Jörgen Eriksson
Kristoffer Holmqvist
Mikael Graffner
Email: publicera@lub.lu.se
+46 (0)46 222 0326
Your most visited
- Sorry, this tool will only work with Javascript available.
Find Publications
Theses, dissertations and research publications (including journal articles, conference abstracts and books) from Lund University are collected in this database. Where possible, the option to download a full text document is available. It is also possible to search for Lund University student theses in the student theses database.
| Title | Optimizing Visual Vocabularies Using Soft Assignment Entropies |
| Author/s | Yubin Kuang, Karl Åström, Lars Kopp, Magnus Oskarsson, Martin Byröd |
| Department/s |
Mathematics (Faculty of Technology)
|
| Full-text | Available as PDF |
| Alternative location (URL) | http://dx.doi.org/10.1007/978-... Restricted Access (Alternative Location) |
| Publication/Series | Lecture Notes in Computer Science |
| Publishing year | 2011 |
| Volume | 6495 |
| Pages | 14 |
| Pages | 255 - 268 |
| Document type | Conference |
| Conference name | 10th Asian Conference on Computer Vision (ACCV 2010) |
| Conference date | 2010-11-08/2010-11-12 |
| Conference location | Queenstown, New Zealand |
| Status | published |
| Quality controlled | yes |
| Language | English |
| Publisher | Springer |
| Abstract English | The state of the art for large database object retrieval in images is based on quantizing descriptors of interest points into visual words. High similarity between matching image representations (as bags of words) is based upon the assumption that matched points in the two images end up in similar words in hard assignment or in similar representations in soft assignment techniques. In this paper we study how ground truth correspondences can be used to generate better visual vocabularies. Matching of image patches can be done e.g. using deformable models or from estimating 3D geometry. For optimization of the vocabulary, we propose minimizing the entropies of soft assignment of points. We base our clustering on hierarchical k-splits. The results from our entropy based clustering are compared with hierarchical k-means. The vocabularies have been tested on real data with decreased entropy and increased true positive rate, as well as better retrieval performance. |
| Subject |
Mathematics and Statistics |
| ISBN/ISSN/Other |
ISSN: 1611-3349 (online) ISSN: 0302-9743 (print) ISBN: 978-3-642-19282-1(online) ISBN: 978-3-642-19281-4 (print) |
| Research group | Mathematical Imaging Group |
Contact
"ReSearch for the Future"
Lund University's "ReSearch for the Future" magazine (Pdf, 10 Mb) presents a range of research from across the University.