Services on Demand
Journal
Article
Indicators
Cited by SciELO
Access statistics
Related links
Cited by Google
Similars in SciELO
Similars in Google
Share
TecnoLógicas
Print version ISSN 0123-7799On-line version ISSN 2256-5337
Abstract
BACCA, Jorge and ARGUELLO, Henry. Sparse Subspace Clustering for Hyperspectral Images using Incomplete Pixels. TecnoL. [online]. 2019, vol.22, n.46, pp.6-19. ISSN 0123-7799. https://doi.org/10.22430/22565337.1205.
Spectral image clustering is an unsupervised method that identifies distributions of pixels using spectral information without requiring a previous training stage. Sparse subspace clustering methods assume that hyperspectral images lie in the union of multiple low-dimensional subspaces. Therefore, sparse subspace clustering assigns spectral signatures to different subspaces, expressing each spectral signature as a sparse linear combination of all the pixels, ensuring that the non-zero elements belong to the same class. Although such methods have achieved good accuracy for unsupervised classification of hyperspectral images, their computational complexity becomes intractable as the number of pixels increases, i.e., when the spatial dimensions of the image become larger. For that reason, this paper proposes to reduce the number of pixels to be classified in the hyperspectral image; subsequently, the clustering results of the missing pixels are obtained by exploiting spatial information. Specifically, this work proposes two methodologies to remove pixels: the first one is based on spatial blue noise distribution, which reduces the probability of removing neighboring pixels, and the second one is a sub-sampling procedure that eliminates every two contiguous pixels, preserving the spatial structure of the scene. The performance of the proposed spectral image clustering framework is evaluated using three datasets, which shows that a similar accuracy is achieved when up to 50% of the pixels are removed. In addition, said framework is up to 7.9 times faster than the classification of the complete data sets.
Keywords : Spectral images; Spectral clustering; Sparse subspace clustering; Sub-sampling; Image classification.