Slashdot Mirror
RGB to become RGBCMY
elgatozorbas writes "The basic color elements of television have not changed much since 1954; a half-century after RCA introduced the first color set, the RGB (red, green and blue) system used then still prevails. But Israeli company Genoa Color Technologies has broken the RGB barrier by adding one to three primary colors such as yellow, cyan and magenta, thus expanding - from 55 to 95 percent - the coverage of the visible color gamut. The promised result of this multi-primary color (MPC) technology is a television picture that, with its truer, more vibrant color and brighter image, looks more like cinema than video. Also covered in IEEE Spectrum."
Does this mean I should hold off on buying an HDTV?
It's almost enough to make me wish I was a mutant mother of a color blind son.
A truly revolutionary idea would be to include and project IR and UV in addition to RGB/CMY. Even though our eyes can't exactly 'see' IR and UV, they still form an important part of our realistic image perception. It's not unlike sounds above 20-25kHz in pitch; we don't 'hear' them, but our brain perceives them nonetheless and they are used for stereo imaging of a space.
Cretin - a powerful and flexible CD reencoder
Cheers,
Erick
http://www.busyweather.com/
Sometimes the most mundane improvements can be the best. All the people who swear by HDTV will be SOL, because they'll have hi-res, but improperly colored, television/movies.
My friends are going to be viridian with envy!
There are a couple of factual errors in this story that makes me feel uneasy.
. html
From the spectrum article:
While film used in cinema contains pigments that can create an infinitely large number of color variations, TV sets combine discrete amounts of red, green, and blue light to create a much more limited color range.
This isn't true: color slide film uses three layers, just like monitors do: http://www.imx.nl/photosite/technical/E100G/E100G
He says that in printing it's common to have inkjet devices that use six, seven, or even eight primaries.
There are good reasons printing uses so many primaries, but it's usually to make an evener tone. My consumer-grade printer has the traditional CMYK (cyan magenta yellow blacK), but it also has two additional colors: light-cyan and light-magenta. They chose these lighter colors so make the blending smoother and the ink spots less noticible; it wasn't to increase the gamut. Printers also use spot-color to make particular colors (such as a company logo) print without needing to use a halftone. These are all just gimicks to get around the fact that printing isn't continuous tone -- in projectors that are continuous tone, these tricks aren't needed.
Basically, it comes down to eyeballs... if you emulate the response curves that your eye is sensitive to, then you can't perceptually do any better.
The traditional RGB's and CMY's don't match these curves, so they define a gamut that can be improved on. For example, take this projector's gamut -- its green is far away from the eye's green, so it can't display the cyans well. But, the color model my company is using for its video product uses a much truer green so we can cover much more of the gamut.
disclaimer: IANACE (color expert), but my most recent project has been color calibration to precise standards.
HIV Crosses Species Barrier... into Muppets
I want to see what it looks like.
Lasers Controlled Games!
Also, each of the three colors commonly used (rgb) are artificially dark, with each one blocking about 2/3 of the light (since the only let that one color through). So if you think about it, your "white" background is really not as bright as it could be. Some DLP projectors I think use red, green, blue, and white to get some of this contrast back. But I think these guys have a more interesting idea. Your cyan pixel, letting through both blue and green light, would be brighter than either your plain blue or plain green or blue&green next to each other.
16 million colors should be enough for anyone.
Adding two extra colors to this kind of projection television has little impact on the price tag, says Simon Lewis, vice president of marketing at Genoa. He says the new Philips color-enhanced set, to be available next year, needs only a few additional filters and optical components to create the yellow and cyan light, with no changes to the more costly microprojection chip.
Right. Right when we've got all these plants around the world cranking out inexpensive TV's using LEDS and LCD, some whizzo comes along and says, "Hey, look, a great idea and all you have to do is retool everything, develop some newer technology and keep selling it all at the same pricing you're currently at!"
Perhaps the main challenge in converting a video stream from a three- to a five-primary color system is doing it in real time, says Maureen C. Stone, ...
Yay, now we really will need a computer in every TV! More components - more to go wrong, more power consumption, etc.
"How the algorithm does that, precisely, is a secret well kept by Genoa. "It's part of their intellectual property," Stone says.
Yay, more intellectual property. This should drive prices down.
<curmudgeon>
Why, back in my day we didn't have remote controls and we had a folded playing card stuck beside the tuner knob to keep the picture from doing funny things, and we liked it!
</curmudgeon>
I'm sure it will look lovely, while watching older stuff from the bad old pre RGBCMY days.
"Gilligan!"
I'm like, totally there, dude!
A feeling of having made the same mistake before: Deja Foobar
There are three primary additive colors and three primary subtractive colors. Cecil explains it rather well.
RGB is a set of orthogonal colors, and a linear combination of RGB can express any color in the universe. Similar comments apply to CMY.
Adding CMY to RGB to create RGBCMY does not buy you anything. Hence, the message starting this discussion thread is misleading.
Why is the television signal so poor in generating an image? The answer is unrelated to RGB. The answer is the the following. Prior to transmission, the analog RGB signal is converted into the digital YCbCr signal. (YCbCr is also an orthogonal set of colors.) Y, luma, is sampled at a reasonable rate, but the sampling system samples Cb and Cr at only half of the sampling rate for Y.
My guess is that RGBCMY is simply a clever attempt to use CMY to restore some of the samples of Cb and Cr that were discarded.
disclaimer: IANACE (color expert), but my most recent project has been color calibration to precise standards.
Parent has very good info, but if anyone wants additional reading, this guy is a color expert
(S(SKK)(SKK))(S(SKK)(SKK))
Yes, our eyes only have three types of cones, but unlike the color projected by a TV, they are not designed to respond to just one frequency of red, one of green, and one of blue. they have broad, overlapping response curves, each cone giving a different level of signal depending on the frequency of the light. The brain figures out the color based on the response of all three types of cones, not just the one that is active.
The stuff above is fact, the rest of this post is my pointless, unscientific, meandering hypothesis:
Obviously we use this concept with RGB signals to create colors like yellow, by tickling both the red and green cones at once with neighboring phosphors, but since the two colors are coming from very very slightly different places, the brain is not necessarily satisfied that it really is the color yellow. Basically, the more spectrum we can cover natively, the less chance there will be of someone's brain mumbling "that color doesn't seem... right"
Random and weird software I've written.
Nothing. This just provides a better way to stimulate them. If one had the technology to vary the intensity of red, green, and blue over an infinite set of real values, then RGB would be able to perfectly replicate any color. In reality, the RGB color model used in displays today varies these values over a finite set of integers. One gets the best ability to reproduce colors that are red, green, or blue. Colors between these on the spectrum can be simulated by mixing these, thanks to the three types of cones we used to process color on the retina, but if in order to reproduce a particular color, we need 255 parts red to 41 parts green, we simply cannot increase the intensity of this color without distorting it (shifting towards green, because we've already maxed red). Thus, any RGB color model is going to more accurately and vibrantly display reds, greens, and blues, and simpler blends of these (where all values are equal, e.g. cyan), anything else is going to be limited in the range, grosser in steps between intensity, and less vibrant at the max. Adding pixels that display actual yellow (light of precisely that wavelength, rather than a blend of red and green wavelenght light exploiting the trick to stimulate our red and green cones to the same levels that actual yellow-wavelength light would), adding these pixels would increase the ability to accurately display these between colors, despite the fact that, in theory, only RGB is necessary. It's easier to add more between color pixels than to up the intensity range and lower the steps between intensities.
"Convictions are more dangerous enemies of truth than lies."
NTSC throws away 3/4 of the colour information, and even HD throws away Half. From the article, it seems as if the chip is doing a lot of guessing and not "really" incresing the colour resolution. This sounds like a good way to go, since the Codec on the DVD won't have to deal with those extra colours; it's handled at display.
Not true, there are a few colors that are out of gamut on an RGB display.
-jim
Most folks don't realize, but there really are four primary colors. Most geeky types are familiar with the red, green, and blue cone cells in our eyes -- but the rod cells that are used for night vision have their own separate response spectrum, weighted heavily toward the blue/violet end of the spectrum.
That means you have four separate "detector systems" in your eye, each of which is sensitive to a different slice of the optical spectrum. In particular, you can distinguish shades of violet and magenta that differ only in the blue-cone/rod response levels.
Ever think about why blue light is used universally to signify "darkness" or "moonlight" on stage? It's because, in low light levels, your cones shut down and your rods -- which in bright light connote blueness -- are the only part of your retina that works well.
It's also the reason why night-vision flashlights are red, and why blue LEDs appear so bright when used as flashlights. The red light doesn't stimulate your rods, preserving their sensitivity; and the blue light gives you extra rod stimulation per unit power, making blue LEDS very efficient as nighttime illumination.
I'll wait for HDR display and feeds, thanks.
Judging from the gamut chart for this RGBCMY, the boost in color range is primarily in yellows and cyans. Gold, as they note, would be a good application. Cyan.. well, that's mostly skies - and those already appear just fine on TV. A fairly decent increase in magentas/purples as well (when taking the assymetric lobe into account), but again.. not seeing its application much.
Unless following the British royal family (lots of golds and purples) a lot, it doesn't appear to offer all that much. Especially considering movie people butcher things anyway (DVD gives a more stable picture, sure.. at the compromise of mpeg artifacting and even encoding issues.. twitches ever 25 frames are annoying - luckily only a few suffer from this).
On the other hand, a higher dynamic range would be immediately noticeable anywhere.
A sequence with the sun glaring into the camera ?
A car's headlights shining at the camera ?
Highlights on objects ?
Blown-out surfaces from bright lighting ?
All that could then more accurately be represented. And thanks to most things still being shot on film, or already on 10bit CCDs with, formally, underexposure but a gain for the operator, a good bit of extra range is already available in previous and current productions.
Whilst RGBCMY would only really be of use for film (as in, actual film) productions, as digital cameras are in much the same RGB limbo that current displays are.
Wow, this is really cool.
o dori.php?=conference
g h.php?pageID=conference
There's a whole bunch of these wide gamut and high dynamic range displays suddenly.
At SIGGRAPH this year, there was a 6-primary (RGBCMY) projection system called IRODORI on display in emerging technologies:
http://www.siggraph.org/s2004/conference/etech/ir
There was also a high dynamic range display (capable of a greater range of brightness) from Sunnybrook Technologies at E-Tech:
http://www.siggraph.org/s2004/conference/etech/hi
And then I saw a few displays on the exhibition floor from NEC with a "WG" specifier for "Wide Gamut". NEC's WG monitor is still RGB but with purer R, G, and B phosphors to obtain a gammut wider than Adobe RGB.
And now there's this one. Way cool.
I can't wait till this becomes more widespread. The question becomes, what will the next color standard be for use in applications and APIs? It doesn't make sense to actually encode color as 6 values for display, since (most) humans only have three kinds of cones. It would make more sense to use something like CIEXYX for color interchange in that case. Especially if we're going to have this wierd mix of HDR and various wide gamut displays around for a while, each which has slightly different needs for color output. Best to just go with a neutral, well-defined intermediate colorspace.
CMY(K) are Subtractive Colours. (You add them together to get black)
... RGB + CMYK negate each other.
... while LCD's naturally use a CMYk approach
Hehe! No, this is quite false, quite a number of ways.
First of all, colors of light are additive, colors of pigment are subtractive. This is true regardless of which colors you choose. If you had a monitor using the CYM model, you could not produce red, because monitors, being light emitting devices, are always additive, never subtractive, mixing C and Y would add their lights, not subtract leaving just the G. Because of this, you cannot get a lot of colors. However, you can get white, by adding C, M, and Y together. Since monitors are additive, adding CYM makes white, not black.
The LCDs we use today are light emitting, not light reflecting. Thus, they naturally use an RGB color model. If they did not emit light on their own but only reflected like, like a sheet of paper, then their natural color model would be CYM(K). But that's just not how things work.
"Convictions are more dangerous enemies of truth than lies."
Well, true they have to expand the gamut of existing RGB data artifically, but this is different from what you can do in photoshop. In this case the display can actually show more real colors than a conventional RGB display. Put the two monitors side-by-side, and you will be able to see colors on a RGBCMY monitor that simply cannot be reprodced on any normal RGB monitor. Have you ever taken a digital picture of a beautifully intense blue stain glass window, or some brightly colored flowers, and been disappointed when you got it home to see how bland the colors were on your monitor. The gamut captured by the camera is part of the problem, but even if it captured the colors perfectly, current monitors still couldn't display the results. These new wide gamut monitors should be able to do much better.
Having to "make up" the additional color data is just a temporary measure until content creation software and image acquisition hardware catches up to the gamuts possible with these new monitors.
I, for one, welcome our new RGBCMY masters.
As I recall, a linear combination of RGB can express any possible color -- if you allow for negative amounts of the components. A really bright yellow might be 1 R + 1 G - .2 B for example.
That's still a linear combination, but just one that's not particular useful in the real world of phosphors and filters.
Thad
I love Mondays. On a Monday, anything is possible.
It's probably simpler than you think.
CMY are really "combinations" of R G and B.
So, what's happening is that they are tossing in "intermediate" colors in roughly the same way as a 6 or 7 color printer. The exact equations are probably proprietary, but the process is pretty standard.
This comes in to play at two places. First, HDTV has a pretty ambitious color gamut, so videos designed around the HDTV gamut will look better, assuming of course that the source footage is equally high quality.
Second, there are colors that your eye can perceive that are not representable by the RGB system.
Overall, the research is already done. There's actually quite a few different ways to represent this data. PhotoCDs already use it. You want to use L*a*b or XYZ or one of the other CIE color systems.
I think it's interesting, but when I read the headline, my first thought was "Gee. What took them so long?"
Gentoo Sucks
CMY are really "combinations" of R G and B.
This is false. C, Y, and M are different wavelengths of light from R, G, and B. Because the human eye only has receptors for R, G, and B, we can't distinguish between equal quantities of R and G and a single wavelength in between the two, namely Y. In other words, we are able to trick the eye into perceiving a full color spectrum using only three different wavelengths of light.
Trees can't go dancing
So do them a big favor
Pretend dancing stinks!
CMY are really "combinations" of R G and B.
They are on your standard RGB monitor, but not in the general case. For example, take a look at the CIE "Tongue" chart displayed e.g. here. With you monitor, you can only display colors in the red, green, blue triangle, but one could add pure cyan at 490nm and actually increase the area/gamut.
Second, there are colors that your eye can perceive that are not representable by the RGB system.
That would be the good old RCA, phosphor based RGB system. If you ran your display with e.g. lasers with 410, 520 and 700nm respectively, you could get a gamut that's almost indistinguishable from the full gamut the average eye can percieve. The smaller area covered in the green region on top of the chart would probably be neglegible due to the decreased capability of the eye to distinguish between greens. So, not RGB is the problem, but the technology to record and display it.
... you're going to need a format that preserves color information in the new 5 color system if you're going to exploit the real improvements in this color technology: closer reproductions of actual color.
Absolutely not true.
For people with normal color vision, in addition to the "rod" pigment (which is not a significant player in color perception and daylight central vision) there are three color receptor pigments located in the "cone" cells, which have broad reception peaks with well-known shapes. The response of those three sets of cells to an image can be accurately modeled by using three sets of sensors and filters that model the three pigments' frequency response.
The problem comes when, given this measurement, you try to stimulate a viewer's cone cells to produce the response equivalent to the light you measured. If you just pick three color phosphors at the peak of the three dyes' response curves, you find that the colors don't stimulate JUST the cones you intended. The green light, for instance, will strongly stimulate the green-responsive cones. But it will also weakly stimulate the red and blue cones. Similarly, red light will strongly stimulate red cones, weakly stimulate green cones, and very weakly stimulate blue cones. Ditto the other way around with blue light.
This has two effects:
First: Even within the range of combinations of stimulus the three light sources can produce, simply playing back the signal will cause the results to be somewhat more pastel than the orignal scene. This can be compensated for to some extent - by subtracting out appropriate amounts of each color's signal from the signals going to the others color emitters.
Second: You can't make the emitters emit a negative amount of light. The result is that there are scene colors, saturated and nearly-saturated colors between the phosphor colors you chose for reproduction, that can produce color sensations that these three screen colors can't reproduce. These scene colors will ALWAYS apper somewhat washed-out if you only reproduce the image with three screen colors.
So with three values you can accurately transmit any color a normal eye can see. But with three phosphors you can't make the eye see some of these colors.
The two-dimensional representation of the relative responses of the three dies looks something like a spearment leaf with the base sliced off. (See figure 12 of this web page. And thank you, canavan) The edge of the leaf represents the response to a pure spectral color, and regions within it to mixes of colors. If you try to reproduce the response with three phosphor colors, you are picking three points on the leaf edge and drawing a triangle between them. By adjusting the relative amounts of light from the three phosphors you can produce a stimulus corresponding to any point WITHIN the triangle. But you can't produce one corresponding to the arcs of the leaf that are outside the triangle.
But by picking more points along the leaf edge you can draw a polygon and hit any point within it. This covers more of the leaf and leaves fewer colors missing. (Indeed, just a couple extra points can give you most of the leaf.)
You still send the signal with the three values corresponding to the response you want from the eye. But now your monitor processes it into more than three colors to put on the screen, to get the eye to respond more closely to the response it would have had to the original scene.
(Note that people with some forms of color blindness have cones with pigments that have abnormal frequency responses. Such people will not see a color TV image as right even with this upgrade, because the camera will not have correctly encoded what THEIR eyes would have seen. They need a camera with a different response, and yet another set of phosphors in the monitor, to get a good match.)
Bantam Dominique roosters crow a four-note song. Once you've heard it as "Happy BIRTHday" you can't NOT hear it that way
I had never tried to think outside the RGB world because it 'technically' displays all colors, though it struck me that the colors in-between RGB will come out dimmer than they should.
No, RGB technically displays more discrete colors than our eye can see. That does not mean it "displays all colors." There are some colors RGB displays that we cannot distinguish between, and there are some colors we can distinguish that RGB cannot display.
24 beers in a case, 24 hours in a day. Coincidence? I think not!
the human eye only has receptors for R, G, and B
Mantis shrimp have at least eleven different receptors, and lots of birds and fish have four or five. So I guess it's the logical direction to go once the human market for RGB monitors reaches saturation.