Sequential dimensionality reduction for extracting localized features

PATTERN RECOGNITION

卷:63

Article

Casalino, Gabriella¹  Gillis, Nicolas² 

[1] Univ Bari A Moro, Dept Informat, Via E Orabona 4, I-70125 Bari, Italy

[2] Univ Mons, Dept Math & Operat Res, Rue Houdain 9, B-7000 Mons, Belgium

关键词: Nonnegative matrix factorization;    Underapproximation;    Sparsity;    Hyperspectral imaging;    Dimensionality reduction;    Spatial information;   

DOI  :  10.1016/j.patcog.2016.09.006

来源: Elsevier

【 摘 要 】

Linear dimensionality reduction techniques are powerful tools for image analysis as they allow the identification of important features in a data set. In particular, nonnegative matrix factorization (NMF) has become very popular as it is able to extract sparse, localized and easily interpretable features by imposing an additive combination of nonnegative basis elements. Nonnegative matrix under approximation (NMU) is a closely related technique that has the advantage to identify features sequentially. In this paper, we propose a variant of NMU that is particularly well suited for image analysis as it incorporates the spatial information, that is, it takes into account the fact that neighboring pixels are more likely to be contained in the same features, and favors the extraction of localized features by looking for sparse basis elements. We show that our new approach competes favorably with comparable state-of-the-art techniques on synthetic, facial and hyperspectral image data sets. (C) 2016 Elsevier Ltd. All rights reserved.

【 授权许可】

Free