Candice Brown Elliott is founder and chief technology officer of Clairvoyante Inc. (Cupertino, Calif.), which develops and licenses unique subpixel architectures and associated algorithms to reduce the power, increase the brightness and lower the cost of manufacturing flat-panel displays. Its PenTile Matrix harnesses Elliott's knowledge of the human visual system to trick the eye into perceiving a display as twice as bright for the same power (or just as bright at half the power). By reducing by one-third to one-half the number of pixels needed for a given resolution, the company's PenTile subpixel rendering also lowers costs. PenTile Matrix technology has been licensed by 10 of the world's leading flat-panel makers, including Samsung Electronics Co. Ltd. and AU Optronics Corp.
EE Times' R. Colin Johnson caught up with Elliott at a recent display conference, where she revealed the specific aspects of the human visual system that Clairvoyante harnessed.
EE Times: What is it about the human visual system that enables you to persuade the eye it is perceiving an LCD screen that's twice as bright as usual, when in fact you are using one-third fewer pixels than standard displays?
Candice Brown Elliott: We are increasing information efficiency over conventional display architectures. The eye detects and encodes only portions of the visual information coming into it, but the brain is able to reconstruct the whole image. We discovered how the brain does this by studying visual illusions they show you what the brain is doing to "fill in the blanks." We designed the PenTile subpixel technology so that the brain fills in the blanks through using our subpixels, by doing the same thing it does normally to fill in the blanks between retinal cells in the eye.
EET: How did you determine just what the eye-brain system normally does to fill in the blanks?
Elliott: Well, I have degrees in both psychology and physics, and I have been working on this problem since the early 1990s, when I worked at Planar Systems. For instance, in 1993 I showed Planar how it could use subpixel rendering to reduce the number of blue subpixels it needed, and tripled the brightness of the electroluminescent full-color panel as a result.
EET: What is the essence of your emphasis on physiology that the eye is smarter than a dumb video camera?
Elliott: The biggest difference between the eye and a video camera is that the eye only processes information that changes either spatially or temporally technically, by using a difference-of-Gaussians, or DOG, wavelet filter, which only responds to the second derivative of an image. The eye is actually a part of the brain, and its DOG-filter architecture only sends changes to update the brain's mental images whenever they are sensed by the eye. For instance, if the eye worked like a video camera, there would be permanent shadows where blood vessels crisscross the retina. But the brain removes these shadows by only sending information about changes. And of course, the blood vessel shadows never change, so their shadows never get transmitted that's a natural design efficiency.
Another thing we do is study optical illusions, where the brain fills in the blanks wrongly. The eye sends only the second derivative of the image. The brain then integrates to get the second integral but it must "guess" what the most likely integration constants were, since that information is missing. It's the wrong guesses that tell us how the system works. That's how we learned to create the impression of a display that appears to be much sharper, higher contrast and higher resolution than [one built] without our technology. By contrast, if one were to naively try performing subpixel rendering as a form of anti-aliasing, the result would be a blurry, fuzzy, low-contrast image.