Region-Based Spatial Preprocessing for Endmember Extraction and Spectral Unmixing

Abstract

In this paper, we develop a new spatial preprocessing strategy which can be applied prior to a spectral-based endmember extraction process for unmixing of hyperspectral data. Our proposed approach directs the endmember searching process to regions which are both spectrally pure and spatially homogeneous in the scene. Our experimental results, conducted using simulated hyperspectral data sets with known endmembers and fractional abundances, reveal that the proposed approach can successfully integrate the spatial and spectral information in the search for more relevant endmembers.

Collinearity and orthogonality of endmembers in linear spectral unmixing

“…Recently, the concept of spatial purity of pixels as an addition to pixel spectral purity was introduced and spatial neighborhoods were used to integrate spectral and spatial information in endmember selection (Mei et al, 2010). Recently (Martin and Plaza, 2011), a region-based spatial preprocessing approach for endmember extractions was proposed that determines a set of spatially representative regions with associated spectra. Theoretically this approach, although not tested in the paper, could significantly reduce collinearity because it selects out of a predefined set of regions those pixels with associated spectra that are both spectrally pure and orthogonal.…”

A survey of methods incorporating spatial information in image classification and spectral unmixing

“…As such, when a geometrically based algorithm is applied to the adjusted image, pixels located in spatially homogenous regions are more likely to be identified as endmembers. Many preprocessing methods adopt segmentation or clustering techniques to partition remotely sensed imagery into homogenous regions and use the mean spectra of each region for endmember extraction algorithms (Thompson et al 2010;Martin and Plaza 2011;Zhang et al 2014a). The spatial-spectral preprocessing (SSPP) algorithm (Martin and Plaza 2012) provides a more integrated framework to combine both spatial homogeneity and spectral purity at the preprocessing level.…”

Supervised Nonlinear Unmixing of Hyperspectral Images Using a Pre-image Methods

“…Hyperspectral analysis techniques can thus benefit from the inherent spatial-spectral duality in hyperspectral scenes. Following this idea, researchers exploited spatial information for endmember estimation (Martin & Plaza 2011;Rogge et al 2007;Zortea & Plaza 2009) and pixel vectors classification (Fauvel et al 2012, to appear;Li et al 2011). Recently, spatial processing methods were also derived for semi-supervised unmixing (Chen et al 2013a).…”