Title

Planar interpolation for optical flow estimation

Abstract

A new technique for optical flow interpolation to be used in the problem of image registration is presented. It is based on the assumption that the fractional portion of each optical flow vector lies on one of eight planes surrounding its closest pixel. These planes represent the local variations in gray level and are estimated from the coordinates and the gray levels of the pixels in a neighborhood of 3 × 3. First, a plane is selected according to the direction of the flow vector under consideration. Then, the gray level is estimated as the third coordinate in a plane equation. We have implemented this method, embedded it in a hierarchical spline-based image registration algorithm and used public domain sequences of images for tests. The first results show that the planar interpolation performs better than a bi-linear interpolation in local flow estimation in terms of image error. Also, this technique is fast and easy to implement.

Publication Date

12-1-1998

Publication Title

Proceedings of the Joint Conference on Information Sciences

Volume

4

Number of Pages

370-373

Document Type

Article; Proceedings Paper

Personal Identifier

scopus

Socpus ID

1642391762 (Scopus)

Source API URL

https://api.elsevier.com/content/abstract/scopus_id/1642391762

This document is currently not available here.

Share

COinS