Content-based feature weighting for scene recognition

Image representation for content-based scene recognition is considered as one of the most challenging problems in computer vision. Motivated by the successful text information retrieval application through a histogram of word counts for each text document, the promising bag of features (BoF) model (...

Full description

Saved in:
Bibliographic Details
Main Author: Zhu, Jing
Other Authors: Ma Kai Kuang
Format: Theses and Dissertations
Language:English
Published: 2010
Subjects:
Online Access:https://hdl.handle.net/10356/38771
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
id sg-ntu-dr.10356-38771
record_format dspace
spelling sg-ntu-dr.10356-387712023-07-04T16:09:51Z Content-based feature weighting for scene recognition Zhu, Jing Ma Kai Kuang School of Electrical and Electronic Engineering DRNTU::Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision Image representation for content-based scene recognition is considered as one of the most challenging problems in computer vision. Motivated by the successful text information retrieval application through a histogram of word counts for each text document, the promising bag of features (BoF) model (or the bag of visual words model) has been developed in recent years and clearly demonstrated its potential as a powerful image representation technique for visual information retrieval application. The BoF model represents each image as a collection of clustered local features, and the centroid of each cluster is known as a visual word, which is analogous to a text word. Since the contextual information (i.e., the inter-word relationship) among the visual words is omitted in the conventional BoF model, a novel feature weighting method, coined as the FeatureRank (FR) in this thesis, has been proposed to address this issue. Furthermore, a non-parametric clustering scheme has been developed, which does not require any prior knowledge regarding the number of clusters involved. Finally, the FeatureRank-based weighting scheme has been developed and incorporated into the BoF model for improving the performance of scene recognition. MASTER OF ENGINEERING (EEE) 2010-05-18T07:00:27Z 2010-05-18T07:00:27Z 2010 2010 Thesis Zhu, J. (2010). Content-based feature weighting for scene recognition. Master’s thesis, Nanyang Technological University, Singapore. https://hdl.handle.net/10356/38771 10.32657/10356/38771 en 86 p. application/pdf
institution Nanyang Technological University
building NTU Library
continent Asia
country Singapore
Singapore
content_provider NTU Library
collection DR-NTU
language English
topic DRNTU::Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision
spellingShingle DRNTU::Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision
Zhu, Jing
Content-based feature weighting for scene recognition
description Image representation for content-based scene recognition is considered as one of the most challenging problems in computer vision. Motivated by the successful text information retrieval application through a histogram of word counts for each text document, the promising bag of features (BoF) model (or the bag of visual words model) has been developed in recent years and clearly demonstrated its potential as a powerful image representation technique for visual information retrieval application. The BoF model represents each image as a collection of clustered local features, and the centroid of each cluster is known as a visual word, which is analogous to a text word. Since the contextual information (i.e., the inter-word relationship) among the visual words is omitted in the conventional BoF model, a novel feature weighting method, coined as the FeatureRank (FR) in this thesis, has been proposed to address this issue. Furthermore, a non-parametric clustering scheme has been developed, which does not require any prior knowledge regarding the number of clusters involved. Finally, the FeatureRank-based weighting scheme has been developed and incorporated into the BoF model for improving the performance of scene recognition.
author2 Ma Kai Kuang
author_facet Ma Kai Kuang
Zhu, Jing
format Theses and Dissertations
author Zhu, Jing
author_sort Zhu, Jing
title Content-based feature weighting for scene recognition
title_short Content-based feature weighting for scene recognition
title_full Content-based feature weighting for scene recognition
title_fullStr Content-based feature weighting for scene recognition
title_full_unstemmed Content-based feature weighting for scene recognition
title_sort content-based feature weighting for scene recognition
publishDate 2010
url https://hdl.handle.net/10356/38771
_version_ 1772826883243638784