Computing inverse optical flow
•We propose four efficient algorithms for computing inverse optical flows.•The accuracy of the algorithms is very high with low running times.•Optical flow discontinuities are preserved in the inverse registration maps.•Stereoscopic and street-lamp occlusions are handled automatically in the process...
Saved in:
Published in | Pattern recognition letters Vol. 52; pp. 32 - 39 |
---|---|
Main Authors | , , |
Format | Journal Article |
Language | English |
Published |
Elsevier B.V
15.01.2015
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | •We propose four efficient algorithms for computing inverse optical flows.•The accuracy of the algorithms is very high with low running times.•Optical flow discontinuities are preserved in the inverse registration maps.•Stereoscopic and street-lamp occlusions are handled automatically in the process.•Disocclusions are filled in a post-processing step using three strategies.
We propose four algorithms for computing the inverse optical flow between two images. We assume that the forward optical flow has already been obtained. The forward and backward flows can be related through a warping formula, which allows us to propose very efficient algorithms. These methods provide high accuracy with low memory requirements and low running times. Additionally, when objects move in a sequence, some regions may appear or disappear. Finding inverse flows in these situations is difficult and, in some cases, it is not possible to obtain a correct solution. Our algorithms deal with occlusions easily and reliably, and disocclusions are treated in a post-processing step. We analyze three approaches for filling disocclusions. In the experimental results, we use standard synthetic sequences to study the performance of our methods. The experiments show that our algorithms clearly outperform current state-of-the-art methods. |
---|---|
ISSN: | 0167-8655 1872-7344 |
DOI: | 10.1016/j.patrec.2014.09.009 |