• CLA Discoveries
• About CLA
• Faculty, Teaching & Research
• Students
• In the Community
• Departments & Majors
• Alumni & Friends
• News & Events
  • CLA Events Calendar
  • For the Media
  • Reach Magazine
  • RSS Feeds
  • Publications & Social Media
• Giving to CLA

Home > News & Events > CLA Today Archive > Summer-Fall 2001 : There's More to Seeing

There's More to Seeing Than Meets the Eye

by Jessica Breed & Eugenia Smith

Communicating with images is as effortless as clicking a mouse; yet scientists are just beginning to understand what "seeing" actually is and how the human brain registers and interprets the visible world.

“It wasn't until the 1960s and 70s, when computer scientists were trying to give robots the ability to see, that we discovered that useful information about the world could not be obtained by simple measurements of image intensities,” says Dan Kersten, professor of psychology and renowned vision researcher.

They discovered through computer-simulations that vision required much more than passive photographic scanning. Seeing was more like a cognitive process, as sophisticated as speaking or reading. Robots could register shapes, lines, and variegations of light and dark, but is that really seeing?

Vision, says Kersten, is a way of "reading" a stream of information from the physical environment. Most of us easily and instantly can "read" a coworker's face or distinguish a parked car from a moving one. But this process of recognition,while instantaneous, requires very complex brain activity, no less than does the subsequent action of greeting or getting out of the way. Up to half of the brain may be used for vision, says Kersten, adding, "If we understand vision and its underlying principles, we will go a long way towards understanding how the brain works.”

With his colleagues in vision research, Kersten is seeking answers to questions that philosophers have asked for centuries: What is seeing? How are seeing and knowing related? If we both look at an object, do we see the same thing? How do we share the experience of seeing and communicate what we see?

“Early vision" researchers study the range of retinal perception—how small a change in brightness or movement we are able to detect. Those studying "high level vision" look at how we make decisions and judgments about visual information—for example, how we "read" environmental cues such as facial expressions or navigate through rush-hour traffic.

“Middle vision" researchers such as Kersten connect the objective sensory data to larger conceptual processes like object and pattern recognition. Using computer animation, MRI scanners, and robotic simulations as well as human subjects, Kersten studies what the brain is doing while the eye is receiving images—how the brain translates seeing into recognition and meaning. [To learn more about Kersten's research visit his website.]

Kersten and his colleagues have discovered that the brain can register and process significant amounts of information that escape our awareness—and so, in a sense, we don't really see what our brains see. The brain may take in data from across an entire landscape, while the conscious mind may register awareness of only a portion of the scene—say, a grove of trees or a flock of birds.

The human brain makes unconscious calculations about such data as shape, color, shadow, and dimensionality much as it does for other cognitive processes like constructing sentences. While we may be unable to articulate the rules of grammar or the physics of vision, we may speak and see with great proficiency. Our brains know the rules, understand the physics.

But we also can trick our brains. We can transform a two-dimensional figure into a three-dimensional image simply by reconceptualizing it.

An avid painter, Kersten says that to understand how we create art is to unravel one of the great mysteries of cognitive science: How do we represent on paper or canvas our knowledge of an object? "We have a lot of implicit knowledge, and it's a great challenge to make it explicit,” he says. Although we all have the ability to recognize faces, translating the "information" we have—nose, mouth, eyebrows—into a recognizable portrait is difficult at best, in part because we don't really know just what makes a face recognizable—or, for that matter, what allows us to see a face from entirely different angles and know it's the same face. "At best,” says Kersten, "we can tell when the painted representation 'looks right.'”

Indeed, there is "a sense in which painting reverses the process of perception,” says Kersten. Rather than beginning with a retinal image and constructing a three-dimensional scene in our mind's eye, we begin with the memory or "conception" of the scene and render it as a two-dimensional image on canvas.

Kersten's studies in vision have important long-term implications for visual communication as well as for medicine, robotics, engineering, and computer graphics—the field that makes possible much of Kersten's work. But a full understanding of how the brain processes visual information can take years. Recently while studying a 3D object computer-generated by another researcher, Kersten remarked to students in the lab, "I’m sort of discouraged. Ten years ago I thought we would have solved this problem by now.”

“You may be discouraged,” one student responded, "but we're glad, because it gives us a challenging problem to work on." "That's what makes this work exciting,” says Kersten, smiling.

“There will always be challenges.”