Limited dynamic range starts to be addressed

[atvinnys]

I just saw an article about DR limitations and apparently some are trying to solve the issue (to some extent of course..)

http://www.dpreview.com/news/0902/09...hinterview.asp

This sounds interesting..
......
For more extreme lighting situations, there is also a built-in high-dynamic range mode. This shoots two, differently exposed, images consecutively and combines them to enable to capture of a greater dynamic range than would be possible in a single exposure. The company claims the CX1 will be able to capture and convey dynamic range of up to 12EV. However, Hongoh stresses that its feature isn't trying to produce the fashionable, heavily-processed 'HDR-look.' 'it aims to portray the scene in as natural a way as possible,' he says.

[Colin Southern]

Interesting.

The problem as I see it isn't so much the capturing of an HDR image as it is "what do you do with it once you've got it" in that the full range can't be printed and (to the best of my knowledge) there's only one monitor that can even get close to displaying them, without all kinds of compression/tone mapping - thus I don't see HDR images becoming the "norm" for photography anytime soon (in fact I'd argue that getting a conventional HDR image to look as acceptable as a conventional range image is probably the bigger challenge).

[Davey]

Hmm I still struggle trying to get HDR to look right, and it's the only reason I'm interested in it. I suppose when monitor tech changes then HDR will be very worthwhile but I think current range is good enough (if tone mapped correctly of course). Safe bet the range of paper isn't going to change any time soon though.

I'd like to see the results of difficult shots that it produces to see how well it manages it. I suspect it can't be as good as pc result, especially considering manual selective blending making comps more flexible. Still in camera it's pretty useful in the right instance. If anyone gets one or finds examples in the future I'd be interested in how it fares.

[Colin Southern]

Part of the problem is that with current display technologies, the displaying of HDR images has to be a compromise; you're never going to be able to look at a photo of the sun on a monitor and have it burn your eyeballs; so there's a limit to how these things can be displayed and in my opinion these limitations can often mean that there is no way to have the scene look totally natural - a lot of the time the "HDR look" is simply an unavoidable consequence of compressing, say, 10 to 14 stops into 4 or 5.

All is not lost though - if you have a spare $49K spare ...

http://www.bit-tech.net/hardware/200...side_hdr_edr/1

[Davey]

If I did have it spare I wouldn't be spending it on a monitor that's for sure. The link info about game HDR is not actually correct since they don't use HDR and display it. The bloom lighting is related to HDR but wouldn't really benefit from that screen, and as for true HDR games that use float 32 they always tonemap so you wouldn't see anything extra and devs are not going to support such a rare feature. Tone mapping in games is good enough these days anyway (eg. stalker with float32 and fdl maxed for instance). There is some hdr stuff used in games but tends to be for light maps etc so again no benefit.

That leaves photos since again film makers wont bother due to it being such a small thing. I think paying that much for a monitor to enjoy images you've taken yourself is not the smartest thing in the world and imagine tone mapping will do a good enough job if done right (and some other tweaks in pp). Scary thing is someone actual has bought one. I reckon a lot of the effect will be falsely percieved one because of people expecting it to be better as opposed to any really huge increased range in a lot of the use of such a monitor. After all there is little (other than hdr photos) where you would be seeing true increased range since most sources don't provide increased range and you can only get out what you put in. Still I wouldn't say no if someone offered me one for free.....you never know.

[Colin Southern]

I think paying that much for a monitor to enjoy images you've taken yourself is not the smartest thing in the world and imagine tone mapping will do a good enough job if done right (and some other tweaks in pp).

Personally, I've got better things to invest my money in too - but - I suspect that the images would none the less look far more dynamic than what we're seeing with our present technology. I suspect that this is the effect I was seeing when I experimented with backlighting some of my canvases with a strong light - the vividness of the colours and tonal ranges looked fantastic.

[Davey]

weird you mention that as was wondering today outside one of the new galleries if they would ever make higher impact with the increased range canvas works by back lighting or something. I thought about photo frames which seem to be becoming popular, perhaps this is why (I always prefered paper/canvas prints myself for photos).

I thought the problem with back lit stuff would be the too harsh a lighting effect, coupled with the fact it would possibly bring the darks up too much. In the end I thought maybe subtle backlighting like using EL film seemed good (I used to have a krill stick until my idiot of an ex wife put the batteries in the wrong way on a camping trip and broke it which is where I got the el film idea since it glows but subtle).

How did you find the image looked when you did that, was it the effect you imagined?

[Colin Southern]

How did you find the image looked when you did that, was it the effect you imagined?

It was actually quite interesting. The canvas has a texture that wasn't overly flattering - but the basic concept appears to be quite sound.

What I need is a way to light the image using white LEDs shining onto some kind of diffuser - haven't figured that one out yet though. I think that back lit canvases hanging on the wall of your average home with subdued room lighting would look quite spectacular.

If one were REALLY keen then one could tweak the position and output of the LEDs to give a super-dynamic looking image.

[Dave Humphries]

Befopre long you'll have (re)invented the electronic picture frame!
I mean, why stop at a few LEDs, why not have 1600 x 1200 (or more) and colour them too?

Not sure it'd work for artist's painted canvas though, if they mess up, they just paint over it and no-one is any the wiser until some shines light through it.

Interesting thoughts though, cheers,

[Davey]

There is always OLED, you could print them too (as they do with some) kind of similar to normal printing technique (but obviously specialised) so you could get super super small pixels cheaply (where as that many micro leds would be expensive. The downside is the life of oled is still poor but most mobile phone screens etc last for a long-ish time. They don't need backlights and many rate them as less harsh compared to backlit screens (lcd etc).

Limited range on painted canvas would be awkward I agree as you say the over painting is an issue. As is pigment opacity, a dye based image would be fine though. The other problem is many painting techniques even without mistakes such as dry shading wouldn't work well being back lit. I'd love to see modern technological advances brought into traditional painting techniques. Even not so recent developments could be incorporated well, eg. I'm yet to see proper blacklight art, ie art piece painted in uv reactive paints and viewed under uv lighting.

May be pictures frames are oled, I don't know they might be lcd. Photography wise though it would be easy enough to print semi opaque dye ink on appropriate medium and backlight it with white EL film or white led or even print the pic in oled inks and pass small current though it so the pigments themselves glow which would give perfect range. I'm sure this is technically viable. Maybe you should reinvent electronic picture frames

[_GUI_]

there's a limit to how these things can be displayed and in my opinion these limitations can often mean that there is no way to have the scene look totally natural - a lot of the time the "HDR look" is simply an unavoidable consequence of compressing, say, 10 to 14 stops into 4 or 5.

I agree. And it is not only because our screen (or output devices in general: projection machines, printed stuff,...) have a limited real DR, but also because in real life our eyes quickly adapt to the different parts of the scene by adjusting the amount of light coming in. That provides us with the feeling of having such a huge DR in our visual system that will never be matched by a flat image we are seeing as a whole on a printed paper or a monitor, without the need to look into different directions because it falls into a relatively narrow angle of view.

IMO a scene with a DR larger that about 12 f-stops (or even less if the final available DR is very low) is more reallistically represented if we assume to loose some range in one or both ends (i.e. we assume some parts will be blown and some others clipped to black), and the remaining 12 stops are then tone mapped in the output device being used.
Compressing more DR into the output can be done, but the cost will always be a non-reallistic feeling (lack of global contrast, halos,...).

BR

[Davey]

Interesting you mention that gui since the most contrasting high range looking paintings tend to be similar in actual range to others seeming to lack contrast.

Usually (in particular with the great classics) the use of contrasting colour especially in large format works gives the impression of great range due to the eyes adjusting to spot local contrast variations (mainly anyway as obviously other factors). Vibrant colours against contrasting shades, light against deeps and so on, done well gives the impression across the whole work of very wide ranges even if technically it's not. You can really make stuff stand out by taking this into account, where as the same range composed differently can look much more compressed or lacking in contrast. If you make the areas the eyes are drawn to have high local contrast then people get the impression of higher range than it is. Again the more different areas you do this in the wider people seem to think it is. They don't actually seem to rate the global contrast but seem to use the sum of these local contrast differences to work out what it is globally. Seeing the whole image at once they'd probably feel differently. Obviously it's hard on photos, especially viewed on small display or print sizes since people assess the whole scene at once.

I suppose you can give the impression of wider range again by breaking the image up more by deepening shadows or scene dependent stuff like colour utilised correctly will fix this. Obviously it's easier to do on big images since the people don't see the whole thing at once. In photography terms I am not really good at all at capturing stuff in that way, and not just due to the limitations of subject itself but at spotting the right lighting to make it stand out more, or due to lack of skills and familiarity in recognising good composition or best point to focus etc.

[_GUI_]

Usually (in particular with the great classics) the use of contrasting colour especially in large format works gives the impression of great range due to the eyes adjusting to spot local contrast variations.

I find this comment highly interesting Davey. This opens a gate to start to think of new ways of trying to tone map images with a final natural perception of the observer in mind.

The variable 'size of the print' has now been introduced in the equation. A huge print of a high DR scene tone mapped that could look unreallistic on a screen monitor or small copy, could however look more natural if seen in a large format where you are obliged to move your head. With the same low contrast format, the enlarged size would help to make your brain think you are enjoying much more DR without that previous unreallistic feeling.

Could this make true a statement such as: "HDR images are best suited in very large formats"?

Unfortunately 99.9% of images are to be seen on small formats not big enough to reach this effect, but the concept is interesting.

BR

[Davey]

Interesting thought, I suppose I'd really need to think about that one. My initial thought is any image needs a certain size/viewing distance to convey a similar perception of range as the scene from which it was captured. I think the problem with hdr I have found is by nature of how it works it naturally closes the gap on the range of luminanc. Obviously this is a good thing from the point of view of noise and blown highlights, but it means the HDR image that is simply just tonemapped (obviously effect varies greatly with method of mapping) and no further processing done will be lacking in luminance contrast (which could also make colours suffer since sometimes it's more their luminance than their saturation that adds to contrast in a scene, ie. deep crimson dress against pale blue background looks a hell of a lot more contrasting than an equally saturated red brought up towards the light end). So many images I have seen are lacking deep shadows and bright highlights.

I think for that reason there will always be time needed with the burn tool as well as curves 'n' contrast adjustments to get it natural again. The reason is the result of what you process is a product of the input, in the case of hdr the "shadows" are captured a lot brighter than they appeared etc. If this wasn't the case the technique would be useless I guess. I'm new to hdr photography as you know and after thinking about it there is no real way around other than capturing for shadows and likewise for highlights > processing (hdr merge, tonemap etc etc) > final tweaks including darkening shadows again. Please correct me if this assumption is wrong and there is a better way, it seems from a technical point of view that this will always be the case however. Obviously with hdr as opposed to single image capture you can choose an optimal level of luminance for the shadows (which will be higher than if you'd exposed for highlights in single capture hence need for hdr merge) and the darks will be more detailed and less noisy than if they were captured at that level initially.

All things considered though depending on the scene a given viewing size might not be sufficient. To really draw the eyes to one thing and make the local contrast have full impact you need to make sure nothing out competes it, nor seems to make it balanced or in harmony with the other elements of the image as it needs to look different for the perception of strong contrast. It might be very high local contrast but we fail to percieve it due to various factors, size is one of these. I personally think hdr is most useful in the capture of a natural looking scene in however many exposures appropriate for an otherwise difficult or impossible single capture scene. You can still have deep shadow, it's just crisper and an ideal level of brightness (rather than muddy, noisy, clipped or close to shadow), likewise still have bright highlights (rather than blown white or un natural looking) but so many hdr images seem to abandon this for muted highlights and brighter shadows seemingly pulling everything toward the mid tones. I'm still experimenting myself and am far from happy yet (due to my own lack of skill and poor familiarity, I only know about making light maps for 3d stuff from hdr which is useless in this case) but what I have found is an otherwise rather severe strong curve adjustment brings the image back to a natural look (like a heavy s shape).

On another note (well a similar one actually) thanks for the programmes you're working on gui, they seem more suited to me in that they seem to be more about natural looks and minimal automation thereby making the process more flexible. The likes of qtpfs seems promising but tends to veer more toward the fully automation little control "one click > result" way of things, coupled with the very unrealistic results (not my thing but actually like the look of the unrealistic style of some HDR, if I want that effect I tend to paint it).

[Dave Humphries]

BR/Davey,

Just a thought, could you experiment with the "large image move head" effect by viewing an HDR picture at 100% so that it is say 4 times the size of the browser window requiring one to drag it around to see it all? This might give an idea whether local ranges ca be different from global ones, although I'm not sure how helpful it would be.

Mainly because this isn't normally the way most people view a large file on the net, their browser will normally "fit to screen", it is only if you click in to magnify you get this effect, for the pixelpeeper's amongst us.

Cheers,