Sequential dimensionality reduction for extracting localized features

详细信息    查看全文

• 作者：Gabriella Casalino^a ; ^{gabriella.casalino@uniba.it}Author Vitae ; Nicolas Gillis^b ; ^{nicolas.gillis@umons.ac.be}Author Vitae
• 关键词：Nonnegative matrix factorization ; Underapproximation ; Sparsity ; Hyperspectral imaging ; Dimensionality reduction ; Spatial information
• 刊名：Pattern Recognition
• 出版年：2017
• 出版时间：March 2017
• 年：2017
• 卷：63
• 期：Complete
• 页码：15-29
• 全文大小：5551 K
• 卷排序：63

文摘

We propose a new nonnegative matrix underapproximation model for images. We design an algorithm that runs in linear time in the dimensions of the input matrix. This allows us to extract sequentially localized and spatially coherent features. We illustrate the effectiveness of our approach on a synthetic data set, facial and hyperspectral images. We show that it competes favorably with comparable state-of-the-art techniques.