How Does Faces & the Local Binary Pattern - Computerphile Work?

Faces and the Local Binary Pattern work by turning faces into numbers that a computer can understand easily.

Imagine you're looking at a face, like your friend’s face when they’re smiling. A computer doesn’t see a smile, it sees pixels, which are tiny colored dots on a screen. Each pixel has a number for how light or dark it is. Now, think of these pixels as little blocks in a puzzle.

How It Works with Numbers

The Local Binary Pattern (or LBP) takes a small group of pixels, like the 8 pixels around one central pixel, and compares them to that center pixel. If the surrounding pixels are lighter, it gives them a number, and if they’re darker, it gives them another number.

It’s kind of like comparing your friend's face in different lights: sometimes their cheeks look lighter than their nose, sometimes darker. The computer uses these numbers to create a pattern, a fingerprint for the face.

From Patterns to Recognition

Once all the patterns are made, the computer can compare them to other faces it’s seen before. If two faces have similar number patterns, the computer thinks they're the same person, just like how you know your friend by their smile even if they’re wearing a hat or glasses!

Take the quiz →

Examples

  1. A computer uses a local binary pattern to recognize a face by comparing light and dark areas on the skin.
  2. Imagine looking at a face like a puzzle, each piece helps identify who it is.
  3. The system checks small sections of a face, similar to how you might describe someone's eyes or nose.

Ask a question

See also

Discussion

Recent activity