File Download

There are no files associated with this item.

  Links for fulltext
     (May Require Subscription)
Supplementary

Article: Representing and recognizing objects with massive local image patches

TitleRepresenting and recognizing objects with massive local image patches
Authors
KeywordsObject detection
Object recognition
Generative learning
Issue Date2012
Citation
Pattern Recognition, 2012, v. 45, n. 1, p. 231-240 How to Cite?
AbstractNatural image patches are fundamental elements for visual pattern modeling and recognition. By studying the intrinsic manifold structures in the space of image patches, this paper proposes an approach for representing and recognizing objects with a massive number of local image patches (e.g. 17×17 pixels). Given a large collection (>104) of proto image patches extracted from objects, we map them into two types of manifolds with different metrics: explicit manifolds of low dimensions for structural primitives, and implicit manifolds of high dimensions for stochastic textures. We define these manifolds grown from patches as the ε-balls, where ε corresponds to the perception residual or fluctuation. Using these ε-balls as features, we present a novel generative learning algorithm by the information projection principle. This algorithm greedily stepwise pursues the object models by selecting sparse and independent ε-balls (say 103 for each category). During the detection and classification phase, only a small number (say 20) of features are activated by a fast KD-tree indexing technique. The proposed method owns two characters. (1) Automatically generating features (ε-balls) from local image patches rather than designing marginal feature carefully and category-specifically. (2) Unlike the weak classifiers in the boosting models, these selected ε-ball features are used to explain object in a generative way and are mutually independent. The advantage and performance of our approach is evaluated on several challenging datasets with the task of localizing objects against appearance variance, occlusion and background clutter. © 2011 Elsevier Ltd. All rights reserved.
Persistent Identifierhttp://hdl.handle.net/10722/273508
ISSN
2023 Impact Factor: 7.5
2023 SCImago Journal Rankings: 2.732
ISI Accession Number ID

 

DC FieldValueLanguage
dc.contributor.authorLin, Liang-
dc.contributor.authorLuo, Ping-
dc.contributor.authorChen, Xiaowu-
dc.contributor.authorZeng, Kun-
dc.date.accessioned2019-08-12T09:55:47Z-
dc.date.available2019-08-12T09:55:47Z-
dc.date.issued2012-
dc.identifier.citationPattern Recognition, 2012, v. 45, n. 1, p. 231-240-
dc.identifier.issn0031-3203-
dc.identifier.urihttp://hdl.handle.net/10722/273508-
dc.description.abstractNatural image patches are fundamental elements for visual pattern modeling and recognition. By studying the intrinsic manifold structures in the space of image patches, this paper proposes an approach for representing and recognizing objects with a massive number of local image patches (e.g. 17×17 pixels). Given a large collection (>104) of proto image patches extracted from objects, we map them into two types of manifolds with different metrics: explicit manifolds of low dimensions for structural primitives, and implicit manifolds of high dimensions for stochastic textures. We define these manifolds grown from patches as the ε-balls, where ε corresponds to the perception residual or fluctuation. Using these ε-balls as features, we present a novel generative learning algorithm by the information projection principle. This algorithm greedily stepwise pursues the object models by selecting sparse and independent ε-balls (say 103 for each category). During the detection and classification phase, only a small number (say 20) of features are activated by a fast KD-tree indexing technique. The proposed method owns two characters. (1) Automatically generating features (ε-balls) from local image patches rather than designing marginal feature carefully and category-specifically. (2) Unlike the weak classifiers in the boosting models, these selected ε-ball features are used to explain object in a generative way and are mutually independent. The advantage and performance of our approach is evaluated on several challenging datasets with the task of localizing objects against appearance variance, occlusion and background clutter. © 2011 Elsevier Ltd. All rights reserved.-
dc.languageeng-
dc.relation.ispartofPattern Recognition-
dc.subjectObject detection-
dc.subjectObject recognition-
dc.subjectGenerative learning-
dc.titleRepresenting and recognizing objects with massive local image patches-
dc.typeArticle-
dc.description.naturelink_to_subscribed_fulltext-
dc.identifier.doi10.1016/j.patcog.2011.06.011-
dc.identifier.scopuseid_2-s2.0-80052726060-
dc.identifier.volume45-
dc.identifier.issue1-
dc.identifier.spage231-
dc.identifier.epage240-
dc.identifier.isiWOS:000295760700019-
dc.identifier.issnl0031-3203-

Export via OAI-PMH Interface in XML Formats


OR


Export to Other Non-XML Formats