   Next: Optical Flow Constraint Equation Up: Optical Flow Previous: Optical Flow

## Introduction

If we take a series of images in time, and there are moving objects in the scene, or perhaps the camera is itself on some moving vehicle, useful information about what the image contains can be obtained by analysing and understanding the difference between images caused by the motion.

For example, given an image of a moving car, deciding which pixels in the image represent motion can help to decide which pixels belong to the car, and which to the static background.

Studying the motion in detail, we can answer such questions as

• How many moving objects there are?
• Which directions they are moving in?
• Whether they are undergoing linear or rotational motion?
• How fast they are moving?

From an image sequence, we calculate a new function called the optical flow. For every pixel,

• A velocity vector is found which says

• how quickly whatever is in that pixel is moving across the image,
• direction it is moving. David Marshall 1994-1997