Evan Lenz

This morning’s reading is for my class. It’s from a book called Visual Intelligence: How We Create What We See. One of the things it talks about is how our brain constructs lines from the light that enters our eyes:

And that is what you do each time you see a line. You construct it from receptor responses. This is not so easy as you might think. Just ask researchers in computer vision. They have worked on “edge detection” or “line finding,” an apparently simple problem, for decades. They have made progress, but their current solutions require much computation–on the order of tens of millions of multiplications and additions just to construct lines in one small image. Even so, their performance is no match to yours.

Constructing a line from what clearly looks like a line seems like a pretty easy problem. (Of course it would.) But after recently working with the “bitmap trace” feature in products like Adobe Illustrator and Macromedia Flash, which tries to recognize lines in order to create vectors from pixels, I can vouch for the fact that computer scientists still have a long way to go when it comes to the “line finding” problem. “I see the line so clearly, why can’t the computer?!”