Dynamic Range-Can't they or Won't they?

dexter75 wrote in post #17741894
Said it before and I'll say it again. DR is the most overrated argument in the world of amateur internet forum photographers. No pro sits around around studying DR specs on DxO and making their decisions based upon it. No pro shooting Canon is fretting because their camera has a whole stop less DR than a Nikon or Sony. Its almost as ridiculous as the megapixel wars of years past where people thought more megapixels clearly meant better quality photos, same with DR. The only people who should even be slightly concerned about this are pro landscape photographers being commissioned by art galleries to make huge prints for lots of money. I think pretty much no one here falls into that category, so just stop

In the last two years I have seen a sea change in the Landscape photography world with the large majority of "pros" that were still shooting Canon moving to either Sony or Nikon. I know some that are still shooting Canon, but they are now a rare breed and if the 5D4 doesn't deliver the goods, that may change. And the reason for the switch isn't because they "make huge prints for lots of money". It's for workflow reasons, which benefits anyone who takes landscape images. That includes me. And really, pretty much anyone else that shoots landscapes. Being able to take one shot and push the shadows so that you have a clean image with all of the DR that you need is a much more elegant solution than having to deal the with the issues that can be caused by using filters or blending separate shots.

Wilt wrote in post #17742395
And I'll make a request of Tareq, who is uniquely outfitted to do this kind of comparison...
We hear that medium format digital has a very wide DR due to the difference in the sensor technology (CCD vs. CMOS). So...
Can you shoot a wide DR shot on both medium format and with existing Canon gear, and post any comparative sections from both which shows how the wider DR medium format shot results in any presentable (on monitor) detail which is NOT VISIBLE in the Canon shot?

Sorry, i am not a trusted member or shooter on the net, i did post different tests here and there but nothing is interesting at all, so i am not interested in this request by you.

Secondly, there are some pro well known photographers did this kind of comparisons and they are more access than me and better gear, so you watch them and i am sure you saw their findings so far, why me? even if i do this comparison for you you will find many mistakes or even excuses to say that you didn't see any difference in prints or on monitors, if your point is valid then i request that medium format must disappear from the world and we all shoot 35m DSLR, or even point and shoot these days are rivals to medium formats and DSLRs so no need for DSLRs too, if the war about DR and mp is useless/worthless then we must all go back and use 10D or Coolpix old generation and it will be great, i don't know why many go from 5Dc to 5D2 or 5D3 if that 12mp is enough or overkill, even my 1D2n 8mp is perfect and i don't need 1DX, so comparison between old and new or small and large will not prove anything rather than if someone can afford something or not more than someone need this or not.

Wilt wrote in post #17742395
Dexter has been taking some recent heat for his statements, but I think the one he makes above is on target! After all, pros shooting for any print media (advertising, company brochures, 10K reports, product literature) have to reduce the DR of a scene into a range which can be offset printed -- which is barely even 6EV of DR -- regardless of B+W vs. color transparency, this has always been a limitation! So then that raises a response from folks, "But if I can capture a wider DR, I can nevertheless compress it to fit my output". That raises the reaction that you hear from others, when HDR techniques have been applied to a shot, "It (HDR) looks artificial!".

That leads me to offer this challenging question, to hear the responses:

So if DR compression (e.g. HDR) results in artificial appearance, and our media (offset printed page, our monitors, even photographic prints) are inherently 'limited DR' media, just why is it so necessary to get any more than 12EV of DR than can be accomplished today (via Sony sensor)?!

And I'll make a request of Tareg, who is uniquely outfitted to do this kind of comparison...
We hear that medium format digital has a very wide DR due to the difference in the sensor technology (CCD vs. CMOS). So...
Can you shoot a wide DR shot on both medium format and with existing Canon gear, and post any comparative sections from both which shows how the wider DR medium format shot results in any presentable (on monitor) detail which is NOT VISIBLE in the Canon shot?

I dont really follow you on the offset printing stuff, but say I process my image, and pull shadows, if banding shows up on the shadows, it may very well show up on print. Same with noise, however noise tends to be minimized somewhat.

I'm not going to debate the necessity of 12 EV, but if one company can do 12 and the other can do 14, then it's an advantage worth noting. Same goes with high ISO. If I can get usable 12800 vs 6400, then it's worth noting.

The pulling up of shadows with Canon, where it produces banded noise, seems to be very much reduced, if not possibly eliminated. There is still more noise in the pulled shadows than a competitor, for example, but it isn't the ugly banding that you have seen in the past, which was nearly impossible to clean up with post processing techniques.

This just leaves us with the issue of more random noise found in pulled shadows than competitors, which is still an issue, but not the huge one that previously existed.

yes, I do understand that banding is pretty much gone at this point, just a note to wilt than DR can come in handy if opportunity arises.

davesrose wrote in post #17742441
I've found since upgrading from the 5Dc to 5D3, I did see an increase in exposure latitude. I could shoot an outdoor scene with bright light and shadows, and get a better range of contrast between white and black (that I could expose more to not clip as much highlights). Only an improvement in DR can contribute to this. If it were just getting a higher bit ADC, and not improvement in sensor, then you would have the same amount of blown highlights (areas of luminance of scene that were not recorded).

Counterpoint: We see professional evaluations of Sony or Nikon using Exmor sensor, and the evaluations point to NOT increases in brightness range (ability to photography brighter objects with detail) but more at the LOWER end of the range of luminances...the detail in the low luminance zones captured without noise/banding.

Yes, because all digital cameras have sensors limited to 14 bits worth of tonal data. I have still seen instances with the D810 that has clipping in highlight areas...for certain scenarios, I believe there still needs to be better DR for any camera sensor.

Wilt wrote in post #17742395
Dexter has been taking some recent heat for his statements, but I think the one he makes above is on target! After all, pros shooting for any print media (advertising, company brochures, 10K reports, product literature) have to reduce the DR of a scene into a range which can be offset printed -- which is barely even 6EV of DR -- regardless of B+W vs. color transparency, this has always been a limitation! So then that raises a response from folks, "But if I can capture a wider DR, I can nevertheless compress it to fit my output". That raises the reaction that you hear from others, when HDR techniques have been applied to a shot, "It (HDR) looks artificial!".

That leads me to offer this challenging question, to hear the responses:

So if DR compression (e.g. HDR) results in artificial appearance, and our media (offset printed page, our monitors, even photographic prints) are inherently 'limited DR' media, just why is it so necessary to get any more than 12EV of DR than can be accomplished today (via Sony sensor)?!

This is true and you bring up a good question. Its simply not necessary. Ive been a pro photographer for 11 years, meaning this is my only source of income for 11 years. I don't have a 9-5 and shoot for fun or side money on the weekend. This is my career and Ive been quite successful and made good money doing it. Ive had agencies and clients tell me they don't like the look of high DR photos as they tend to look too "cartoonish" or too "digital" are words Ive often heard used. Most prefer more of a film look which is why many high end fashion shooters still shoot MF film to this day. Of course this may vary based on what you shoot, but for what I do, thats how it is. Yes, I have taken some heat but its cool. When people who shoot as a hobby who have no idea what its like to have your income depend on your photos or what dealing with demanding clients is like tells me Im wrong, sorry if I don't agree. Doesn't mean Im right either, its just my opinion and its based on my experiences. No reason to get heated, say you don't agree and move on.

So I myself have posted a few times in these threads, and agreed that the final result is not likely to show a need for much more DR. I also do agree with Dextrer that the need for DR seems over stated.
I often bring up slides..

We like contrast, we like punch, we add it in post a LOT.

However, Bcaps brings up a very good point that also sounds LOT like my own defense of larger color Gamut, 16 bit Tiff etc.

ie: more in the beginning is better even if you are going to end up printing 8 bit sRGB.

In this case, more DR in the beginning making for a better less clipped work flow from the beginning even if we end up with 9 stops of DR in out final out put.

I can not argue the premise as i always argue the same one for 16 bit/Pro?Adobe RGB etc.

Seems in keeping with a general philosophy I have adhered to since about 2003.
Better is just better.

No?

Wilt wrote in post #17742395
Dexter has been taking some recent heat for his statements, but I think the one he makes above is on target! After all, pros shooting for any print media (advertising, company brochures, 10K reports, product literature) have to reduce the DR of a scene into a range which can be offset printed -- which is barely even 6EV of DR -- regardless of B+W vs. color transparency, this has always been a limitation! So then that raises a response from folks, "But if I can capture a wider DR, I can nevertheless compress it to fit my output". That raises the reaction that you hear from others, when HDR techniques have been applied to a shot, "It (HDR) looks artificial!".

If we accept that monitors (and especially print media) have limited DR, then we must always be looking for ways to represent an image that could have too much DR for that output medium. A simple curves or contrast tweak could be sufficient, or dragging the shadows and highlights sliders in Lightroom. Or we could look into HDR techniques.

When people talk about the "HDR look" they are (probably without realising) usually referring to the tone mapping algorithm that was used to map a high DR input image into a lower DR output. The most commonly used technique (at least in the early days) was one where brightness levels are manipulated in specific areas in relation to other local detail (often called "local contrast" or "local adaptation") - in order to fool your eye into thinking there's a larger DR in the image than is actually being displayed. It's a similar visual trick to showing a dark grey square on a black background next to another dark grey square on a white background - the squares look different, but it's your brain playing a trick on you.

When pushed hard, those local contrast techniques result in weird looking halos around objects, and give that hyper-real "grungy" HDR look.

But... there are more subtle techniques, so compressing a large DR range down to something that can be displayed can be done without it looking overly artificial.

Wilt wrote in post #17742395
That leads me to offer this challenging question, to hear the responses:

So if DR compression (e.g. HDR) results in artificial appearance, and our media (offset printed page, our monitors, even photographic prints) are inherently 'limited DR' media, just why is it so necessary to get any more than 12EV of DR than can be accomplished today (via Sony sensor)?!

As noted, DR compression doesn't necessarily guarantee an artificial result - there are many ways of skinning a cat.

As to why you'd want to capture more than 12EV: when you observe a scene, you look at different elements, your eyes are adjusting to the light levels in each area, and your brain is putting together an image of the whole. If I'm standing in a dark building and looking at the wall/window frame, then look outside to a bright outdoor scene, I don't "see" a clipped outside any more than I "see" a pitch black interior. If I took a single photo of the whole though, I'd get one of the two (depending on my exposure).

More DR at capture allows a better chance of a recreation of what we "see". Whether the tone mapping techniques used to create the final image are pleasing/realistic/art​ificial is down to the intention of the shooter (and his PP skill) and the artistic interpretation of the viewer.

davesrose wrote in post #17742656
Yes, because all digital cameras have sensors limited to 14 bits worth of tonal data. I have still seen instances with the D810 that has clipping in highlight areas...for certain scenarios, I believe there still needs to be better DR for any camera sensor.

If you have an image that has clipping in highlight areas where you wanted detail, the shot was overexposed. Simple as that.

Remember that highlight recovery is a red herring; a pixel is either over exposed (clipped) or it's not. The issue is that the histograms in our cameras usually give us data based on the JPEG rather than the raw, so detail that might be flagged as clipped in-camera may (or may not) be clipped in the raw file. If our cameras would give us raw histograms then there'd be no ambiguity (and what was clipped would really be clipped - no chance of recovery).

Circling back - the solution to a photo with clipped highlights is obviously to reduce the exposure sufficiently until any (desired) highlights are no longer clipped. That of course reduces the exposure of all areas in the image, meaning that some/much of the image may be in lower stops and thus has a worse signal to noise ratio.

If you then push (brighten) that darker detail in post and it's visually unacceptable due to noise; that's insufficient DR.

In a nutshell: the more DR we have at capture, the better chance you have in holding detail in your highlights (due to taking a darker exposure) whilst still getting acceptable quality in shadow areas.

sploo wrote in post #17742803
If you have an image that has clipping in highlight areas where you wanted detail, the shot was overexposed. Simple as that.

But the point is that there's still situations where an exposure will clip in either your highlights, shadows, or both. It was true with film, and it's true with any photographer shooting RAW: the photographer finds a full exposure range that gets you as much tonal range as possible. Lets say I do shoot a scene that I notice is clipped in the highlights....I then "underexpose" further down to get more detail in the highlights. If my scene has a higher luminance range then what my sensor is able to record, then if I only expose for highlights, I will get blocked up shadows (and won't be able to "recover" what's not there). From my experience with the D810, there is just a little bit more wiggle room post processing shadows in photos ...it does not always eliminate the need for multiple exposures in a HDR scene, though.

I think another factor, that leads to other arguments, is tone-mapping a RAW file to 8bpc space. Many examples of DR of the Sony sensors that I see is taking a "normal" exposure (one that's not exposing a full 12-14EV). In those instances, the Sony EXMOR technology has been and still is better, because the lowest "stops" of shadows are cleaner than Canon's technology.

davesrose wrote in post #17742838
But the point is that there's still situations where an exposure will clip in either your highlights, shadows, or both.

Remember: you're not clipping shadows. What's happening is that, at some point, a pixel contains a signal so low (dark) that it's swamped by noise. It's not clipped in the way that a highlight is clipped (photons overflowing the well of a pixel on a sensor).

davesrose wrote in post #17742838
Lets say I do shoot a scene that I notice is clipped in the highlights....I then "underexpose" further down to get more detail in the highlights. If my scene has a higher luminance range then what my sensor is able to record, then if I only expose for highlights, I will get blocked up shadows (and won't be able to "recover" what's not there).

In an ideal world we'd have cameras with so much capture DR that you could always expose for the brightest highlight in the scene (holding some detail), safe in the knowledge that all the darker pixels would have sufficiently low noise that they'd be usable.

Of course, that isn't reality, so we must always make that judgement call of how much and what highlight detail we're willing to sacrifice in order to get sufficient signal in the darker areas of an image (or bracket and blend, of course).

davesrose wrote in post #17742838
From my experience with the D810, there is just a little bit more wiggle room post processing shadows in photos ...it does not always eliminate the need for multiple exposures in a HDR scene, though.

From what I've seen, it's quite a lot more wiggle room (vs Canon), but, agreed, it's not a panacea, and there will always be scenes that benefit from multiple exposures.

sploo wrote in post #17742850
Remember: you're not clipping shadows. What's happening is that, at some point, a pixel contains a signal so low (dark) that it's swamped by noise. It's not clipped in the way that a highlight is clipped (photons overflowing the well of a pixel on a sensor).

But you are...all cameras are currently limited to 14bpc tonal range now. Lets say that you can keep perfect exposure with the D810. The scene is a perfect 14EV, the sensor records 14EV, the RAW is a full 14bpc. The linearly converted jpeg will not have any clipped highlights or shadows....you might want to adjust gamma curves to bring out more detail in an area that looks too washed out or dark. That then gets into manual tone-mapping...adjusting gamma curves for viewing on 8bpc display. But you can have a scene that's over 14EV. In that scenario, if you expose for just highlights, then there will be stops of "shadows" clipped. I know you like to say that my HDR imaging experience has nothing to do with digital photography, but HDR graphics format standards were set to 16bpc and 32bpc for a reason IMO the difference between the first 2 stops of improved tonal range of the Sony sensors is less significant in the grand scheme of things....but yes, there's no argument that Nikon/Sony is inherently cleaner at shadow recovery.

sploo wrote in post #17742803
If you have an image that has clipping in highlight areas where you wanted detail, the shot was overexposed. Simple as that.

Yup. Exposed too far to the right. The old days of picking the midpoint and setting it somewhere are gone; if maximum DR is the goal and you haven't set the exposure to "almost saturate" those pixels where you want to retain detail, you've thrown away DR. This separates the capture from the processing, where "exposure" may (need to) be altered - along with the inevitable tone-mapping that is required to get the capture DR presented in the display/print. Thus the top of the DR is fixed - regardless of what the light meter says for the 18% average gray scene - and there is nothing to "recover" - you either got it or it's saturated/clipped. All we are left to assess is the performance at the bottom of the DR. An analogy might be measuring the length of a stick - I align one end with the 0 on the ruler (=set the highlights to just below clipping), and read off at the other end (in the shadows). If you set the clipping to any other mark, you won't get the same length.

sploo wrote in post #17742803
Remember that highlight recovery is a red herring; a pixel is either over exposed (clipped) or it's not. The issue is that the histograms in our cameras usually give us data based on the JPEG rather than the raw, so detail that might be flagged as clipped in-camera may (or may not) be clipped in the raw file. If our cameras would give us raw histograms then there'd be no ambiguity (and what was clipped would really be clipped - no chance of recovery).
Circling back - the solution to a photo with clipped highlights is obviously to reduce the exposure sufficiently until any (desired) highlights are no longer clipped. That of course reduces the exposure of all areas in the image, meaning that some/much of the image may be in lower stops and thus has a worse signal to noise ratio.

If you then push (brighten) that darker detail in post and it's visually unacceptable due to noise; that's insufficient DR.

In a nutshell: the more DR we have at capture, the better chance you have in holding detail in your highlights (due to taking a darker exposure) whilst still getting acceptable quality in shadow areas.

Perhaps we are defining terms again. I don't use the term "clipping" for shadows. I am familiar with the term "clipping" from my science career and photography both meaning to exceed some threshold of a storage capacity. In photography, the term for shadows that I have heard most often - going back even to film - is "blocked" - they are dark and indistinguishable from what you'd get with a lens cap on. If you try to push them, the detail is indistinguishable from noise. The bottom of the DR is where enough photons were captured that the image detail can just be distinguished from the noise.

When we go from 14 stop (or even 11 stop) capture space to 8 bit presentation space, we don't just chop off 6 bits - we tone-map/gamma correct/manipulate often in a very non-linear way (that's what the "curves" tool is all about). If the exposure was set to just capture the right highlight detail (say a neon sign in a night street scene), it may be that the photographer wants the final image to contain some detail in an area that is 12 stops down from that highlight (the textures in a dark alley). He can do that better with a camera that can capture 14 stops, than with one that captures 11 stops, even though both end up presented in the same 8 bit space. The tone-mapping is just a little different. Yes, it is very easy to screw up the tone mapping - you see it all the time in real estate photos (HDR or otherwise) where the "highlight/shadow" tool was overused (or the HDR-8bit tonemapping was poor), or used with a big radius and everything has haloes. That's not what we are discussing in the difference between 10, 12 and 14 stop DR captures.

davesrose wrote in post #17742873
But you are...all cameras are currently limited to 14bpc tonal range now. Lets say that you can keep perfect exposure with the D810. The scene is a perfect 14EV, the sensor records 14EV, the RAW is a full 14bpc.

Forget digital for the moment and think about what happens in the analog space. If you (are unfortunate enough to) remember old audio cassette tapes you'll recall that they had quite a high noise floor (tape hiss). Any signal (music) that dropped below a certain volume level on the recording was drowned out by that noise.

You could digitise the content of a tape cassette to an digital audio format with quite a large bit depth, but it'd be pretty pointless as the least significant bits would contain nothing but noise.

With a camera sensor, each pixel can take up to some maximum amount of (analog) charge; see "Full Well Capacity" here http://www.clarkvision​.com …rmance.summary/​#full_well

You'll see from that graph that most modern cameras are in the tens of thousands of electrons per pixel range. But, just like cassette tape, there's noise.

In practice then, the least significant bits in the digital data in a raw file do contain information, but it's not usable data as the signal is lost to noise.

Granted when I'm looking at the least significant bit in a raw file (either 0 or 1) then below that I've "clipped", in that there is no darker shadow detail. The issue is that this lower bit is likely to already be useless because it's just noise.

At the sensor level, I guess you could say that the point at which a well either receives a single photon (some light) or no photons (no light) is the lower clipping point (in the sense that when a well overflows it's the upper clipping point). The issue is that noise makes that lower boundary somewhat moot, so it's more a case of "drowned" (out by noise) than clipped.

EDIT: I've just noticed a comment in a caption further down the page I linked above, and this describes it perfectly: "Ultimately, with zero electronics noise, dynamic range would be limited by the number of photons collected"

davesrose wrote in post #17742873
The linearly converted jpeg will not have any clipped highlights or shadows

You mean will have completely clipped highlights and shadows, surely? Unless the camera just squashed the whole 14bit range into 8bits - which would probably look a bit odd if done in a linear fashion (very low contrast output)

AJSJones wrote in post #17742955
Perhaps we are defining terms again. I don't use the term "clipping" for shadows. I am familiar with the term "clipping" from my science career and photography both meaning to exceed some threshold of a storage capacity. In photography, the term for shadows that I have heard most often - going back even to film - is "blocked" - they are dark and indistinguishable from what you'd get with a lens cap on. If you try to push them, the detail is indistinguishable from noise. The bottom of the DR is where enough photons were captured that the image detail can just be distinguished from the noise.

As noted above, I think of it as "drowned" (out by noise), but I guess I'm introducing another term