Motion Estimation

The motion estimation techniques fall into two categories:

1. Feature-Based Methods

  • Extract visual features (corners, textured areas) and track them over multiple frames

  • Sparse motion fields, but more robust tracking

  • Suitable when image motion is large (10s of pixels)

    We have seen much of this method in Feature Recognition.

2. Direct, Dense Methods

  • Directly recover image motion at each pixel from spatio-temporal image brightness variantions
  • Dense motion fields, but sensitive to appearance variations.
  • Suitable for video and when image motion is small.

2.1 Optic Flow

Optic flow is the apparent motion of objects or surfaces.

image-20210202175443444

2.1.1 Problem Definition

How to estimate pixel motion from image to

  • Given a pixel in , look for nearby pixels of the same color in

image-20210202175542379

2.1.2 Assumptions

  • Color Constancy: a point in looks the same in . For grayscale images, this is brightness constancy.
  • Small Motion: points do not move very far

Taylor series expansion of : Combining these two constraints together:

2.2.3 Brightness Constancy Constraint Equation

2 unknowns but 1 equation!

What does this constraint mean?

  • The component of the flow in the gradient direction is determined
  • The component of the flow parallel to an edge is unknown

Aperture problem

image-20210202180907796

2.2 Smooth Optical Flow

How much we violate motion brightness equation?

2.2.1 Formulate Error in Optical Flow Constraint

  • Global Error Function

  • Smoothness constraint: Motion field tends to vary smoothly over the image. This punishes large changes to over the image.

Combining both constraint, find at each image point that minimizes:

2.3 Dense Flow Summary

  • Impose a constraint on the flow field in general to make the problem solvable
  • Strength: Allow you to bias your solution with a prior
  • But there are better ways to increase the number of equations.

results matching ""

    No results matching ""