Parallel optical flow using local voting

James J. Little, Heinrich Bulthoff, Tomaso Poggio

Research output: Chapter in Book/Report/Conference proceedingConference contribution

22 Citations (Scopus)

Abstract

A parallel algorithm is described for computing optical flow from short-range motion. Regularizing optical flow computation leads to a formulation which minimizes matching error and, at the same time, maximizes smoothness of the optical flow. An approximation to the full regularization computation is developed in which corresponding points are found by comparing local patches of the images. Selection among competing matches is performed using a winner-take-all scheme. The algorithm accommodates many different image transformations uniformly, with similar results, from brightness to edges. The optical flow computed from different image transformations, such as edge detection and direct brightness computation, can be simply combined. The algorithm is easily implemented using local operations on a fine-grained computer, and has been implemented on a Connection Machine. Experiments with natural images show that the scheme is effective and robust against noise. The algorithm leads to dense optical flow fields; in addition, information from matching facilitates segmentation.

Original languageEnglish
Title of host publicationUnknown Host Publication Title
Place of PublicationNew York, NY, USA
PublisherPubl by IEEE
Pages454-459
Number of pages6
ISBN (Print)0818608838
Publication statusPublished - 1988 Dec 1
Externally publishedYes

Fingerprint

Optical flows
Luminance
Edge detection
Parallel algorithms
Flow fields
Experiments

ASJC Scopus subject areas

  • Engineering(all)

Cite this

Little, J. J., Bulthoff, H., & Poggio, T. (1988). Parallel optical flow using local voting. In Unknown Host Publication Title (pp. 454-459). New York, NY, USA: Publ by IEEE.

Parallel optical flow using local voting. / Little, James J.; Bulthoff, Heinrich; Poggio, Tomaso.

Unknown Host Publication Title. New York, NY, USA : Publ by IEEE, 1988. p. 454-459.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Little, JJ, Bulthoff, H & Poggio, T 1988, Parallel optical flow using local voting. in Unknown Host Publication Title. Publ by IEEE, New York, NY, USA, pp. 454-459.
Little JJ, Bulthoff H, Poggio T. Parallel optical flow using local voting. In Unknown Host Publication Title. New York, NY, USA: Publ by IEEE. 1988. p. 454-459
Little, James J. ; Bulthoff, Heinrich ; Poggio, Tomaso. / Parallel optical flow using local voting. Unknown Host Publication Title. New York, NY, USA : Publ by IEEE, 1988. pp. 454-459
@inproceedings{592a07de40564505938e2a8983533baa,
title = "Parallel optical flow using local voting",
abstract = "A parallel algorithm is described for computing optical flow from short-range motion. Regularizing optical flow computation leads to a formulation which minimizes matching error and, at the same time, maximizes smoothness of the optical flow. An approximation to the full regularization computation is developed in which corresponding points are found by comparing local patches of the images. Selection among competing matches is performed using a winner-take-all scheme. The algorithm accommodates many different image transformations uniformly, with similar results, from brightness to edges. The optical flow computed from different image transformations, such as edge detection and direct brightness computation, can be simply combined. The algorithm is easily implemented using local operations on a fine-grained computer, and has been implemented on a Connection Machine. Experiments with natural images show that the scheme is effective and robust against noise. The algorithm leads to dense optical flow fields; in addition, information from matching facilitates segmentation.",
author = "Little, {James J.} and Heinrich Bulthoff and Tomaso Poggio",
year = "1988",
month = "12",
day = "1",
language = "English",
isbn = "0818608838",
pages = "454--459",
booktitle = "Unknown Host Publication Title",
publisher = "Publ by IEEE",

}

TY - GEN

T1 - Parallel optical flow using local voting

AU - Little, James J.

AU - Bulthoff, Heinrich

AU - Poggio, Tomaso

PY - 1988/12/1

Y1 - 1988/12/1

N2 - A parallel algorithm is described for computing optical flow from short-range motion. Regularizing optical flow computation leads to a formulation which minimizes matching error and, at the same time, maximizes smoothness of the optical flow. An approximation to the full regularization computation is developed in which corresponding points are found by comparing local patches of the images. Selection among competing matches is performed using a winner-take-all scheme. The algorithm accommodates many different image transformations uniformly, with similar results, from brightness to edges. The optical flow computed from different image transformations, such as edge detection and direct brightness computation, can be simply combined. The algorithm is easily implemented using local operations on a fine-grained computer, and has been implemented on a Connection Machine. Experiments with natural images show that the scheme is effective and robust against noise. The algorithm leads to dense optical flow fields; in addition, information from matching facilitates segmentation.

AB - A parallel algorithm is described for computing optical flow from short-range motion. Regularizing optical flow computation leads to a formulation which minimizes matching error and, at the same time, maximizes smoothness of the optical flow. An approximation to the full regularization computation is developed in which corresponding points are found by comparing local patches of the images. Selection among competing matches is performed using a winner-take-all scheme. The algorithm accommodates many different image transformations uniformly, with similar results, from brightness to edges. The optical flow computed from different image transformations, such as edge detection and direct brightness computation, can be simply combined. The algorithm is easily implemented using local operations on a fine-grained computer, and has been implemented on a Connection Machine. Experiments with natural images show that the scheme is effective and robust against noise. The algorithm leads to dense optical flow fields; in addition, information from matching facilitates segmentation.

UR - http://www.scopus.com/inward/record.url?scp=0024174150&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0024174150&partnerID=8YFLogxK

M3 - Conference contribution

SN - 0818608838

SP - 454

EP - 459

BT - Unknown Host Publication Title

PB - Publ by IEEE

CY - New York, NY, USA

ER -