Toward reliable optical flow with subpixel accuracy

Kurt A. Steinkraus

doi:10.1117/12.457437

18 February 2002 Toward reliable optical flow with subpixel accuracy

Kurt A. Steinkraus

Proceedings Volume 4573, Mobile Robots XVI; (2002) https://doi.org/10.1117/12.457437
Event: Intelligent Systems and Advanced Manufacturing, 2001, Boston, MA, United States

Abstract

When computing optical flow using region-based matching, several problems arise. One is that the optical flow cannot be reliably computed at most locations; particularly troublesome are those with little or no texture. Some confidence measures exist for optical flow, but they either succumb to the aperture problem, or they require that the optical flow be computed, a potentially costly computation with results discarded if unreliable. Another problem is that the normal patch dissimilarity measures do not take image sampling into account, causing the optical flow not to be found in high-contrast areas. I first give an efficient algorithm for determining image locations likely to have reliably correct optical flow. I then derive a patch dissimilarity measure insensitive to image sampling, by extending the work of Birchfield and Tomasi to the computation of two-dimensional optical flow. Finally, I show how to adapt sub-pixel optical flow estimation by quadratic fitting to this new patch dissimilarity measure.

Citation Download Citation

Kurt A. Steinkraus "Toward reliable optical flow with subpixel accuracy", Proc. SPIE 4573, Mobile Robots XVI, (18 February 2002); https://doi.org/10.1117/12.457437

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available