Abstract
We present a temporal accumulation scheme which disambiguates different kinds of visual 3D descriptors within one coherent framework. The accumulation consists of a twofold process: First, by means of a Bayesian filtering outliers become eliminated and second, the precision of the extracted information becomes enhanced by means of an unscented Kalman filtering process. It is a particular property of our algorithm to be able to deal with different kinds of visual descriptors by the very same mechanism. We show quantitative and qualitative results.
Originalsprog | Engelsk |
---|---|
Bogserie | Lecture Notes in Computer Science |
Vol/bind | 6688 LNCS |
Sider (fra-til) | 79-90 |
Antal sider | 12 |
ISSN | 0302-9743 |
DOI | |
Status | Udgivet - 2011 |
Begivenhed | 17th Scandinavian Conference on Image Analysis - Ystad, Sverige Varighed: 23 maj 2011 → 27 maj 2011 |
Konference
Konference | 17th Scandinavian Conference on Image Analysis |
---|---|
Land/Område | Sverige |
By | Ystad |
Periode | 23/05/2011 → 27/05/2011 |