Is the 4th Yellow Pixel of Sharp Quattron Hype?

Nom du Keyboard writes "Sharp Aquos brand televisions are making a big deal about their Quattron technology of adding a 4th yellow pixel to their RGB sets. While you can read a glowing review of it here, the engineer in me is skeptical because of how all the source material for this set is produced in 3-color RGB. I also know how just making a picture brighter and saturating the colors a bit can make it more appealing to many viewers over a more accurate rendition – so much for side-by-side comparisons. And I laugh at how you are supposed to see the advantages of 4-color technology in ads on your 3-color sets at home as you watch their commercials. It sounds more like hype to extract a higher profit margin than the next great advance in home television. So is it real?"

Comments

Clearly missing a trick.

[queazocotal]

To get truly astonishing pictures, they should add a black pixel, to improve contrast.

Re:Clearly missing a trick.

[jjoelc]

joking aside... some of the newer TVs with LED backlighting actually do something like this... Lighting up the picture with thousands(ish?) of independent LEDs (as opposed to a couple of souped up flourescent tubes) means they can selectively dim or turn off entirely sections of the backlighting. So when large parts of the scene are dirk, large parts of the backlighting is dimmed as well, thus increasing the contrast. It also saves a bit of power, making it easier for them to meet energy star standards, etc...

RGB

[Kell Bengal]

It strikes me that a better use of a fourth colour pixel would be to represent all those greens the RGB colour space doesn't actually represent.

Re:RGB

[Anonymous Coward]

Only is the camera recording the picture recorded that same color.

As it has been stated, adding a new color on the TV is literally the last place that it needs to be. (First the camera that films, then the storage medium(DVD?), then broadcast(HDMI?) THEN the TV )

Re:RGB

[Anonymous Coward]

That 1931 color gamut is misleading because it overempasizes greens. In fact, the original NTSC green primary was much closer to the peak, but as a result, yellows were too muted, so they changed it. But you're right - a turquoise primary would increase the RGB gamut significantly.

The ideal would be that all color information in video would be in device-independent xy color space instead of RGB. See LogLUV encoding for example: http://www.anyhere.com/gward/papers/jgtpap1.pdf

Re:RGB

[forkazoo]

Only one problem. No Y encoded in the data stream, so it has to be interpolated.

In some cases, it could actually be useful. While most cameras shoot with RGB sensors, most video compression is in some variation of YUV (1) color space. If you shoot on something like a Red One (2) camera, you get a RAW format with more than 8 bits (3) of color information. If you have a sensible post pipeline, you can go to YUV for your distribution format and have plenty of color data to completely fill out the 8 bit YUV data. YUV and RGB don't have identical color reproduction and gamut, so you can wind up with the odd situation where you shot on an RGB sensor, and you decimated to 8 bit data for distribution, but a normal 8 bit RGB display can't quite show every color that you have.

I wouldn't expect brick-shittingly amazing results on such a system. I'd need to see it in person and see a measured gamut chart to have any particular opinion on this particular display, but I can't dismiss the concept out of hand.

(1) : Y in YUV isn't Yellow, it's Luma. Still, the imperfect conversion between YUV and RGB means that a fourth primary could make it possible to more accurately show YUV data on an RGBY display.

(2) : "Red" is a brand name. "Red" in the name of the camera doesn't specifically imply any relationship to RGB color space or anything like that. The camera does use a standard RGB Bayer pattern sensor, though.

(3) : 8 bit color in this context is always "per component" rather than "per pixel" and doesn't imply old school 256 total colors palleted mode. In a X11 config file for example, this would be referred to as 24 bit color. Video guys are more interested in per-component colors because they always do operations on components. When you are writing misc. GUI software, you are generally more concerned with bits per-pixel because you would never care about how much space it takes to upload a fraction of a pixel to a video card since you have to upload a full pixel to display it.

(4) : This footnote doesn't correspond to anything in the text. After all that, I'm now just in the habit of writing footnotes.

Re:RGB

[Twinbee]

Parent is correct. Any colours around green and cyan are usually terribly unsaturated on most monitors. In fact, even in 'real life', it isn't theoretically possible to experience true cyan/aqua because the nearest direct wavelength will stimulate the red eye cone to some extent creating colour pollution.

There is a trick around this, which can be found by over-saturating the red cone. This weakens it temporarily, and then when shortly afterwards you see anything resembling cyan, it will appear as close to the true qualia as you could ever expect. The "Eclipse of Mars" illusion that follows in the below link demonstrates this for those who are curious:

http://www.skytopia.com/project/illusion/2illusion.html

Re:RGB

[rduke15]

I have the impression that you are implying that the bits per channel are related to the color gamut. That more bits per pixel or channel produce a wider color gamut. That is not the case, and the 2 are unrelated. More bits per pixel only give you more shades within a given gamut. In practice, more bits per channel are desirable in video production to allow finer control over color correction, without producing artifacts like banding.

Re:RGB

[jipn4]

That's misleading. A lack of a fully saturated green on a monitor is a limitation with the phosphors or dyes it uses. But monochromatic light of around 515 nm is pure, fully saturated green. Fully saturated green stimulates both your M and L cones ("G" and "R" cones); that's the way your eye works.

You can achieve non-physical responses from your photoreceptors via oversaturation, drugs, or electrical stimulation. That's interesting, but it isn't "green" and it isn't a "true qualia". Thinking of that as "green" is simply because you think of the M cone as a "green" cone and the L cone as a "red" cone, but those are just arbitrary names.

Re:RGB

[Twinbee]

That's my point though. How can a apx 515nm wavelength be a fully saturated green if the L cone is also being activated to some degree? That would be the extra pollution I'm talking about. I believe a much purer green would result if you somehow disabled the L cone. Unless you think we might see a more cyan/blue-like hue here?

To get a definitive answer, I would be interested to see what one would experience if you disabled two of the three S/M/L cones. I'm suspecting you would see pure red (disable S+M), green (disable S+L) and blue (M+L). Any research into that?

That's interesting, but it isn't "green"

What is it then?

Re:RGB

[6350']

and so it does.
http://en.wikipedia.org/wiki/Imaginary_color

Neat snippet from the article:
"At Walt Disney World, Kodak engineered Epcot's pavement to be a certain hue of pink so that the grass would look greener through the reverse of this effect."

Sneaky!

Re:RGB

[jipn4]

That's my point though. How can a apx 515nm wavelength be a fully saturated green if the L cone is also being activated to some degree?

Because all light of a single wavelength is automatically "pure"; it doesn't matter what your cone responses are. The cone responses are just a code to transmit that information to your brain. Your cone responses are such that they overlap (for good reason), but that doesn't keep you from seeing pure colors.

And actually, you perceive color contrast anyway, not absolute RGB values or wavelengths. So, even if you get a group of cones to produce a pure "green" response somehow, that will simply be processed as being part of a strong red/green contrast and result just in a vivid green percept.

Re:RGB

[jareds]

You seem to be under the impression that every fully saturated color is a spectral color, but this is false. If "hue" only includes the pure visible spectrum, then HSB will not include magenta (or, actually, the whole triangle defined by white, red, and violet, on CIE chromaticity diagram). Look in particular at the color wheel and the visible spectrum on the magenta page. People do use HSB, but the range of the hue must include a non-spectral "line of purples" to wrap around.

Not necessarily fake

[russotto]

Adding an extra phosphor can extend your gamut, increase your dynamic range within your gamut, or give you finer quantization within the gamut, or some combination of all three. The fact that your source material is provided as three quantities (YCbCr, not RGB) doesn't mean four phoshors won't help.

Doesn't mean it will, either.

Re:Not necessarily fake

[kc8apf]

If you've ever had a display calibrated, you'd know that even the existing RGB color space can't be completely recreated with existing RGB-based displays. The problem is in the inability of LEDs or LCD or plasma panels to produce light uniformly in the three color channels. If you can add a 4th channel that lets the RGB color space be more accurately produced by the display, then you will see an improvement. It won't make the source any better, but the output generated by the display for that input will be better.

Re:Yellow... yawn

[uglyduckling]

Like Octarine?

Re:Yes

[FatSean]

How do you decide which pixel to sacrifice for the colour gambit?

Re:Human retinas

[MRe_nl]

http://en.wikipedia.org/wiki/Opponent_process

What's wrong?

[rm999]

Representing yellow with a mix of green and red is already a hack. What's wrong with software determining that the color of a pixel is yellow and actually lighting up a yellow light?

Maybe a yellow light looks more convincing than a red and green light right next to each other. I'd want to see for myself before making blanket judgments.

It *could* be good

[__david__]

First, check out http://en.wikipedia.org/wiki/Gamut for reference. The sample gamut picture in the top right shows a typical CRT--lets assume for the sake of argument that LCDs are similar.

If you add a yellow LED to that it just isn't going to add much. The yellow part of the spectrum is already fairly well represented.

*But* if they also change the hue of the green LED toward the blue spectrum then it has a good chance of really opening up the gamut.

The people saying RGB is enough don't understand chromaticity--go look for gamut plots of your favorite output devices and see how little of the full spectrum of colors they can actually reproduce. Printers are especially embarrassing. Your eyes can really see a whole lot of color detail.

Re:Yes

[LynnwoodRooster]

The red one, just like on Star Trek.

Submitter fail.

[blair1q]

"And I laugh at how you are supposed to see the advantages of 4-color technology in ads on your 3-color sets at home as you watch their commercials."

But the script of the commercial is written almost entirely with deference to that fact.

The estimable Mr. Takei tells you, while you're no doubt ogling his adam's apple instead of listening, that he can't actually show you the difference itself, but, "I can show you this," wherupon he looks at the screen and gives his review in a single, somewhat gaudily overacted word.

I'm not sure how anyone misses that, since his behavior is utterly bizarre without the concept of telling-not-showing being in play.

Pictures just about sums it up

[phantomcircuit]

http://regmedia.co.uk/2010/05/07/quattron_4.jpg That just about sums up the entire article.

Local dimming has a problem

[tepples]

A lot of TV sets that use local dimming have a big problem showing starfields. The average color in a starfield is pretty dark, so the LED goes dim and not bright enough to show the stars. It really takes the punch out of Star Wars Special^n Edition if you can't see the stars.

Re:Yellow... yawn

[eggnoglatte]

XYZ space is not perceptually uniform. In particular, the green/cyan area in XYZ occupies a much larger area than would be justified by the eye's ability to distinguish colors in that range. Yellow on the other hand is very under-represented in XYZ.

If you look at the gamuts in a perceptually uniform space such as LUV, you'll find that LCD panels are actually fairly limited in the yellows.

Re:Yellow... yawn

[eggnoglatte]

True, but the regular LCD color gamut is smaller than the sRGB/Rec 709 gamut that is encoded in the HDTV video standard.

Basically, LCD panels use relatively wide spectrum color filters, so that they don't loos too much light in absorption. The result is a relatively small gamut - smaller than plasma or CRT.

Re:Careful What You Laugh At

[iwaybandit]

Didn't see that ad myself, but they probably used this effect.
Fechner color

is an illusion of color seen when looking at certain rapidly changing or moving black-and-white patterns. They are also called pattern induced flicker colors (PIFCs). Not everyone sees the same colors.