me.

Please re-read what rsbones wrote - you capture the same amount of shadow detail no matter what ISO you set your camera on as long as the exposure (aperture and shuttere speed) and available light is the same.

Basically yes, the same amount of light hits the sensor. However at iso 100, 6 stops underexposed, you will NOT be able to recover them without getting an INSANELY noisy, un-recognizable photo. I've read through this thread several times. Done tests. Viewed others' tests. Read other threads on other forums. I, along with most other photographers in the world, agree that this whole argument is pointless right now. Sure, the same detail may be there in theory. But by the time you adjust the exposure in post, you've introduced so much noise and screwed up the colors so much that the image is unrecognizable and ultimately unusable. Maybe in 10 years technology will have advanced to a point where iso is simply something you can change in post, but we aren't even close to that point yet.

 

[EchoingWhisper]

BS. YOU CAN ABSOLUTELY CLIP SHADOWS. If you take a photo that is severely underexposed the shadow areas will be SO DARK THAT THE CAMERA DOES NOT RECORD DETAIL IN THEM.

One more thing - if you read my previous post - underexpose by using a lower ISO wouldn't clip shadows. I didn't say that it was not possible to clip shadows - what I said may not be what you're thinking. What I meant is that underexposing by using a lower ISO wouldn't clip more shadows compared to using a higher ISO. A monitor capable of displaying high bits would show that it isn't really clipped.

I TOTALLY disagree. Again, if I take a photo at f/4, 1/30th, and iso 100, but the proper exposure (per the histogram) for said image would be f/4, 1/30th, and iso 6400, THE EXPOSURE TAKEN AT ISO 100 WILL LOSE DETAIL IN THE SHADOW AREAS. There is no way you can possibly argue against this. You just can't.

You're wrong again. Please read the links rsbones gave or just Google ISOless sensor and read their entire thread and please confirm what you know - proof. Your example has the same amount of available light. You can't possibly lose detail in the shadow area by using a lower ISO if your camera is using high bit eg. 14 bit or your camera has low read noise. The reason you couldn't recover details from the shadow areas is because either your software isn't capable of doing it or your camera has high read noise. The D80 has definitely high read noise. No one would ever imagine not raising the ISO when the picture is dark just a generation ago - because the read noise was so high it was best amplifying the signal in camera.

 

[EchoingWhisper]

Okay, we're arguing slightly different concepts here. Let's take it to the REAL world, and not technical BS.

Are the details there? Maybe. Let's for the sake of argument say they are. Because I for one don't care if they're there or not. It doesn't matter.

Why? Because there is NO way that you'll ever restore them without introducing enough noise to render the image completely useless. I don't care how good your software is. Example:

First image: Take at iso 3200 on my Nikon D80. Admittedly crappy noise performance but that should only help your argument because the next image is taken at iso 100, where my camera produces clean photos. For the sake of my argument, this first shot is straight out of camera, just converted to JPG. Both are 100% crops btw.



Image 2, taken at iso 100 on my D80. Upped the exposure and fill light to balance the histogram with image one, AND applied the strongest noise reduction settings in lightroom to try and restore it to a useable image. You're still left with Chicken **** for an image. There is no recovering more detail out of it. There is no way to fix the noise to make it useable. The first shot however, I could apply some noise reduction and make it a somewhat useable image.


Now don't go and say the results in the second image would be better with a different camera. That's bogus because the image came out of the camera clean because it was taken at iso 100.

low ISO =/= low noise. You seem to have a lot of misconceptions about what causes noise, what clipped out means, etc. That seems to be where all your confusion is coming from. The statement "That's bogus because it came out of the camera clean because it was taken at ISO 100" doesn't make any sense at all. What do you even mean by "came out of the camera clean"? Your camera takes clean photos at ISO 100 WHEN PROPERLY EXPOSED. High ISO can amplify noise, but that doesn't mean that low ISO's don't also amplify noise, they're just less sensitive to it, so it would take longer exposure to the noise for it to be noticeable with a low ISO. ISO doesn't make noise, cameras and environment make noise. ISOs, in digital cameras at least, AMPLIFY noise. WHen you brighten a picture with software, you're essentially doing the same thing, you're amplifying the lowest light signals from the sensor to be brighter. So, if your camera produced a lot of noise, it doesn't matter what ISO you shot it at, the noise your camera created will be amplified (along with any environmental noise). Again, ISO doesn't create noise, so saying your camera takes clean shots at ISO 100 is completely meaningless. All that matters is how much that noise was amplified, and how good the programs that reduced it were.

This is the whole point that everybody has ben trying to make. If you have low enough noise, you can underexpose almost much as you want to. There is no such thing as 'clipping' on the left side of the histogram. Sure, with cameras that create lots of noise the left side can become pretty unusable pretty quickly. And some older sensors don't pick up low levels of light very well, especially in short exposures, meaning that they can run into noise problems really fast from environmental noise. But as long as the sensor registered photons, you don't lose that information, and it's certainly not like you lose information on the right side, where it just disappears, as the sensor cannot physically record above a certain level.

Clipping has to do with overloading the sensor on the right side of the histogram. The left side being unusable has to do with the signal to noise ratio.

Then why does just about EVERY professional photographer use the ETTR method rather than just shooting everything at iso 100 and fixing it later. THERE IS NO REAL WORLD APPLICATION FOR WHAT YOU ARE ARGUING, AS IT CAN BE DONE BETTER BE EXPOSING PROPERLY IN CAMERA IN THE FIRST PLACE.

ETTR is correct. Using a lower ISO is not exposing the right. It is best getting the most information you can have exposing to the right - but not by using a higher ISO.