Lucas–Kanade optical flow algorithm

E295649

The Lucas–Kanade optical flow algorithm is a widely used computer vision method for estimating the motion of features between consecutive images by assuming locally constant motion and solving a least-squares problem.

All labels observed (5)

How this entity was disambiguated

Statements (49)

Predicate Object
instanceOf computer vision algorithm
feature tracking method
optical flow algorithm
assumption brightness constancy
locally constant motion
small motion between frames
basedOn optical flow constraint equation
category differential optical flow method
commonVariant Lucas–Kanade optical flow algorithm self-linksurface differs
surface form: inverse compositional Lucas–Kanade

pyramidal Lucas–Kanade optical flow
field computer vision
image processing
robotics
video analysis
implementedIn MATLAB Computer Vision Toolbox
OpenCV
scikit-image
influenced modern feature tracking pipelines
real-time visual tracking systems
introducedBy Bruce D. Lucas
Takeo Kanade NERFINISHED
limitation assumes constant motion within window
fails on large motions without pyramids
sensitive to illumination changes
mathematicalTool Gauss–Newton optimization
normal equations
operatesOn consecutive image frames
local image neighborhoods
originalPublicationTitle Lucas–Kanade optical flow algorithm self-linksurface differs
surface form: An iterative image registration technique with an application to stereo vision
originalPublicationVenue Proceedings of Imaging Understanding Workshop
output displacement vector of features
optical flow field at selected points
publicationYear 1981
relatedTo Horn–Schunck optical flow method
Kanade–Lucas–Tomasi feature tracker
surface form: KLT feature tracker

Lucas–Kanade optical flow algorithm self-linksurface differs
surface form: pyramidal Lucas–Kanade method
requires good spatial texture in local window
solves 2D motion vector for image patch
typicalInput grayscale image sequence
typicalUseCase feature tracking in video
image registration
motion estimation
stereo matching refinement
visual odometry
uses least-squares estimation
overdetermined linear system
spatial image gradients
temporal image gradients
windowType fixed-size image patch

How these facts were elicited

Referenced by (6)

Full triples — surface form annotated when it differs from this entity's canonical label.

Takeo Kanade knownFor Lucas–Kanade optical flow algorithm
Lucas–Kanade optical flow algorithm originalPublicationTitle Lucas–Kanade optical flow algorithm self-linksurface differs
this entity surface form: An iterative image registration technique with an application to stereo vision
Lucas–Kanade optical flow algorithm relatedTo Lucas–Kanade optical flow algorithm self-linksurface differs
this entity surface form: pyramidal Lucas–Kanade method
Lucas–Kanade optical flow algorithm commonVariant Lucas–Kanade optical flow algorithm self-linksurface differs
this entity surface form: inverse compositional Lucas–Kanade
Kanade–Lucas–Tomasi feature tracker basedOn Lucas–Kanade optical flow algorithm
this entity surface form: Lucas–Kanade optical flow method
Kanade–Lucas–Tomasi feature tracker relatedTo Lucas–Kanade optical flow algorithm
this entity surface form: pyramidal Lucas–Kanade method