So what will the manufacturers do?

Video and live-view are logical progressions, but once you have that feature, it's hard to upsell to the next model unless the existing one is flawed. We'll see some migration away from the current modal control system and towards manual controls for basics like aperture and shutter-speed. But other than the general improvements of one camera over the other, there is no longer a compelling reason for the consumer to upgrade. Features and size reduction will remain purchase incentives for some time, though. In two to three years we'll see the "post-modern" or "neo-classical" control interface make a strong play. Does this mean we've hit the wall as far as image-quality is concerned?  Absolutely not!

There are two areas of image-quality improvement:

1. Dynamic Range

2. Wider Gamut

It is important to realize that these are related issues and are a direct result of the physical design of the sensor. It's physics!

With printers, we've gone from the four primary print colors of cyan, magenta, yellow and black to colors so numerous that the ink replacement costs are comparable to buying a computer, DSLR or expensive lens. The Epson Stylus Pro 7900/9900 takes nearly a dozen ink cartridges! It is only a matter of time before the chip manufacturers follow suit and a whole new battlefield is created.

Standard Bayer RGB color array

The commonly used Bayer filter array uses a tricolor sensor with green, red and blue sensitive detectors. These individual pixels can be mixed with their nearest neighbors to create the full RGB gamut of colors. In essence, each detector is an analog device producing a microvoltage (or microcurrent, depending on chip design) based on the amount of photons striking the detector. This microvoltage or microcurrent is then measured and assigned a numerical value. The numerical value is recorded in the image file for later conversion to a color and brightness value ultimately used by an output device which converts these numerical values into a pulsed laser beam (printer), LCD display, or inkjet printer squirting equivalent amounts of pigments or dyes on paper.

The sooner the detector's output is converted to numbers, the greater the potential SNR (signal to noise ratio). This is why we are starting to see more sensor chips with on-board A-D (Analog to Digital) Converters.

We have seen several variations of the color array used in sensors, with yellow, cyan and magenta specified in some. The red, green and blue array had distinct mixing advantages and for the most part proved to be the best three colors to use to record visible light when only three colors are to be used.

Several years ago, Sony introduced the DSC-F828 which replaced half of the green detectors with emerald detectors. The end result was a distinct advantage in extending the gamut, but had the unintended consequence of strange artifacts when certain colors landed at the red-emerald transition. The concept was sound, but the execution was flawed.

This fourth-color approach also has a film parallel as Fujifilm incorporated a fourth color layer in many color print films. Fujifilm Reala is considered by many to be among the best films ever produced because of it's ability to capture more accurate and pleasing colors in mixed and artificial lighting.

With greater pixel density of newer sensors, we can break from the RGB model and start incorporating additional detectors of other colors. Kodak has sensors with a fourth non-color detector. Fuji uses six detectors, three full-size, three smaller-sized for the matrix.

One highly likely sensor design will incorporate a six-color array like this:

Six Color Array

The six color array is modeled on the standard three-color array but allows additional mixing of colors which would greatly increase the potential gamut of the image. During conversion, we will still end up with a standard RGB image until our editors support RGBCYK files which is unlikely for a while. Even so, it doesn't matter how the image is stored, as it is conversion from the physical world into the electronic world and vice-versa that is the problem.

The advantages of the six color array over a three color array is specifically in the "derived colors", such as yellow and magenta. The ability to capture yellows and more nuanced greens is an area where three color detectors struggle. Having red, magenta and yellow detectors all available means far greater gamut and tonal separation in the reds. Human skintones should be greatly improved too with fewer green and magenta shifts.

The human eye also produces derived colors. The human eye is usually unable to directly see the color "Red" as the cones sensitive to the longer wavelengths peak at yellow-orange. The visual cortex derives the color red by the absence of the color green. In otherwords, it is doing a "mix-minus" approach to color determination. When the green sensitive cones see light and the yellow-orange cones see light, then the visual cortex determines that the color is not red, but if the yellow-orange cones see light and the green cones do not then the color is red. To a very limited extent, the Bayer algorithm is doing the same thing, but only through the additive process and doesn't directly apply a "mix-minus" calculation. If it did so, then we would be able to get a wider gamut during capture.

To provide for a six color array, as shown above, only requires a doubling of pixel count and with in-camera processing, combining pixels is possible for a standard RGB file output

Another critical advantage to the six color array is improved white-balance and color response in artificial lighting. Most artificial lighting today has a very poor CI (color index) and RGB-based sensors do not properly assign the actual colors to the appropriate RGB values. By having detectors for yellow, cyan and magenta in additon to red, green and blue, the image-processing is able to see and adjust for these off-colored light sources and map the resulting image to the visible spectrum matching what the human eye sees.

The addition of more color detectors to the sensor is highly likely and is just a matter of time before they appear on the market. This has the potential for being a "game-changer" in the photography industry and over a period of a year or two will substantially obsolete the existing cameras as the advantages to the new sensor design is very obvious.

Non-Visible Color Detectors

Kodak is producing sensors now with the equivalent to the human eye's rods. These detectors do not see colors, but only the amount of light striking the sensor. Combined with the color detectors, this sensor has the potential of much greater light sensitivity and far better color fidelity in high and low values. In a normal tri-color sensor, colors are poorly captured in the first and last stops of dynamic range. The Kodak sensor should have far better capture of the pastels.

One development we may not see in general-purpose cameras is the IR and UV sensitive detectors. Sensors are typically covered with filter blocking the majority of the non visible-spectrum colors. When this filtering is inadequate, we experience maroon blacks, mysterious green shadows and other oddities. However, as a result of aggressive filtering we lose the ability to capture certain pigments in the natural world. An African Violet, for instance, is extremely difficult to photograph as the majority of the color that the human eye sees is actually outside of the range of the visible spectrum. The African Violet turns blue. The filters are present because today's RGB detectors have the nasty habit of seeing equally well the out-of-bounds colors and have no ability to determine the difference between IR and UV light. If a sensor has IR and UV detectors which are not sensitive to the visible spectrum, the processing algorithm is able to do a "mix-minus" which is identical to the way the human visual cortex processes the color red. This style of chip design would have tremendous value in astronomy, science, medical and forensic work. It wouldn't necessarily be seen in a $1500 DSLR, but would be likely in a special-purpose model of the high-end Nikon and Canon models.