| Pictured top, an octopus retina. (Bottom) The silicon chip pictured above mimics the structure and function of the octopus retina. |
And, like an octopus, the chip can't distinguish between oriented, horizontally mirrored images, such as the letter X. Future chips will be able to see polarized light, particularly underwater. This turns out to be especially important for the octopus.
"Different animals have different visual structures with different properties for how they see the world," says Titus. "If we want to develop a chip that can see far distances, we should mimic the visual system of an eagle. Likewise, if we want to develop a chip for seeing primarily in the desert, we should mimic the eyes of zebras or lions."
Titus' goal is to build a complete artificial vision system able to perform multiple visual functions. His chips use analog circuitry, which need far less operational power than digital circuitry. Because the chips use little power they are suitable for autonomous robots or other exploration devices that have long run times.
Also in the works is a silicon retina chip that mimics edge detection, a form of data compression performed by biological eyes. Here, the chip uses visual information from the edges of an image to identify objects rather than from the whole picture.
"Edge detection is a basic-level process of what the retina does," explains Titus. "The retina reduces the amount of visual information we take in and extracts the most useful information. The challenge is understanding how the brain uses this information to process and reconstitute the image. Then (the question is whether) we can implement this process on a chip."