Slashdot Mirror
High Dynamic Range Monitors
An anonymous reader writes, "We are seeing more and more about high dynamic range (HDR) images, where the photographer brackets the exposures and then combines the images to increase the dynamic range of the photo. The next step is going to be monitors that can display the wider dynamic range these images offer, as well as being more true-to-life, as they come closer to matching the capabilities of the ol' Mark I eyeball. The guys who seem to be furthest along with this are a company called Brightside Technologies. Here is a detailed review of the Brightside tech." With a price tag of $49K for a 37" monitor (with a contrast ratio of 200K to 1), HDR isn't exactly ready for the living room yet.
Of course one of the other principal arenas where monitors like this are valuable is in medical imaging. One of the serious shortcomings in the migration of radiology to digital formats is the reduced quality of the images as compared to film. The dynamic range of film is simply so much greater than can be achieved with standard CRTs or LCD monitors that there is a real danger of missing out on very subtle changes in X-Rays for example. While it's true that image processing can make up for some differences, digital still can't quite compete with film for many purposes including data density in many cases.
Visit Jonesblog and say hello.
I've been seeing these at Siggraph for years. They do look very nice. You basically need a very bright light source (not hard) that doesn't generate too much heat (a little harder) and a way to modulate that light over a very large range (harder). It would be fun to have a converter for DSLR RAW images to display in HDR, or the usual bracketed ones.
The examples they usually use are things like light streaming through stained glass in a church, where normally you'd either only see the stained glass properly exposed, or the rest of the room, but not both. It does work to very good effect in those instances, and heightens the "window into the world" effect that high resolution displays have. If this were to be combined with 2X HD resolution 60P motion video (about 4,000 pixels across) it would kick serious ass as the next 'Imax' lifelike motion picture display.
Oddly enough, the captcha for the post reply screen right now is 'aperture'.
Since "true" HDR consumera camera's don't exist (anyone know?), it can be faked, quite convincingly, I might add.
i.e.
"It's a feature in Photoshop CS2 or Photomatix or FDRTools."
Even black and white can be support HDR. This is a great B&W example of why 8-bit greyscale just doesn't cut it.
--
"The difference between Religion and Philosophy, is that one is put into practise"
Mammography has gone completely digital. Why? Because the quality of the imagery is lightyears better than what you can get for film. Couple that with rapid processing from a scanner laser and throw in algorithms that contrast enhance areas of nearly neutral density and you have a recipe for catching growths that would otherwise miss detection.
A good, excellent radiologist could detect subtle differences of about 80% that of a standard person. I'd give you the exact quote but it's been a while since I remembered the data- suffice to say I was impressed at the level (in controlled lighting situations) that they were able to see in film.
A good medical display is a peeled LCD- all the colors have been chemically removed from the surface- and has typically a brighter backlight and another polarizer to knock down the lmin even further. This gives you better dynamic range that is easily adjusted faster than film can- want to zoom in? No problem- touch and zoom- or if you had film, grab a loupe (or crane your head closer). Digital wins hands down.
Yes, if you digitize a negative you have a data density that can't be reached very easily (I used to estimate this for a job for large quantities of imagery and at high quality ratios- 2 micron spot sizes). But frankly alot of that information is useless- you don't need to know what isn't of relevance.
The most important aspect of digital imaging is proper viewing environment- something no one seems to get. Reduce the lighting of the area to 0.5 fc and remove any sources of glare off the monitor. Wear dark clothing. Have wall wash lighting appropriate to about 3-9 fc. Have surfaces neutral gray. Ceiling black.
Digital definately competes with film in many markets for medical xray- Mammography was just the easiest to choose because it has been such a radical change in such a short time period.
I should note I used to work for Eastman Kodak and did work with other individuals on these digital products (specifically, algorithms)... but I'm not biased because of that. Just the simple truth- from the raw data I've seen I'll feel happy and safe knowing my wife gets a digital mammagram every year.
The BrightSide DR37-R EDR display theoretically has an infinite contrast ratio. How? Because it can turn individual LED backlights off completely (see How It Works), it has a black luminance of zero. When you divide any brightness value by this zero black value, you get infinity.
It goes from 0 to 4000cd/m^2. Their comparison model, the LVM-37w1, goes from 0.55 to 550cd/m^2.
So this toy gets as close to true black as you can get - "off", thus constrained by the ambient light level. For white, they manage 4000cd/m^2, or comparable to fairly bright interior lighting.
Consider me impressed, but realistically, this only amounts to roughly an 8x brightness improvement over the best normal displays, with true-black thrown in as a perk (they suspiciously don't mention the next lowest intensity, no doubt because it goes back into the realm of a contrast ratio of only a few thousand.
Excellent point. Truth be told I'd much rather see the color depth approached first. They've gotten better but for film level work none of them display full color resolution. Frustriating that the software will handle 48 bit, three channels of 16, but the monitors won't. Mostly becomes an issue when you are working with a lot for gradient images, skies and such. You still get some pixelation that isn't in the actual image file. Then again if you're doing TV who cares. They call it NTSC, never the same color, for a reason.
Don't be an idiot. You don't have to have "a black hole in the display" for a pixel to have effectively zero brightness, you just have to have it not generate any light (excluding blackbody radiation, which is negligible in the visible spectrum at room temp). One of the photos in TFA is of the monitor displaying a black screen in a dark room; you can't tell it from the surroundings. The pixels can be individually completely switched off (actually, that's not strictly true, a group of a few pixels can be switched off), giving a contrast ratio of (max brightness)/0 -- hence the divide by zero error, as the grandparent said.
What's purple and commutes? An Abelian grape.
No. Think about it: unless you're really pressing your nose right up to the screen, for a monitor to display a reflection of the image on the face of whoever's looking at it, it would have to radiate at a single angle (probably perpendicular) only. You wouldn't be able to see the whole screen, only a few pixels per eye at any one time. Ever stood in front of a projector screen and looked at the projector? Like that. It would be utterly useless as a monitor.
N.B. if you have something like the left side of the screen one colour and the right side a different one, you may well be able to see that by looking at your face, but that's more due to the fact that your face isn't flat; the left side slopes backwards from centre to edge, and vice versa for the right side. You certainly wouldn't be able to see detail.
What's purple and commutes? An Abelian grape.
> f the room isn't entirely dark (and a room with an HDR display in it isn't perfectly dark if you're not watching only renditions of /dev/zero), then the reflectiveness of the display surface limits the contrast, unless it's a black hole, as I mentioned before
Nope. The specified contrast is the ratio of EMITTED, RADIATED light from a bright pixel to EMITTED, RADIATED light from a dark pixel. Certainly, ambient light will reduce the effective contrast in reality, but the definition of specified contrast ratio assumes no ambient or reflected light. Obviously. How could it be otherwise, or the contrast ratio would be meaningless unless you specify everything from amount of ambient light to the colour of the walls along with it.
What's purple and commutes? An Abelian grape.
I can't wait till this goes mainstream. Then I'll be able to watch a video of a solar eclipse and actually get blinded by the image. Coool.
My bicyles
I can't seem to find a reference to it online... I'd appreciate one if someone has one... but circa 1960 the Polaroid company developed a film for recording nuclear tests, which was similar to three-layer color film except that the three layers, instead of being sensitized to different colors, were given emulsions with widely different sensitivities. The fastest emulation was similar to Kodak Royal-X Pan, ISO 1600, and the slowest was similar to Kodak Microfile, and if I recall correctly had an ISO speed of something like 0.1
The result was to extend the useful dynamic range of the film by a factor of 10000 or so--more than a dozen additional f-stops of latitude, or extra Ansel zones, if you like.
The film was processed in regular Kodacolor chemistry (IIRC), each layer coming out a different color. In color, the result was a "false color" image displaying a huge dynamic range of light intensity; or, it could be printed as black-and-white using different filters to select different intensity ranges.
In effect, the photographer was automatically bracketing every shot by a dozen F-stops, in a single shutter click.
It was an incredibly neat hack. I wonder whatever became of it?
"How to Do Nothing," kids activities, back in print!
There is no such thing as a 0.0 - 1.0 visual range. The human visual system is floating point, pretty much literally. You have your exponent, which is how well adapted your eyes are to the light, how dilated your pupil is, etc. You have your mantissa, which is the relative intensity within your current visual field. Physiologially, we have about 28 bits of exponent and about 10 bits of mantissa. So, proper HDR is floating point. But we're not quite there yet.
In both audio and video, this whole idea of quantizing a 0.0 - 1.0 interval is a compromise wrought by insufficient numerical resolution. It has nothing to do with physics or perception or anything else. Once you realize that, you should also realize that the idea of "going outside" the 0.0 - 1.0 range is absurd. You don't go outside the range, you expand the range so as to better approximate the incredible human senses. As long as we're using fixed point image formats and digital video standards, there will always be a range, and we'll always be inside the 0.0 - 1.0 range, and it will always be a compromise.
Audio professionals have worked out their terminology far better than graphics guys have. Audio guys talk about dB, decibels. The reference point is 0db, which is as loud as your amp will go. When you add more bits, you're adding more quiet, not more loud. If you want more loud you buy a bigger amp. Each additional bit gives you 3dB more quiet, and you'd better hope your equipment has a low enough noise floor that you can hear all that fresh new quiet.
So what are you saying? What's the difference between HDR and 48 bit color? To use an analogy to audio, you seem to be saying that HDR is about more loud, and 48 bit color is about more quiet. But as you go on to point out, they're really just the same thing. No matter what, you've got a clipping level (the maximum luminance of your output device), you've got a noise floor (the minimum luminance of your output device), and hopefully you've got enough quantization levels in between for perceptual linearity. That's why HDR and color depth are joined at the hip. You can't get one without the other. There is no meaningful distinction.