Features for recognition: Viewpoint invariance for non-planar scenes
Most current local feature detectors/descriptors implicitly assume that the scene is (locally) planar, an assumption that is violated at surface discontinuities. We show that this restriction is, at least in theory, unnecessary, as one can construct local features that are viewpoint-invariant for ge...
| Main Authors: | , |
|---|---|
| Format: | Journal article |
| Language: | English |
| Published: |
2005
|
| _version_ | 1826289940227424256 |
|---|---|
| author | Vedaldi, A Soatto, S |
| author_facet | Vedaldi, A Soatto, S |
| author_sort | Vedaldi, A |
| collection | OXFORD |
| description | Most current local feature detectors/descriptors implicitly assume that the scene is (locally) planar, an assumption that is violated at surface discontinuities. We show that this restriction is, at least in theory, unnecessary, as one can construct local features that are viewpoint-invariant for generic non-planar scenes. However, we show that any such feature necessarily sacrifices shape information, in the sense of being non shape-discriminative. Finally, we show that if viewpoint is factored out as part of the matching process, rather than explicitly in the representation, then shape is discriminative indeed. We illustrate our theoretical results empirically by showing that, even for simple scenes, current affine descriptors fail where even a naive 3-D viewpoint invariant succeeds in matching. © 2005 IEEE. |
| first_indexed | 2024-03-07T02:36:34Z |
| format | Journal article |
| id | oxford-uuid:a8fa6ccb-cc55-4858-ae30-7d5b9111854b |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-03-07T02:36:34Z |
| publishDate | 2005 |
| record_format | dspace |
| spelling | oxford-uuid:a8fa6ccb-cc55-4858-ae30-7d5b9111854b2022-03-27T03:05:20ZFeatures for recognition: Viewpoint invariance for non-planar scenesJournal articlehttp://purl.org/coar/resource_type/c_dcae04bcuuid:a8fa6ccb-cc55-4858-ae30-7d5b9111854bEnglishSymplectic Elements at Oxford2005Vedaldi, ASoatto, SMost current local feature detectors/descriptors implicitly assume that the scene is (locally) planar, an assumption that is violated at surface discontinuities. We show that this restriction is, at least in theory, unnecessary, as one can construct local features that are viewpoint-invariant for generic non-planar scenes. However, we show that any such feature necessarily sacrifices shape information, in the sense of being non shape-discriminative. Finally, we show that if viewpoint is factored out as part of the matching process, rather than explicitly in the representation, then shape is discriminative indeed. We illustrate our theoretical results empirically by showing that, even for simple scenes, current affine descriptors fail where even a naive 3-D viewpoint invariant succeeds in matching. © 2005 IEEE. |
| spellingShingle | Vedaldi, A Soatto, S Features for recognition: Viewpoint invariance for non-planar scenes |
| title | Features for recognition: Viewpoint invariance for non-planar scenes |
| title_full | Features for recognition: Viewpoint invariance for non-planar scenes |
| title_fullStr | Features for recognition: Viewpoint invariance for non-planar scenes |
| title_full_unstemmed | Features for recognition: Viewpoint invariance for non-planar scenes |
| title_short | Features for recognition: Viewpoint invariance for non-planar scenes |
| title_sort | features for recognition viewpoint invariance for non planar scenes |
| work_keys_str_mv | AT vedaldia featuresforrecognitionviewpointinvariancefornonplanarscenes AT soattos featuresforrecognitionviewpointinvariancefornonplanarscenes |