Abstract
Spectral-spatial classification has been widely applied for remote sensing applications, especially for hyperspectral imagery. Traditional methods mainly focus on local spatial similarity and neglect nonlocal spatial similarity. Recently, nonlocal self-similarity (NLSS) has gradually gained support since it can be used to support spatial coherence tasks. However, these methods are biased towards the direct use of spatial information as a whole, while discriminative spectral information is not well exploited. In this paper, we propose a novel method to couple both nonlocal spatial and local spectral similarity together in a single framework. In particular, the proposed approach exploits nonlocal spatial similarities by searching non-overlapped patches, whereas spectral similarity is analyzed locally within the locally discovered patches. By fusion of nonlocal and local information, we then apply group sparse representation (GSR) for classification based on a group structured prior. Experimental results on three real hyperspectral data sets demonstrate the efficiency of the proposed approach, and the improvements are significant over the methods that consider either nonlocal or local similarity.
Highlights
Recent advances in remote sensing sensors, especially for hyperspectral imagery (HSI), increase the possibility of more accurate discrimination of materials of interest [1,2]
Step 1: Extract a S1 × S1 sized patch denoted as Pi,j with centered pixel of xi,j from X; Step 2: Search the nonlocal spatial similar patch Pa,b towards Pi,j according to Equation (6); Step 3: Search the local spectral similar pixel xu,v in Pa,b towards the current testing pixel xi,j according to Equation (7); Step 4: Extract two S2 × S2 sized patches with the center pixels of xi,j and xu,v, and transform them in to two-dimensional formed matrix denoted as Xi,j and Xu,v ; Step 5: Obtain the fusion matrix X f of Xi,j and Xu,v, and use them using group SRC (GSRC) to obtain the coefficient matrix A
The superiority of local spectral similarity through Nonlocal Spatial and Local Spectral Similarity (NSLS) is confirmed by the best overall classification accuracies obtained by the proposed NSLS-GSRC method in all cases, which allows for an overall consideration of local and nonlocal spatial information
Summary
Recent advances in remote sensing sensors, especially for hyperspectral imagery (HSI), increase the possibility of more accurate discrimination of materials of interest [1,2]. Nonlocal self-similarity (NLSS) defines the spatial consistency of materials in a global distribution [31] It can provide references of global structure prior by exploiting the spatial similarity in nonlocal area, such that the discrimination of a pixel can be more precisely processed through the global similarity constraint [32]. In [37], NLSS was introduced into SR to reconstruct the dictionary for separation of signal and noise, and contributed to more concise and accurate restoration of HSI Though these NLSS-based methods have shown their superiority based on global structured priors in spectral and spatial domains, they essentially tend to the direct use of nonlocal spatial similarity with concentration of spatial information from all the available regions, while the spectral features can still be exploited for more accurate discrimination. Based on the exploration of local constraint under local and nonlocal spatial consistency
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
