Local and Nonlocal Context-Aware Elastic Net Representation-Based Classification for Hyperspectral Images

Sparse and Low-Rank Representation With Key Connectivity for Hyperspectral Image Classification

“…Because 1 l -based SR may yield unstable representation results, Tang et al [20] incorporated manifold learning into SR to exploit the smoothness across neighboring samples. Moreover, considering that the limited labeled samples of different classes are unbalance and the learnt representation is hard to reflect the particular characteristics of each class, the graph based context-aware elastic net (ELN) [21] was proposed for HSI classification by taking full advantages of SR and CR, which promoted local and global consistence preserving. Recently, the combination of class probability and sparse representation has displayed excellent performance in HSI classification, mainly because class structure information helps the SR model to obtain a highly effective graphical expression.…”

A Bilateral Filter Based Post-Processing Approach for Supervised Spectral-Spatial Hyperspectral Image Classification

“…GCKs (Generalized CKs) [6], or MFL (Multiple Feature Learning) [7], these spectral-spatial features are integrated. voting [8], MRF (Markov Random Field) [9,10], graph regularization [11] or ERWs (Extended Random Walkers) [12].…”

“…Among them the kernel-based and morphological approaches results showed the good accuracies to classify HSI [5,14]. In our previous work [11], a graph based spectral and spatial self-similarity of local and non-local neighbors using the ELN (Elastic Net)-coding neighborhood graph is incorporated. Furthermore, Kang et.…”

“…Benefited with the weighted averaging approach of BF, for the post-processing step it is well suited to add the geometric information with the pixel-wise spectral-only classifier. Previous research [9,11,19] have indicated that post-processing classification is an important step in improving the quality of classifiers. In last decades'…”