Abstract:Hyperspectral unmixing is an important technique for remote sensing image exploitation. It aims to decompose a mixed pixel into a collection of spectrally pure components (called endmembers), and their corresponding proportions (called fractional abundances). In recent years, many studies have revealed that unmixing using spectral information alone does not sufficiently incorporate the spatial information in the remotely sensed hyperspectral image, as the pixels are treated as isolated entities without taking into account the existing local correlation among them. To address this issue, several spatial preprocessing methods have been developed to include spatial information in the spectral unmixing process. In this paper, we present a new spatial preprocessing method which presents several advantages over existing methods. The proposed method is derived from the Simple Linear Iterative Clustering (SLIC) method, which adapts the global search scope of the clustering into local regions. As a result, the spatial correlation and the spectral similarity are intrinsically incorporated at the clustering step, which results in O(N) computational complexity of the clustering procedure with N being the number of pixels in the image. First, a regional clustering is iteratively performed by using spatial and spectral information simultaneously. The obtained result is a set of clustered partitions that exhibit both spectral similarity and spatial correlation. Then, for each partition we select a subset of candidate pixels with high spectral purity. Finally, the obtained candidate pixels are gathered together and fed to a spectral-based endmember extraction method to extract the final endmembers and their corresponding fractional abundances. Our newly developed method naturally integrates the spatial and the spectral information to retain the most relevant endmember candidates. Our experimental results, conducted using both synthetic and real hyperspectral scenes, indicate that the proposed method can obtain accurate unmixing results with less than 0.5% of the number of pixels used by other state-of-the-art methods. This confirms the advantages of integrating spatial and spectral information for hyperspectral unmixing purposes.

Regional Clustering-Based Spatial Preprocessing for Hyperspectral Unmixing

Local Sparse Representation Based Spatial Preprocessing For Endmember Extraction

Improving Spectral-Based Endmember Finding by Exploring Spatial Context for Hyperspectral Unmixing

Using Linear Spectral Unmixing for Subpixel Mapping of Hyperspectral Imagery: A Quantitative Assessment

Spectral and Spatial Complexity-Based Hyperspectral Unmixing

A spatial-spectral clustering-based algorithm for endmember extraction and hyperspectral unmixing

Superpixel construction for hyperspectral unmixing.

Region-Based Spatial Preprocessing for Endmember Extraction and Spectral Unmixing.

Hybrid Unmixing Based on Adaptive Region Segmentation for Hyperspectral Imagery

Correntropy-Based Spatial-Spectral Robust Sparsity-Regularized Hyperspectral Unmixing

Superpixel-Guided Sparse Unmixing For Remotely Sensed Hyperspectral Imagery

Band Selection Based Hyperspectral Unmixing

Sketch-Based Region Adaptive Sparse Unmixing Applied to Hyperspectral Image

Joint Local Block Grouping with Noise-Adjusted Principal Component Analysis for Hyperspectral Remote-Sensing Imagery Sparse Unmixing

Combined Nonlocal Spatial Information and Spatial Group Sparsity in NMF for Hyperspectral Unmixing

Local Spectral Similarity-Guided Sparse Unmixing of Hyperspectral Images with Spatial Graph Regularization

Unsupervised Hyperspectral Remote Sensing Image Clustering Based on Adaptive Density

Spatial-spectral preprocessing for volume-based endmember extraction algorithms using unsupervised clustering

Spatial-Spectral Preprocessing Prior to Endmember Identification and Unmixing of Remotely Sensed Hyperspectral Data

Hypergraph-Regularized Sparse NMF for Hyperspectral Unmixing.

Spatial Discontinuity-Weighted Sparse Unmixing of Hyperspectral Images.