THE UNIVERSITY OF WESTERN AUSTRALIA

📌 Preview

• Optical Flow and Motion Field

• Tracking

• Applications

• Optical Flow Constraints

• Lucas Kanade Tracking Algorithm

• Mean Shift Algorithms

🌠 OPTICAL FLOW AND MOTION FIELD

• Motion Field:

  • Projection of 3D relative velocity vectors onto the image plane.

  Optic Flow:

  • Observed 2D displacement of pixels (brightness patterns) in the image.

Most of the time, motion field is what we want to find. But optic flow is what we actually measure from the image.

$Optic Flow \neq Motion Field$

💈 Barber's pole illusion

Aperture Problem

• We can only measure the component of optic flow in the direction of intensity gradient.

• We can not measure the component that is tangent to the intensity gradient.

The Optical Flow Constraints

Techniques for Computing Optical Flow

Differential Technique: Lucas Kanade Motion

Solving The Aperture Problem

Lucas Canade Optical Flow

Implications

Farneback's Two Frame Optical Flow

TRACKING

• Using optical flow, we can track pixels or corners over multiple frames

• Sometimes we don't want to track every pixel or every corner

• We mat want to track a specific object such as

  • A tennis ball

  • Pedestrians

  • Cars

  • Or simple blobs

Lucas-Kanade Tracking

Tracking with Mean-Shift

Mean Shift and Colour Models

Mean Shift On Weight Images

CamShift (Continuously Adaptive Mean Shift)