Luminous Landscape Forum
Equipment & Techniques => Cameras, Lenses and Shooting gear => Topic started by: sgwrx on September 16, 2017, 12:41:06 am
-
hello. so since i've read about the dynamic range of some of the newer PhaseOne digital systems, i've wondered what exactly this looks like? and what does it mean? if i take a photograph of scene with bright sky in the background and use my canon 6d, then take the same image with same exposure settings with one of the HDR PhaseOne'rs, the plop both of these image into a raw converter, will the PhaseOne with 10 or 12 stops of DR have that bright sky under control without me having to do any adjustments?
i know for instance, that you can use a highlight recovery slider in a raw converter and get some of that sky back. or, a shadows slider to increase the visibility of shadow details.
i work in a 3d rendering program that can generate 32bit HDR images. i can save those out and open in photoshop and use an exposure slider or whatever, to basically set whatever exposure "range" i want for the rendered image. technically i don't know how to equate the number of stops the image is. but ultimately i get a nice clear image no matter what. in addition to the 3d rendering program's simulated sun and sky light, i can use HDR photographs to light a scene... those are typically 10 or so photographs blended together from under exposure to over exposure. but even with these, as i increase/decrease the exposure of the rendered image, i still run into the sky can be over exposed and all "white" or shadows can be all "black".
or, is that not even the issue in digital camera systems? is the issue the noise, "blockiness", lack of details in shadows as you try to "open" them up in a scene that might be exposed for that bright sky.
thanks,
-
What's the difference in luminosity between the darkest shadow tone short of black and the brightest highlight short of white that a given sensor + supporting electronics can capture and quantize? That's the camera's dynamic range. As to whether a given camera is good at squeezing its data into a file needing minimal or no tweaking…that's a very different matter. To start: how do you define "good at?" :)
-Dave-
-
to answer my own question - since 0-255 is the best we can do in computers and displays and such, i think the biggest benefit of HDR images is first noise suppression in lower tones and then the ability to make an HDR type image that squeezes in the 0-255 that we have on our computers.
-
to answer my own question - since 0-255 is the best we can do in computers and displays and such, i think the biggest benefit of HDR images is first noise suppression in lower tones and then the ability to make an HDR type image that squeezes in the 0-255 that we have on our computers.
That is not true. 10 bit monitors are widely available and prices are coming down. Here (http://wolfcrow.com/blog/fun-comparison-which-is-the-best-4k-monitor-for-color-grading/) is a link to a review of 10 bit 4K wide gamut monitors. The BBC has a good article (http://www.bbc.co.uk/rd/sites/50335ff370b5c262af000004/assets/586cdf8106d63ee287051b35/BBC_HDRTV_FAQ.pdf) on HDR TV. Adobe Photoshop supports 10 bit color.
The weak link in the reproduction chain is the print. Tone mapping is necessary to compress a HDR gamut to what can be printed, as Karl Lang presents here. (https://www.ppa.com/files/pdfs/renderingtheprint.pdf)
Regards,
Bill
-
The low noise in shadows is the reason it has higher dynamic range Such a sensor means more extreme contrast conditions can be captured and still record usable information that can later be manipulated back into useful output.
An extreme example, single shot CMOS IQ3 100. Exposed so the sun itself wouldn't clip to maintain the color information in it. the IQ180 wasn't too bad as long as you were at base ISO, but I don't know if it could have pulled this off. Which other sensors could have also pulled it off? not sure, but probably several of the more modern Sony sensors.
unmodified RAW then processed final image.
-
There are two aspects of dynamic range at play here - input and ouput. They have nothing to do with each other.
Any given scene has a certain amount of dynamic range - the difference between the darkest point and the brightest point in the scene. A sensor with a given amount of dynamic range may be able to capture detail within the entire DR of the scene (if the sensor's DR exceeds the DR of the scene) or only part of it (if the DR of the scene exceeds the DR of the sensor). For instance, a sensor capable of recording 11.5 stops of DR can capture detail within any 11.5 stops of the scene's DR - it can be the shadows (leaving the highlights to blow out), the highlights (losing the shadows to noise) or somewhere in between (losing a bit to each end). This input DR determines a relative brightness range over which detail can be simultaneously recorded by the sensor - a sensor capable of 11.5 stops of DR can record detail in areas 11.5 stops apart brightness-wise.
Output DR is the difference between the brightest and darkest point the monitor, printer/paper combination or other display system is capable of displaying. This is usually significantly less than the input DR of the sensor. But this doesn't mean a 5-stop print can't display a 14-stop image. Remember, input DR denotes the range in which the sensor can record detail. This detail can then be compressed into whatever DR the output system is capable of displaying, with each pixel retaining its brightness relative to the rest of the image, and, thus, detail being retained. Remember, when you produce a print or display an image of a landscape, you don't expect the paper or monitor to have as much brightness or contrast as the scene where you took the photo. Rather, it is the relationships between areas of bright and dark - not the absolute values themselves, but whether you can distinguish a dark line (e.g. a crack in a granite wall) from the surrounding, brighter area - which determine whether detail is retained and, thus, the dynamic range capabilities of the system.
This is where tone-mapping comes in. The input data is compressed to fit the output requirements, while maintaining each brightness level's relationship with all the others within the image. It also has nothing to do with bit depth, except for the number of possible levels being restricted to the number of levels permissible by the bit depth. This doesn't restrict the range of usable input DR, since conversion is almost always non-linear. In a way, tonemapping is just like taking an image printed at 90x60" and 100ppi and re-printing it at 30x20" and 300ppi - the amount of detail is the same, just that the points are closer together.
-
Shadowbkade that was a very well written and thoughtful explanation.
-
Hear, hear, shadowblade! Great explanation.
-
to answer my own question - since 0-255 is the best we can do in computers and displays and such, i think the biggest benefit of HDR images is first noise suppression in lower tones and then the ability to make an HDR type image that squeezes in the 0-255 that we have on our computers.
Those are just numbers; not necessarily tone or color(s) we might capture. And only 8-bits per color using that granularity. You can divide up the numbers more if you have the data to do so.
See:
http://digitaldog.net/files/ColorNumbersColorGamut.pdf
-
There are two aspects of dynamic range at play here - input and ouput.
Hear, hear. And the input side can be split into two relevant parts, Scene dynamic range, and Capture dynamic range.
The latter, Capture dynamic range, can be subdivided into sensor dynamic range, and lens+sensor dynamic range. The veiling glare that the lens creates can significantly reduce the dynamic range that's presented to the sensor, turning 12-15 stops maybe to something like 9 stops maximum. That's usually quite different in CGI created imagery, which can have arbitrarily high Scene DR with zero Capture losses.
The larger dynamic range of the sensor still helps, even with a reduced lens-projected DR, because the shadows are cleaner and we can do more extreme tonemapping on it in output, including boosted shadow contrast. HDR images will also need tonemapping if we want to avoid dull low contrast output.
Cheers,
Bart
-
wow thanks for responses! relative brightness range... and tone mapping. things i thought i knew, but now know more completely. so forgetting the details i've learned here to some extent, when i use ACR or C1 to develop two images, one which reveals the shadows and one which reveals the highlights, then combine those two in photoshop, i'm bascially doing my own variation of tone mapping. example, a shot inside looking out, where the exterior is very bright white...
ok, so tone mapping is also what's going on when people actually shoot 5 or 6 or 10 different exposures of a scene, from very under exposed to very over exposed, and stack those images together into a digital HDR image. tone mapping.
finally, in my 3d program, when i lower the brightness of the sun (i feel powerful in the 3d world ha!) so that my interior lights show enough brightness to see the interior, and the exterior still has somewhat blue sky and i can see outside (rather than being blown out white) i'm tone mapping.
awesome!
ok, to the point of the 0-255 as it relates to a monitor, in some world, we could develop a monitor that displays "black" or 0, and then kind of go beyond 255 white, by have super-bright pixels. i think what i'm trying to say here is, real dynamic range in a monitor could potentially be developed by have super bright LED's in it (kind of like OLED tvs?) but also a graphics card and computer and image to be able to display that. hm. or in terms of light bulbs, i could have a 1 watt light bulbs to 1,000 watt light bulbs and based on brightness you'd have a wide or high dynamic range display depending on what it's displaying.
-
The low noise in shadows is the reason it has higher dynamic range Such a sensor means more extreme contrast conditions can be captured and still record usable information that can later be manipulated back into useful output.
An extreme example, single shot CMOS IQ3 100. Exposed so the sun itself wouldn't clip to maintain the color information in it. the IQ180 wasn't too bad as long as you were at base ISO, but I don't know if it could have pulled this off. Which other sensors could have also pulled it off? not sure, but probably several of the more modern Sony sensors.
unmodified RAW then processed final image.
thanks! this is an awesome example!
-
oh, one more question - this is why a lot or most HDR photos look horrible huh? because we expect that "outside" would be blown out white if you are taking an interior photo during the day. but we see an exterior which is similar in brightness to the interior and that looks wanky.
typically, i will leave my exterior in that case, overly bright, but with enough detail and color to see what's out there. i do this also in my 3d renders.
-
oh, one more question - this is why a lot or most HDR photos look horrible huh?
Photographers with no artistic sense.
-
oh, one more question - this is why a lot or most HDR photos look horrible huh? because we expect that "outside" would be blown out white if you are taking an interior photo during the day. but we see an exterior which is similar in brightness to the interior and that looks wanky.
typically, i will leave my exterior in that case, overly bright, but with enough detail and color to see what's out there. i do this also in my 3d renders.
That's due to bad tonemapping, not HDR.
Every digital photo - HDR or not - is tonemapped. The thing is, with a standard exposure, the RAW converter applies a default tonemapping curve for you, subject to a limited number of manual controls. With an HDR image, you need to set the curve yourself. This gives you more opportunity to screw it up. You can screw up a non-HDR file just as easily if you import it into tonemapping software and apply a bad curve.
-
That's due to bad tonemapping, not HDR.
Every digital photo - HDR or not - is tonemapped. The thing is, with a standard exposure, the RAW converter applies a default tonemapping curve for you, subject to a limited number of manual controls. With an HDR image, you need to set the curve yourself. This gives you more opportunity to screw it up. You can screw up a non-HDR file just as easily if you import it into tonemapping software and apply a bad curve.
Ok sure. Even in the software they talk about applying a certain tone curve to the raw image. there's probably one applied in the camera at capture as well.
-
While shooting this weekend, this thread came to mind so I thought I'd take a shot with an iPhone that has relatively limited dynamic range and a Sony a7R Mk II with excellent dynamic range to illustrate what dynamic range really means. To top it off, I even used the HDR feature on the iPhone and the a7R Mk II still annihilated the iPhone. Just look at the detail in the buildings and in the iPhone picture and you can't even really see the river boats since they are just silhouetted against the far bank:
iPhone first, then a7R Mk II:
-
Nice demonstration, EJ, if just a little bit lop-sided. ;D I think that the high-end camera companies need not give up hope of survival in the face of the phone onslaught.
You are local to STL?
-
nice! so this is where i started to conclude that the main benefit must be in single capture noise. but now i'm even questioning that. see the thing is, you can take an hdr image with any camera, it really boils down to how many exposures (1 or multiple) and what type of post processing or in camera processing there is. also, what is closest to what our eye's see i suppose. and again, at least on a computer monitor, it all boils down to what can you compress into 0 and 255 (in each of the colors) and still display. but who knows.
everything is just better about the sony image to me.
-
nice! so this is where i started to conclude that the main benefit must be in single capture noise. but now i'm even questioning that. see the thing is, you can take an hdr image with any camera, it really boils down to how many exposures (1 or multiple) and what type of post processing or in camera processing there is. also, what is closest to what our eye's see i suppose. and again, at least on a computer monitor, it all boils down to what can you compress into 0 and 255 (in each of the colors) and still display. but who knows.
everything is just better about the sony image to me.
Nothing whatsoever to do with the 0-255 RGB display. That's just the number of levels - it tells you nothing about the contrast between the brightest and darkest possible parts (Dmin and Dmax) and nothing about the difference between each level from 0-255 (it isn't necessarily linear). And all that's just output dynamic range, which has nothing to do with input dynamic range. You can map any input DR to any output, and still take advantage of all the DR of the input even if the output DR is restricted. Prints on matte paper don't have more than about 5 stops, but you can map all 14 stops of input DR into that - you're not just restricted to showing the highlights, or the shadows, and having the other end blow out.
-
While shooting this weekend, this thread came to mind so I thought I'd take a shot with an iPhone that has relatively limited dynamic range and a Sony a7R Mk II with excellent dynamic range to illustrate what dynamic range really means. To top it off, I even used the HDR feature on the iPhone and the a7R Mk II still annihilated the iPhone. Just look at the detail in the buildings and in the iPhone picture and you can't even really see the river boats since they are just silhouetted against the far bank:
iPhone first, then a7R Mk II:
was the Sony shot in raw and then processed or is this an in camera jpeg? Just curious, since the iPhone software might have made the decision to leave the shadows dark, not necessarily representative of what the HDR process could do if it was left to the photographer.
Not trying to defend the iPhone here, I shoot sony too and the iPhone isn’t any where close in DR. But it does make me curious how the iPhone would fair if the captures were raw and then went through similar post processing.
-
While shooting this weekend, this thread came to mind so I thought I'd take a shot with an iPhone that has relatively limited dynamic range and a Sony a7R Mk II with excellent dynamic range to illustrate what dynamic range really means. To top it off, I even used the HDR feature on the iPhone and the a7R Mk II still annihilated the iPhone.
Mmmm: this does legitimately illustrate the advantages of high dynamic range capture (whether single shot or via HDR composite) when photographing scenes of high Subject Brightness Range (why have photographic discussions mostly abandoned the old standard jargon of SBR?)
But this reminds me a bit of the notorious article at LuLa some years ago arguing the superiority of medium format cameras by comparing to what turned out to be an iPhone image. Here one missing detail is that this is the four year old iPhone 5S; two years older than the Sony a7R Mk II.
So for more irrelevant fun, let's compare high SBR handling between this year's hottest new toys: the iPhone X and the Nikon D850!
P. S. I should add: thanks to shadowblade for the good explanation, and to Wayne Fox for the illustration of getting an artistically worthwhile result from a scene with way too much SBR for "straight" printing or display.
-
OK, lots of questions/comments:
- No I am just visiting the STL area
- Sony shot in RAW, even had I shot it in JPEG it would still be very different, in fact the embedded JPEG in the RAW file clearly shows this.
- Yes, my newer iPhone did not survive the Pantanal and I have an 8 Plus waiting at home so I am temporarily using an old phone but even with a 7 Plus, there is absolutely no comparison in dynamic range and remember I shot the iPhone shot with HDR on!
- The benefit is definitely not just noise - it is how many tonal levels the sensor records between solid black and solid white
- Not at all trying to advocate one format over the other, simply pointing out the difference between a sensor that has lower dynamic range with one that has higher dynamic range using what is probably an extreme difference in equipment for illustrative purposes. I just happened to think of this thread as I was shooting at dusk...
-
to put some numbers to it:
according to http://photonstophotos.net/Charts/PDR.htm
the Apple Iphone7 has at 25 asa the same dynamic range as the d850 at 2500asa
at 25 asa the iphone has 7.15 stops of dynamic range the d850 has at 64 asa 11.63 stops = about 4.5 stops more.
at asa 200 the iPhone has the same dynamic range as the d850 at 12.800asa= 4.7 stops.
But i think what is interesting is the different approach camera and phone companies have.
The iphone companies are computer based and use advanced software to get the best out of the captured information.
They do a very good job in processing the image to a kind of general taste. Part of the work i have to do myself in photoshop.
-
A DR question: (These three images are typical of what I photograph).
For digital images, I use a 9-stop DR sensor camera. If I use a 12-stop DR sensor camera, what differences might I see (other than increase in resolution size)?
Will there be smoother transitions between tonality? A better gradation between colors?
Optimistically...
-
A DR question: (These three images are typical of what I photograph).
For digital images, I use a 9-stop DR sensor camera. If I use a 12-stop DR sensor camera, what differences might I see (other than increase in resolution size)?
Will there be smoother transitions between tonality? A better gradation between colors?
Optimistically...
In the last photo less noise and more detail in the dark areas...- yes smoother transitions.
This also depends on your way of handling the files
But maybe with an other camera you do not like the colour you now have- always a bit different.
And if you have a controlled environment you just make sure you can arrange you only need 9 stops.
-
In the last photo less noise and more detail in the dark areas...- yes smoother transitions.
This also depends on your way of handling the files
But maybe with an other camera you do not like the colour you now have- always a bit different.
And if you have a controlled environment you just make sure you can arrange you only need 9 stops.
Not necessarily.
Yes, you will have less noise in the shadows, since they may be six stops above the noise floor rather than two - the SNR in the shadows may be much higher (and you will be able to capture detail in deeper ones). But transitions and graduations will not necessarily be more gradual - that depends on the bit depth rather than the dynamic range. More bit depth means finer graduations. If you have a sensor with high bit depth but low dynamic range, you may have 32 different levels within the darkest stop, and it only goes up from there. Conversely, if you have high dynamic range but low bit depth, you may have one level at the darkest stop, with no levels at the secomd darkest and only two levels in the third darkest.
And this is all complicated by noise. It's only meaningful to have more levels of tonality in the shadows if there is enough of a difference between the levels to make detail distinguishable from noise. Although having the difference between levels slightly smaller than the noise is, in itself, no bad thing if it's not overdone - it ensures that all recordable detail is captured and helps eliminate posterisation, a bit like adding noise to eliminate posterisation in post-processing, where curves are being stretched and levels pulled further apart.
-
Hi,
Dynamic range is about recovery of shadow detail. All digital sensor handle highlights the same. If you have a lot of DR you may underexpose to protect highlights and still have clean shadow detail.
I would suggest that DR is somewhat overrated.
Technically, DR is usually the full well capacity (FWC) of the sensor divided by readout noise. So a sensor with a good FWC and low readout noise has a good DR. This is an area where modern CMOS has made great progress. Signal readout is much cleaner compared to CCD or old CMOS technology, at least at base ISO. All CMOS based MFDs except Leica use Sony technology.
Best regards
Erik
hello. so since i've read about the dynamic range of some of the newer PhaseOne digital systems, i've wondered what exactly this looks like? and what does it mean? if i take a photograph of scene with bright sky in the background and use my canon 6d, then take the same image with same exposure settings with one of the HDR PhaseOne'rs, the plop both of these image into a raw converter, will the PhaseOne with 10 or 12 stops of DR have that bright sky under control without me having to do any adjustments?
i know for instance, that you can use a highlight recovery slider in a raw converter and get some of that sky back. or, a shadows slider to increase the visibility of shadow details.
i work in a 3d rendering program that can generate 32bit HDR images. i can save those out and open in photoshop and use an exposure slider or whatever, to basically set whatever exposure "range" i want for the rendered image. technically i don't know how to equate the number of stops the image is. but ultimately i get a nice clear image no matter what. in addition to the 3d rendering program's simulated sun and sky light, i can use HDR photographs to light a scene... those are typically 10 or so photographs blended together from under exposure to over exposure. but even with these, as i increase/decrease the exposure of the rendered image, i still run into the sky can be over exposed and all "white" or shadows can be all "black".
or, is that not even the issue in digital camera systems? is the issue the noise, "blockiness", lack of details in shadows as you try to "open" them up in a scene that might be exposed for that bright sky.
thanks,
-
Dynamic range is about recovery of shadow detail.
Dynamic range is about the capture of shadow to highlight data/detail.
You can't recover what wasn't captured.
This is a bit like those that state they 'alter' exposure in post'. Not possible. Exposure like DR is an attribute of the capture of the data.
DR isn't at all over-rated IF you desire to capture as much of what you 'saw' and what existed at the scene. More data on this concept:
http://www.lumita.com/site_media/work/whitepapers/files/pscs3_rendering_image.pdf (http://www.lumita.com/site_media/work/whitepapers/files/pscs3_rendering_image.pdf)
Page 3-4 but do read the entire piece.
-
In the last photo less noise and more detail in the dark areas...- yes smoother transitions.
This also depends on your way of handling the files
Like how you exposed them:
https://photographylife.com/how-to-use-the-full-dynamic-range-of-your-camera
-
Hi,
The factors involved are the full well capacity (FWC) , that is how many electron charges the pixel can hold and the readout noise of the sensor. DR is simply FWC divided with the "readout noise" of the sensor. This is a tiny bit oversimplified.
It could be argued that FWC should be utilized fully at exposure, exposing non specular highlights just below FWC, this is called expose to the right and maximises DR.
Highlights can never go beyond FWC, but you can protect highlights by underexposing the image.
The other end is dark detail. Here noise sets the limit. Engineering DR tells how far the noise floor is from the ceiling defined by FWC.
Engineering DR is plenty on modern sensors, especially with CMOS sensors using column converters doing conversions in parallell. But, you cannot show like 13 EV of DR on a screen especially not in surround light. So, if we want to utilize the DR of a modern camera, we need to map tones.
The images below show a white text over a gradient going from Lab(100, 0,0) to Lab(80, 0, 0), printed on paper. The images were shot on a P45+ (2007 year's CCD) and a Sony A7rII (2015 year's CMOS)
(https://4.img-dpreview.com/files/p/E~forums/58984278/2ae5eacd49d34b248c71368a4a313844). The text begins with Lorem ipsum and the "m" in Lorem is the first discernible character in the original print.
Getting a scene with a full dynamic range utilizing the sensor is no easy task. In the example below, a tabletop setup was made, illuminated by a single flash with a grid to reduce light spread. The image was split into a highlight/shadow part. An additional flash at minimum power was bounced trough a door to get some fill light. This scene had a luminance range of around 14 stop.
The highlight parts were pretty much similar:
(https://2.img-dpreview.com/files/p/E~forums/58984278/7ca61b80d0494593a586be910e162a8c)
Brightening the shadows was not easy, as I run into the 5EV exposure limitation of Lightroom, combined with the built in exposure bias of LR. So, I resorted to RawTherapee to show shadow detail, where I could adjust exposure at will:
(https://1.img-dpreview.com/files/p/E~forums/58984278/2a9d310695ac4026a2b88ee8d40ec2d9)
This image shows that the A7rII handles the shadow detail well. The P45+ image is noisy. With the A900, that is the same era as the P45+, the image is blotchy. This is probably in part due to the A900 using a 12 bit file format, 14 bits are needed to handle 14EV of dynamic range.
Some examples are shown in this discussion:
https://www.dpreview.com/forums/thread/4083662
Best regards
Erik
-
Engineering DR is plenty on modern sensors, especially with CMOS sensors using column converters doing conversions in parallell.
Plenty, perhaps. But not enough for many scenes which the Karl Lang article clearly outlined visually and numerically.
The article by the author of RawDigger doesn't appear to believe that getting a scene with a full dynamic range utilizing the sensor is difficult task when the scene DR doesn't exceed the sensor DR.
-
And what about lens contrast limitations?
Shooting and combining +/- 5EV images can give an enormous 'dynamic range' - but it doesn't at all.
Yes, shadow become less noisy but there not much new details - 'cause camera body and in-lens reflections eats almost all.
Zeiss is better, old zoom lens is worthier but over 14EV is practical ceiling.
-
Interesting discussion. Can someone here point me to a good tonemapping tutorial? And software? (And if you have a good technique, feel free to explain it.) I'm a commercial photographer and tend to light everything I shoot, so I'm not usually dealing with images needing HDR techniques or bracketing. But occasionally I do shoot landscapes / found scenes, and I'd like to be more capable when it comes to rendering and processing them artfully.
I shoot predominantly with a Hasselblad H5D-50 (CCD).
Thank you,
ethan pines
-
Interesting discussion. Can someone here point me to a good tonemapping tutorial? And software? (And if you have a good technique, feel free to explain it.) I'm a commercial photographer and tend to light everything I shoot, so I'm not usually dealing with images needing HDR techniques or bracketing. But occasionally I do shoot landscapes / found scenes, and I'd like to be more capable when it comes to rendering and processing them artfully.
I shoot predominantly with a Hasselblad H5D-50 (CCD).
Thank you,
ethan pines
You're tonemapping any time you apply a Curves adjustment to an image, a channel or part of an image. In other words, every time you use almost any tool in Photoshop, Capture One or any other RAW converter or editor. It's the same whether it's a regular file or a merged HDR file - just that the HDR file gives you more data to work with.
-
I'm going to rephrase my question in a new thread. Thanks.
-
I'm going to rephrase my question in a new thread. Thanks.
Excellent idea considering one can pull a curve without affecting tones. ;)