Motion estimation

Why estimate visual motion?

Today’s lecture

Readings

Patch-based motion estimation

Classes of Techniques

Patch matching (revisited)

The Brightness Constraint

The Brightness Constraint

Gradient Constraint (or the Optical Flow Constraint)

Patch Translation [Lucas-Kanade]

Local Patch Analysis

The Aperture Problem

SSD Surface – Textured area

SSD Surface -- Edge

SSD – homogeneous area

Iterative Refinement

Optical Flow: Iterative Estimation

Optical Flow: Iterative Estimation

Optical Flow: Iterative Estimation

Optical Flow: Iterative Estimation

Optical Flow: Iterative Estimation

Optical Flow: Aliasing

Slide 24

Slide 25

Parametric motion estimation

Global (parametric) motion models

Motion models

Example:  Affine Motion

Other 2D Motion Models

3D Motion Models

Patch matching (revisited)

Correlation and SSD

Discrete Search vs. Gradient Based

Shi-Tomasi feature tracker

Tracking results

Tracking - dissimilarity

Tracking results

Correlation Window Size

Robust Estimation

Robust Estimation

Robust Estimation

Robust Estimation

Image Morphing

Image Warping – non-parametric

Image Warping – non-parametric

Image Morphing

Image Morphing

Warp specification

Warp specification

Warp specification

Warp specification

Final Morph Result

Layered Scene Representations

Motion representations

Block-based motion prediction

Layered motion

Layered motion

Layers for video summarization

Background modeling (MPEG-4)

What are layers?

How do we form them?

How do we estimate the layers?

Layer synthesis

Results

Bibliography

Bibliography

Bibliography

Bibliography

Bibliography

Bibliography