Relaxed matching kernels for robust image comparison
The popular bag-of-features representation for object recognition collects signatures of local image patches and discards spatial information. Some have recently attempted to at least partially overcome this limitation, for instance by "spatial pyramids" and "proximity" kernels....
| 主要な著者: | , |
|---|---|
| フォーマット: | Journal article |
| 出版事項: |
2008
|
| _version_ | 1826306564661706752 |
|---|---|
| author | Vedaldi, A Soatto, S |
| author_facet | Vedaldi, A Soatto, S |
| author_sort | Vedaldi, A |
| collection | OXFORD |
| description | The popular bag-of-features representation for object recognition collects signatures of local image patches and discards spatial information. Some have recently attempted to at least partially overcome this limitation, for instance by "spatial pyramids" and "proximity" kernels. We introduce the general formalism of "relaxed matching kernels" (RMKs) that includes such approaches as special cases, allow us to derive useful general properties of these kernels, and to introduce new ones. As an example, we introduce a kernel based on matching graphs of features and one based on matching information-compressed features. We show that all RMKs are competitive and outperform in several cases recently published state-of-the-art results on standard datasets. However, we also show that a proper implementation of a baseline bag-of-features algorithm can be extremely competitive, and outperform the other methods in some cases. ©2008 IEEE. |
| first_indexed | 2024-03-07T06:49:51Z |
| format | Journal article |
| id | oxford-uuid:fc2b1759-b812-4011-87bb-34e9c6058007 |
| institution | University of Oxford |
| last_indexed | 2024-03-07T06:49:51Z |
| publishDate | 2008 |
| record_format | dspace |
| spelling | oxford-uuid:fc2b1759-b812-4011-87bb-34e9c60580072022-03-27T13:18:50ZRelaxed matching kernels for robust image comparisonJournal articlehttp://purl.org/coar/resource_type/c_dcae04bcuuid:fc2b1759-b812-4011-87bb-34e9c6058007Symplectic Elements at Oxford2008Vedaldi, ASoatto, SThe popular bag-of-features representation for object recognition collects signatures of local image patches and discards spatial information. Some have recently attempted to at least partially overcome this limitation, for instance by "spatial pyramids" and "proximity" kernels. We introduce the general formalism of "relaxed matching kernels" (RMKs) that includes such approaches as special cases, allow us to derive useful general properties of these kernels, and to introduce new ones. As an example, we introduce a kernel based on matching graphs of features and one based on matching information-compressed features. We show that all RMKs are competitive and outperform in several cases recently published state-of-the-art results on standard datasets. However, we also show that a proper implementation of a baseline bag-of-features algorithm can be extremely competitive, and outperform the other methods in some cases. ©2008 IEEE. |
| spellingShingle | Vedaldi, A Soatto, S Relaxed matching kernels for robust image comparison |
| title | Relaxed matching kernels for robust image comparison |
| title_full | Relaxed matching kernels for robust image comparison |
| title_fullStr | Relaxed matching kernels for robust image comparison |
| title_full_unstemmed | Relaxed matching kernels for robust image comparison |
| title_short | Relaxed matching kernels for robust image comparison |
| title_sort | relaxed matching kernels for robust image comparison |
| work_keys_str_mv | AT vedaldia relaxedmatchingkernelsforrobustimagecomparison AT soattos relaxedmatchingkernelsforrobustimagecomparison |