Google announced an updated version of their Nexus 7 tablet this morning. Central to Google’s pitch was the improved display with both more pixels and more color. The device does feature an impressively high resolution, packing 2.3 million pixels into a 7″ form factor. But, I’m more interested in the color performance and, on this point, Google was vague offering only that the display, “has a 30% wider range of colors.”
What do they mean by that?
It depends on their frame of reference- what color space they are using and what color gamut standard they are comparing against. Since Google talked about the accuracy of HD video at their event, let’s assume that they are referring to the HDTV broadcast standard (rec.709) and using the common CIE 1976 (u’ v’) color space.
When I measured last year’s Nexus 7, I found it could only reproduce about 82%* of the colors found in the rec.709 standard. Color reproduction was not accurate and a little bit undersaturated on this device:
Color gamut of Google’s previous generation Nexus 7 versus the HDTV broadcast standard (rec.709). Plotted in CIE 1976 (u’ v’).
With just a simple calculation, increasing 82% by 30%, you’d get about 106% coverage of the HDTV broadcast standard. While that’s actually a slightly wider color gamut than the standard, it is not uncommon for device makers to use a wider color gamut in order to guarantee the color spec across all devices with some room for manufacturing tolerances. This means video and web content should be displayed accurately and it could make for a great looking display.
We’ll order and measure one as soon as they are available to verify so stay tuned…
* note: I always measure coverage of broadcast standards, not simply total area since that can be misleading. However, in this case, coverage and area are nearly the same since the Nexus 7’s gamut is smaller than rec.709.
Last week I looked at the three “P’s” of human color perception– physical, physiological and psychological– as a way to help define a color gamut for the ideal display. Based on real world examples from art and commerce, I concluded that the range of colors found in nature, as measured by Pointer, provided the best fit with our two design goals which were an accurate and exciting, immersive experience.
This week, I’d like to get a little more practical and take a look at existing color gamut standards to see what we might realistically be able to achieve today.
What fits best?
Color gamut of 4,000 surface colors found in nature as measured by Pointer in 1980 against the color gamut of the iPhone 5.
The first thing you’ll notice about Pointer’s gamut (pictured above again) is that it’s a pretty odd, squiggly shape. This means it is going to be difficult to cover efficiently with a three primary system that mixes just red, green and blue to create all the colors we see, like the LCD found in the iPhone. In order to cover Pointer’s with just those three colors, we’d need to make them extremely saturated. There are proposed standards that take this approach, such as rec.2020, but since they are not practical to implement today from a technology standpoint I’ve decided to ignore them for this discussion.
For the near future, we’ll need to rely on just three colors to get the job done, so what can we do now? Let’s look at two popular wide color gamut standards: Adobe 1998 and DCI-P3:
Current wide color gamut standards Adobe RGB 1998, commonly used by pro photographers and designers, and DCI-P3, used in digital cinema, compared to Pointer’s gamut in CIE 1976
Let’s start with Adobe 1998. Many people are familiar with this color gamut since it is found as an option on many consumer cameras and it is popular among creative professionals. It certainly covers a significantly wider range of colors than the HDTV broadcast standard with a very deep green point. The rich cyans that we talked about in the movie “The Ring” would look great in Adobe 1998. But, we’re not getting any more of those exciting reds and oranges. In fact, Adobe’s red point is identical to the HDTV broadcast standard.
What about DCI-P3 then? Designed to match the color gamut of color film and used in cinemas all over the world, DCI-P3 has a very wide gamut. The reds are particularly deep and, of course, all of the colors from the movies we looked at are covered. Still, it’s missing a lot of the deep greens found in Adobe 1998 and only just fits the green Pantone color of the year. So DCI-P3 is not quite perfect either.
What about a hybrid, custom gamut?
What if we combined the green from Adobe with the red from DCI-P3 and their shared blue point? We’d end up with pretty good, high 90’s percentage coverage of Pointer’s gamut, coverage of all of the existing HDTV broadcast content, full coverage of cinema content from Hollywood and a superior ecommerce experience with most of the colors from the natural world covered.
Hybrid color gamut standard that combines the green point from Adobe 1998 with the deep red of DCI-P3
Looks pretty great and we can make displays now that cover this color gamut with today’s technology. But how would it work on the content side? Would we need to get together and agree on this new standard and then wait for years while it is slowly adopted by content creators and display makers?
Next week we’ll look at how content delivery might evolve to support gamuts like this without the need for major changes to broadcast standards.
Last week I set out to define the ultimate consumer display experience in terms of color performance. I laid out some potential color performance design goals for an ideal display, suggesting that such a display should be both accurate and capable of creating an exciting, immersive experience that jumps off the shelf at retail.
Can we achieve both goals? To find out, let’s start by looking at how we perceive color.
The color of objects that our eyes see in nature is determined by three things: physical, physiological and psychological:
The color of objects that our eyes see in nature is determined by three things: physical, physiological and psychological.
The physical component of our color perception is a constant based on the laws of nature. It is a combination of the quality of the illumination or light source, in this case meaning spectrum it contains, and the reflectance of the object. In the image above, the ball appears red to the eye because it is reflecting red light, while absorbing most the other colors from the light source.
The physiological part of our vision is also a relative constant that is based on the electrochemical processes of the eye. The back of the retina contains photoreceptor nerve cells which transform incoming light into electrical impulses. These electrical impulses are sent to the optic nerve of the eye and onto the brain, which processes and creates the image we see. And that’s where the psychological component comes in.
Let’s look at how each of these components might affect display color performance, starting with the physical, which ought to be something we can measure.
Fortunately, a guy named Pointer has done this for us. For his 1980 publication, Pointer measured over 4,000 samples and was able to define a color gamut of real surface colors, of objects found in nature. The result is commonly called “Pointer’s Gamut:”
Color gamut of over 4,000 colors found in nature as measured by Pointer against the color gamut of the iPhone 5.
This already seems like a great place to start. It immediately looks like a great fit our first ultimate color experience criteria which was accuracy. If we could accurately capture and reproduce all of the colors found in the natural world it would make for a much improved, more accurate ecommerce experience, for example.
But how important are those extra colors? Looking at Pointer’s gamut mapped against the color gamut of the latest iPhone in the chart above, you have to wonder if we really come across these deep cyans and reds in everyday life. Are they just infrequent, rare colors or something worth pursuing for our display?
Turns out we do. As an example, Pantone’s color of the year for 2012 was a deep emerald green that falls outside of both the iPhone’s gamut and the HDTV broadcast standard. This is an important and popular color that appears a bit too yellowish on your computer monitor when you are shopping for the perfect tie on Amazon. So there are some really important colors outside of what the iPhone can display today.
But, what about our second criteria, the lifelike, exciting, immersive experience we want to give consumers? Is the gamut of the natural world enough?
If we look at the second component of the visual system, the physiological component, we’ll see that we can actually perceive a much wider range of colors. The cells in the back of retina can actually detect the entire range of the CIE diagram. That’s almost double the range of colors that Pointer found in nature:
Color gamut of the average human eye vs gamut of colors found in nature as measured by Pointer
This is starting to sound like a much more immersive experience. Maybe we ought to pursue the full color capability of the human eye just like the industry has done for high, “retina” resolutions.
It sounds great but it would be a tall order. It would take quite a lot of power, brightness and extra bit depth to even begin to think about covering a color space this large. There certainly would be a high price to pay in terms of design tradeoffs to get there. So are there any truly valuable colors contained in that extra space, similar to the Pantone color in Pointer’s gamut, that would make us want to go for it?
This is where the psychological component comes into play.
Seeing is not passive. Our brains add meaning to the light that our eyes detect based on context and experience and memory. We are continuously and actively re-visualizing the light that comes out of our retinas.
This may seem hard to believe but this fun demo created by neuroscientist Beau Lotto does a great job of showing just how much our brains actively interpret and change what we see.
The color of the chips has not changed in the video above, just our perception of the color. What’s happening here is our experience is telling us that the color chip in shadow must actually be a much brighter color than the chip under direct illumination, so our brain is just making the correction for us on the fly.
Artists absolutely play on this psychological element of our perception of color, sometimes using totally unrealistic or hyper real colors to make us feel or experience something new or help tell a story. In fact, one of the most influential art instructors of the 20th century, Josef Albers, once said that, “the purpose of art is not to represent nature but instead to re-present it.”
Monet’s The Poppy Field, near Argenteuil
So, whether it’s Monet using saturated and contrasting colors with equal luminance to trick our brains into seeing poppy flowers sway in an imaginary breeze in a 19th century painting or modern films which sometimes rely on the wider gamut capabilities of color film and digital cinema projection to create uniquely cinematic experiences for audiences.
Movies like “The Ring,” for example, which used a deep cyan cast throughout much of the film to create tension and help tell a scary story. Or Michael Bay’s “Transformers” movies, which use deeply saturated oranges, reds and teal greens to create an exciting, eye-popping palette appropriate for a summer blockbuster sci-fi movie about giant robots:
There’s certainly a place for wild, unexpected colors in art. But, as we go through some of these examples, I think we’ll actually find that there is a huge range of expression possible within the gamut of surface colors that Pointer measured. The full range of gamut detectable by the human eye, while exciting to think about, is not really necessary to deliver both accurate and pleasing (engaging) color to our visual system.
So where does that leave us?
In my next post I’ll look at existing wide color gamut standards and content delivery mechanisms to see both what we can do today and what’s next for wide color gamut displays.
Finally getting around to posting a follow-up to a follow-up to John The Math Guy’s recent series on color gamut size, colorblindness and tablet displays. I thought I might be able to at least shed a little more light on his question about the differences in color accuracy between some of these devices.
In his testing, John found no statistically significant difference in scores among different people taking the EnChroma colorblindness test on different devices. I found this somewhat surprising since, in my experience, even tablets with similar color gamuts tend to show colors with very different levels of accuracy.
As you can see, the iPad mini and Nexus 7 each produce very different colors, even for those colors that are actually inside their gamuts.
For example, even though the iPad mini has enough gamut coverage to accurately display the Gretag chart’s deepest blue, it cannot do so without distorting the image in another way. This is because of data in the underlying image standard- most content today is encoded in the sRGB standard. If the iPad were to show that Gretag blue correctly, it would not have enough color saturation headroom left over to show you a different color if a deeper blue, say right at the bottom of the sRGB triangle, were called for.
A good real world example of this can be found in the picture below of my bloodhound, Louisa, racing down the beach at Carmel, CA. The middle of the sky in this image is right on the edge of the iPad’s color gamut, very similar to the Gretag blue in the charts above, while the deepest blues found in the ocean fall outside the iPad’s gamut.
If the iPad were striving for accuracy at all costs, it might map both colors right on top of each other at the edge of the gamut. There’d be no visible difference between the two in this case and the quality of the image would suffer but at least the sky would be accurate. In order to avoid this scenario, the designers of these devices have decided to compromise on accuracy so they can show a full range of color differences to the user.
They do this by remapping colors inward, away from the edges of the gamut, effectively compressing the gamut even further so that otherwise out-of-gamut colors can be seen. This is a good solution given the gamut limitations of the device since it results in more pleasing, if less accurate images.
As newer devices trend towards wider color gamuts this kind of compromise should become a thing of the past. In fact, tablet designers may be working on the reverse issue- how to avoid oversaturating images that were encoded for smaller gamuts.
Great, how does this relate to colorblindness again?
iPad mini vs Nexus 7 color accuracy comparison in CIE 1976
Taking another look at the Gretag results from the two devices plotted on top of each other, there clearly are major differences. But, in the reds and greens, two colors associated with a common form of color blindness, the devices are relatively close. So, the simple answer may just be that colorblindness tests do not require pinpoint accuracy to be effective, at least as basic screening tools.
For many who are new to the world of display measurement, the prevalence of two distinct, but often-interchanged color spaces can be a source of confusion. Since my recent post about the color performance of Apple’s new iPad, a number of people have asked about this topic, so I thought it would be worth a closer look.
In the world of displays and color images, there exists a variety of separate standards for mapping color, CIE 1931 and CIE 1976 being the most popular among them. Despite its age, CIE 1931, named for the year of its adoption, remains a well-worn and familiar shorthand throughout the display industry. As a marketer of high color gamut display components, I can tell you from firsthand experience that CIE 1931 is the primary language of our customers. When a customer tells me that their current display “can do 72% of NTSC,” they implicitly mean 72% of NTSC 1953 color gamut as mapped against CIE 1931.
“…we strongly encourage people to abandon the use of the 1931 CIE color diagram for determining the color gamut… The 1976 CIE (u’,v’) color diagram should be used instead. Unfortunately, many continue to use the (x,y) chromaticity values and the 1931 diagram for gamut areas.”
So why are there two standards, and why are we trying to declare one of them obsolete? Let me explain.
What is a color space?
First, a little background on color spaces and how they work.
While there are a number of different types of color spaces, we are specifically interested in chromaticity diagrams, which only measure color quality, independent of other factors like luminance. A color space is a uniform representation of visible light. It maps the all of the colors visible to the human eye onto an x-y grid and assigns them measureable values. This allows us to make uniform measurements and comparisons between colors, and offers certainty that images look the same from display to display when used to create color gamut standards.
In 1931, the Commission internationale de l’éclairage or CIE (International Commission on Illumination in English) defined the most commonly used color space. Here’s a look at the anatomy of the CIE 1931 color space:
What makes a good color space?
An effective color space should map with reasonable accuracy and consistancy to the human perception of color. Content creators want to be sure that the color they see on their display is the same color you see on your display.
This is where the CIE 1931 standard falls apart. Based on the work of David MacAdam in the 1940’s, we learn that the variance in percieved color, when mapped in the CIE 1931 color space, is not linear from color to color. In other words, if you show a group of people the same green, then map what they see against the CIE 1931 color space, they will report seeing a wide decprepancy of different hues of green. However, if you show the same group a blue image, there will be much more agreement on what color blue they are seeing. This uneveness creates problems when trying to make uniform measurements with CIE 1931.
The result of MacAdam’s work is visualized by the MacAdam Elipses. Each elipse represents the range of colors respondents reported seeing when shown a single color, which was the dot in the center of each elipse:
A better standard
It was not until 1976 that the CIE was able to settle on a significantly more linear color space. If we reproduce MacAdam’s work using the new standard, variations in percieve color are minimalized and the MacAdam’s Elipses mapped on a 1976 CIE diagram appear much more evenly sized and circular, as opposed to oblong. This makes color comparisons using CIE 1976 significantly more meaningful.
The difference of the CIE 1976 color space, particularly in blue and green, is immediately apparent. As an example, lets look at the color gamut measurements of the iPad 2 and new iPad we used in an earlier article. Both charts do a reasonably good job of conveying the new iPad’s increased gamut coverage at all three primaries. But, the 1976 chart captures the dramatic perceptual difference in blue (from aqua to deep blue) that you actually see when looking at the displays side by side:
The increased gamut of the new iPad is worth testing. Next time you find yourself in an Apple store, grab an iPad 2, hold it alongside a new iPad, Google up a color bar image and see the difference for yourself.
So, why do we still use CIE 1931 at all? The only real answer is that old habits die hard. The industry has relied on CIE 1931 since its inception, and change is coming slowly.
Fortunately, CIE 1931’s grip is loosening over time. The ICDM’s new measurement standard should eventually force all remaining stragglers to switch over to the more accurate 1976 standard. Until then, you can familiarize yourself with a decent color space conversion calculator, such as the handy converter we built just for this purpose:
Last Friday Apple released an updated version of one of their hottest products, called simply “the new iPad.” Central to the update is a brand new display featuring significantly more resolution and color saturation. Since the resolution bit has been covered to death by others and we’re interested in color here we thought we’d take a closer look at Apple’s color saturation claims.
Using the new iPad, particularly next to an “iPad 2,” the reds and greens are noticeably better, but the blues in particular are quite striking. It actually makes the blue on the iPad 2 seem more ‘aqua’ than pure blue. The color data bears this out. According to our measurements, Apple has significantly increased the saturation in all three primaries, most notably in blue:
The key color claim that Apple made on stage at the iPad announcement was that the new iPad has 44% more color saturation. What they mean by that of course depends on the context. There are a couple of different color measurement standards that Apple could be gauging the performance of the new iPad against such as CIE 1931 or CIE 1976.
An easy way to think about these standards is a bit like the temperature measures that we are all familiar with, Celsius and Fahrenheit, in that they are different ways communicating the same information. Saying, “it’s 5 degrees warmer today” means something very different to users of each system and its much the same way with color spaces, only we’re talking about measuring how the eye perceives color, not how warm it is outside.
We should also note that when people in the display industry talk about color saturation as a percentage, it is common practice to refer to a color gamut standard within a CIE color space. There are many color gamut standards in use today including: NTSC, sRGB, Adobe RGB 1998, DCI-P3, and rec 709. Each of these standards is a subset of a CIE color space. They are typically used by content creators to ensure the compatibility of their work from device to device. For example, if I create an image in Adobe RGB, I would like to display it on a screen that can show all of the colors in Adobe RGB in order to make sure it accurately reproduces all the colors in my original shot.
Based on our measurements it looks like Apple is referring to the NTSC gamut within a color space. But which color space do they mean?
A 44% improvement within the CIE 1931 color space would give the new iPad the equivalent of the sRGB standard used by HDTV broadcasts, Blu-Ray and much of the web. Given the significance of achieving that standard, some thought Apple must have been trying to say “sRGB” without confusing consumers by describing the meaning of various color standards.
According to our data, this is not the case. The new iPad only manages about 26% more saturation over the iPad 2 when measured against the CIE 1931 NTSC color space. However, the unit we measured showed a 48% increase in saturation when measured in the CIE 1976 color space, so that must be Apples frame of reference.
Measurements and standards aside, the new display looks great. The improvement in color performance will greatly enhance the user experience, and as we discussed yesterday, show’s what Apple is betting on for the functionality of future devices.
In our next post we will explain exactly how Apple achieved this improved color performance and look at ways they can improve the next generation.