Robust Double Spatial Regularization Sparse Hyperspectral Unmixing

Abstract

With the help of endmember spectral library, sparse unmixing techniques have been successfully applied to hyperspectral image interpretation. The inclusion of spatial information in the sparse unmixing significantly improves the resulting fractional abundances. However, most existing spatial sparse unmixing algorithms are sensitive to noise and produce unstable solutions. To alleviate this drawback, a new robust double spatial regularization sparse unmixing (RDSRSU) method is proposed, which simultaneously exploits the spatial structure information from hyperspectral images and estimated abundance maps to mitigate the negative influence of noise on unmixing, so as to achieve robust sparse unmixing. To this end, a pre-calculated spatial weighting factor is introduced to maintain the original spatial information of the hyperspectral image. Meanwhile, the total variation spatial regularizer is used to capture the piecewise smooth structure of each abundance map. The experimental results, conducted by two sets of simulated data, as well as Cuprite and Mangrove real hyperspectral data, uncover that the proposed RDSRSU algorithm can offer better anti-noise ability and obtain more accurate results over those gave by other advanced sparse unmixing algorithms.

Superpixel-Based Weighted Sparse Regression and Spectral Similarity Constrained for Hyperspectral Unmixing

“…The performance of the WSRSSU algorithm is evaluated by using two synthetic datasets and three real hyperspectral images. Comparison algorithms involved in the experiment include SUnSAL [27], SUnSAL-TV [28], MUA SLIC [31], SUSRLR-TV [32], SBWCRLRU [33], and RDSRSU [14]. For quantitative comparison, two image quality evaluation indicators are adopted.…”

“…LSMM considers that an observed mixed spectrum is approximately expressed as a linear combination of endmember spectra and their abundance fractions [13]. Compared to NLSMM, LSMM has computational tractability and flexibility [14]. Hence, this article focuses on LSMM-based spectral unmixing algorithms.…”

“…It makes a major contribution to reducing the error caused by endmember extraction, considers a part of the spectra in the spectral library as endmembers, and directly estimate the abundance maps by solving constrained sparse regression problem [11]. Over the past ten years, several sparse unmixing algorithms [13], [14], [27], [28], [29], [30], [31], [32], [33] have been developed. This problem is described as a linear fitting problem, that is, an observed hyperspectral vector can be linearly fitted by spectral signatures in an available spectral library [27].…”

“…In recent years, sparse unmixing is not limited to one sparse regression, and double regression model [14], [31], [46], [47], [48] has become popular. The double regression model achieves unmixing through two sparse regressions.…”

Fast Hyperspectral Unmixing Using a Multiscale Sparse Regularization

“…The proposed FastUn algorithm is compared with state-ofart TV-based and competitive multiscale based unmixing algorithms: SUnSAL-TV [9], S 2 WSU [11], MUA [16], SUSRLR-TV [17], DRSUTV [10] and RDRSU [20]. These algorithms are quantitatively compared in term of signal reconstruction error (SRE) defined as SRE = 10 log 10…”