Subspace learning from image gradient orientations

We introduce the notion of subspace learning from image gradient orientations for appearance-based object recognition. As image data is typically noisy and noise is substantially different from Gaussian, traditional subspace learning from pixel intensities fails very often to estimate reliably the l...

Full description

Bibliographic Details
Main Authors: Tzimiropoulos, Georgios, Zafeiriou, Stefanos, Pantic, Maja
Format: Article
Published: Institute of Electrical and Electronics Engineers 2012
Subjects:
Online Access:https://eprints.nottingham.ac.uk/31424/
_version_ 1848794198352855040
author Tzimiropoulos, Georgios
Zafeiriou, Stefanos
Pantic, Maja
author_facet Tzimiropoulos, Georgios
Zafeiriou, Stefanos
Pantic, Maja
author_sort Tzimiropoulos, Georgios
building Nottingham Research Data Repository
collection Online Access
description We introduce the notion of subspace learning from image gradient orientations for appearance-based object recognition. As image data is typically noisy and noise is substantially different from Gaussian, traditional subspace learning from pixel intensities fails very often to estimate reliably the low-dimensional subspace of a given data population. We show that replacing pixel intensities with gradient orientations and the 2 norm with a cosine-based distance measure offers, to some extend, a remedy to this problem. Within this framework, which we coin IGO (Image Gradient Orientations) subspace learning, we first formulate and study the properties of Principal Component Analysis of image gradient orientations (IGO-PCA). We then show its connection to previously proposed robust PCA techniques both theoretically and experimentally. Finally, we derive a number of other popular subspace learning techniques, namely Linear Discriminant Analysis (LDA), Locally Linear Embedding (LLE) and Laplacian Eigenmaps (LE). Experimental results show that our algorithms outperform significantly popular methods such as Gabor features and Local Binary Patterns and achieve state-of-the-art performance for difficult problems such as illumination- and occlusion-robust face recognition. In addition to this, the proposed IGO-methods require the eigen- ecomposition of simple covariance matrices and are as computationally efficient as their corresponding 2 norm intensity-based counterparts. Matlab code for the methods presented in this paper can be found at http://ibug.doc.ic.ac.uk/resources
first_indexed 2025-11-14T19:12:23Z
format Article
id nottingham-31424
institution University of Nottingham Malaysia Campus
institution_category Local University
last_indexed 2025-11-14T19:12:23Z
publishDate 2012
publisher Institute of Electrical and Electronics Engineers
recordtype eprints
repository_type Digital Repository
spelling nottingham-314242020-05-04T20:21:07Z https://eprints.nottingham.ac.uk/31424/ Subspace learning from image gradient orientations Tzimiropoulos, Georgios Zafeiriou, Stefanos Pantic, Maja We introduce the notion of subspace learning from image gradient orientations for appearance-based object recognition. As image data is typically noisy and noise is substantially different from Gaussian, traditional subspace learning from pixel intensities fails very often to estimate reliably the low-dimensional subspace of a given data population. We show that replacing pixel intensities with gradient orientations and the 2 norm with a cosine-based distance measure offers, to some extend, a remedy to this problem. Within this framework, which we coin IGO (Image Gradient Orientations) subspace learning, we first formulate and study the properties of Principal Component Analysis of image gradient orientations (IGO-PCA). We then show its connection to previously proposed robust PCA techniques both theoretically and experimentally. Finally, we derive a number of other popular subspace learning techniques, namely Linear Discriminant Analysis (LDA), Locally Linear Embedding (LLE) and Laplacian Eigenmaps (LE). Experimental results show that our algorithms outperform significantly popular methods such as Gabor features and Local Binary Patterns and achieve state-of-the-art performance for difficult problems such as illumination- and occlusion-robust face recognition. In addition to this, the proposed IGO-methods require the eigen- ecomposition of simple covariance matrices and are as computationally efficient as their corresponding 2 norm intensity-based counterparts. Matlab code for the methods presented in this paper can be found at http://ibug.doc.ic.ac.uk/resources Institute of Electrical and Electronics Engineers 2012-12 Article PeerReviewed Tzimiropoulos, Georgios, Zafeiriou, Stefanos and Pantic, Maja (2012) Subspace learning from image gradient orientations. IEEE Transactions on Pattern Analysis and Machine Intelligence, 34 (12). pp. 2454-2466. ISSN 1939-3539 image gradient orientations robust principal component analysis discriminant analysis non-linear dimensionality reduction face recognition http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=6138864 doi:10.1109/TPAMI.2012.40 doi:10.1109/TPAMI.2012.40
spellingShingle image gradient orientations
robust principal component analysis
discriminant analysis
non-linear dimensionality reduction
face recognition
Tzimiropoulos, Georgios
Zafeiriou, Stefanos
Pantic, Maja
Subspace learning from image gradient orientations
title Subspace learning from image gradient orientations
title_full Subspace learning from image gradient orientations
title_fullStr Subspace learning from image gradient orientations
title_full_unstemmed Subspace learning from image gradient orientations
title_short Subspace learning from image gradient orientations
title_sort subspace learning from image gradient orientations
topic image gradient orientations
robust principal component analysis
discriminant analysis
non-linear dimensionality reduction
face recognition
url https://eprints.nottingham.ac.uk/31424/
https://eprints.nottingham.ac.uk/31424/
https://eprints.nottingham.ac.uk/31424/