Dynamic Range & Resolution

[Michael Gerald-Yamasaki]

Greetings. So here are a couple of things that I've puzzled over between what I understand from computer graphics and what I understand from digital photography:

Dynamic Range - Most of our computer displays are 8-bit displays (there are newer 10-bit displays (AKA, confusingly, 30 bit displays) but probably not in common use). They are able to display 256 shades of gray - white to black. I would think this would be something like 8 stops of dynamic range (8 doublings from 1 to 256). Does this not mean that we can only see 8 stops of DR on our screens no matter what the DR of the captured data? On a similar note I've not seen any processing software that attempts to show anything greater than the display 8-bit range (would have to show, for instance, two versions - a lower part of an expanded range and a higher part of an expanded range. Agreed not a very satisfying way to show a DR over 8 stops).

How does higher DR of our newer cameras express itself in the limited DR of the computer screen? Or does it?

Resolution - Similar to the dynamic range question, unless your looking at fat bits or at least 100%, how can the high resolution images from our cameras be displayed to show differences of, say, even a 2-3 MP crop and a 36 MP FF capture? I admit to looking at an image and thinking the IQ seems low, I wonder if this is a big crop (leaving only 3,4,5 MP). But how can that really express itself on a little more than 2MP screen (let alone a 250MB jpeg)?

Is it all just a mass hallucination, and we are really just seeing 8 stops DR and 2 MP resolution, no matter the source?

Thanks for thinking on this...

Cheers,

-Michael-

[John Chardine]

Hi Michael- My way of thinking about the DR question is that the range itself is determined by the contrast that your display is capable of producing. Many today have bright whites and deep blacks and if you measure the intensity of light from these two extremes it comes to more than 8-stops. I would think the 8-bit constraint comes into play in how this DR is divided up from black to white, not in the absolute dynamic range. So, I would expect a modern, high quality, calibrated display to be more or less able to display an image from a modern digital camera faithfully subject to the constraints of the smaller gamut of the display.

[Flavio Rose]

A few points:

The dynamic range of a screen is about seven stops in my experience. That is all we can see even if the jpeg being displayed contains more. Now, what we can do if we have more stops in our capture is use curves to compress x > 7 stops into seven. This "improves" many, many images by making the shadows more visible. However, it is a big source of noise because the stops of the image which are more than seven stops below maximum are pretty noisy when made visible (except, I read, for captures made at low ISOs in modern non-Canon sensors).

Also, jpegs can contain more than eight stops of content because of gamma scaling.

Re resolution, I agree with you: I see on my 1280xsomething screen only a megapixel or so of content. The higher resolution of sensors is mostly wasted for the vast majority of people who view photos only on computer/pad screens -- unless they're crazy bird photographers like me who do huge crops. That is why my bird lenses are so expensive and yet I shoot most other stuff with kit lenses.

However, just to be clear, I think for a one-megapixel jpeg you need four megapixels of capture as those are usually described. That is because the one-megapixel jpeg has R, G, and B in all pixels, whereas the way manufacturers describe their sensors, each pixel has just one of R, G, and B. If you go 1:1 from capture to jpeg, the other two values for a pixel are interpolated from neighboring pixels (so-called "Bayer interpolation").

[John Chardine]

This seems to be quite useful reference on the subject display DR (although the paper might be long in the tooth; I could not see a date other than some citations in the 2000s):

http://matttrent.s3.amazonaws.com/heroku/attachments/research/papers/Siggraph.2004-HDR_Display.pdf

[Michael Gerald-Yamasaki]

John,

Thanks for posting the paper. The discussion on human perception is pretty interesting, particularly the chart on Just Noticeable DIfference Levels for Different Maximum Luminance Levels. I'd interpret this chart as saying for very high end (in terms of brightness, luminance) displays 10 stops is about as much as you could expect in terms of just noticeable differences between steps - that would be the number distinguishing 12-bit or 14-bit captures meaning that even 12-bits is overkill.

As far a luminance levels, looking around it seems something around 2000 nits (probably a slight misrepresentative name for cd/m2 but I like it) would look like 11 stops but only if you had a really black black. Computer screens (from what I can find) look to have about a 10-20 nit luminance level at their blackest knocking out, say, 4 of those stops. So high end displays are looking at maybe 7 stops (in luminance difference from black to white), while typical LCD perhaps a little as 5.

Of course, maybe I misunderstand the numbers.

Numbers aside one thing looking at this stuff that i found interesting is the blacks are where it's at for DR. You quickly lose DR if your blacks aren't that black, which is pretty descriptive of what I've seen in printing (DMAX and all that).

So, getting back to the OP... the number of steps on a conventional display is limited to 256 while the brightness levels support substantially less than 8 doublings in luminance. Short of HDR techniques (localized mapping from intensity data, say, raw data to display intensities with different mappings across an image), it seems to me that our displays (and probably our prints, as well) can't possibly represent the DR our sensors are capable of capturing.

No?

Cheers,

-Michael-

[John Chardine]

It makes sense Michael. The only thing I will say is that despite these limitations, a good photograph put up on my Apple 27" Thunderbolt (IPS) display looks stunning and I am sure there are many other higher-end displays that perform as good or better.

[Michael Gerald-Yamasaki]

It makes sense Michael. The only thing I will say is that despite these limitations, a good photograph put up on my Apple 27" Thunderbolt (IPS) display looks stunning and I am sure there are many other higher-end displays that perform as good or better.

Sure thing. There are plenty of stunners posted here on BPN with the small jpeg limitation... I just wonder about the exclamations about the amazing DR on the new cameras as to how one would know given the limitations of our display tech.

(not to mention that given the choice between 12 bit and 14 bit captures I still choose 14 bits even though I've yet to see a real difference even in extreme post-processing :w3 ).

Cheers,

-Michael-

[Flavio Rose]

I just wonder about the exclamations about the amazing DR on the new cameras as to how one would know given the limitations of our display tech.

You will only know by raising shadows so that the lower stops become visible and, lo and behold, they're less noisy than you're used to. You might be motivated to do this shadow raising because you underexposed (i.e., a boo-boo) or because you really want to compress say 9 stops into 7 for some esthetic reason, be it with curves or an HDR program.