Nonlocal Total Variation Subpixel Mapping for Hyperspectral Remote Sensing Imagery

Abstract

Subpixel mapping is a method of enhancing the spatial resolution of images, which involves dividing a mixed pixel into subpixels and assigning each subpixel to a definite land-cover class. Traditionally, subpixel mapping is based on the assumption of spatial dependence, and the spatial correlation information among pixels and subpixels is considered in the prediction of the spatial locations of land-cover classes within the mixed pixels. In this paper, a novel subpixel mapping method for hyperspectral remote sensing imagery based on a nonlocal method, namely nonlocal total variation subpixel mapping (NLTVSM), is proposed to use the nonlocal self-similarity prior to improve the performance of the subpixel mapping task. Differing from the existing spatial regularization subpixel mapping technique, in NLTVSM, the nonlocal total variation is used as a spatial regularizer to exploit the similar patterns and structures in the image. In this way, the proposed method can obtain an optimal subpixel mapping result and accuracy by considering the nonlocal spatial information. Compared with the classical and state-of-the-art subpixel mapping approaches, the experimental results using a simulated hyperspectral image, two synthetic hyperspectral remote sensing images, and a real hyperspectral image confirm that the proposed algorithm can obtain better results in both visual and quantitative evaluations.

Contemporary Liquid Brine Exploration on Mars: From Spectral Unmixing to Subpixel Mapping

“…Benefiting from the expansibility of the MAP estimation model, various spatial regularization terms can be incorporated for the indication of the detailed‐scale spatial distribution. For instance, total variation (TV)‐based MAP (Ling et al, ; Zhong et al, ) uses the order variation in the horizontal and vertical directions to construct the spatial regularization term, for the purpose of smoothness and boundary preservation; nonlocal based MAP (Feng, Zhong, Wu, et al, ; Feng, Zhong, Xu, et al, ) adopts the nonlocal spatial similarity in the same image for information increment; multishifted image based MAP (Chen et al, ) uses multiple images with subpixel scale shift for information increment; and the sparse representation based MAP approach models the sparse property of the image for the construction of the spatial regularization term (Feng, Zhong, Wu, et al, ; Feng, Zhong, Xu, et al, ). The MAP model combined with spatial regularization term has been successfully applied to regularize the SPM problem, improving the SPM result compared to the traditional methods.…”

Image Fusion for Spatial Enhancement of Hyperspectral Image via Pixel Group Based Non-Local Sparse Representation

“…However, owing to the technical and budget constraints, there is a tradeoff between spectral resolution and spatial resolution, which often implies low spatial resolution of HSIs. This fact may severely impede the practical use of HSIs and, therefore, various spatial resolution enhancement algorithms [2][3][4][5] have been proposed with spatial and spectral fusion approaches playing an important role. In contrast to hyperspectral sensors, multispectral sensors produce images with relatively higher spatial resolution but less spectral bands.…”

Spatial-Spectral Classification Based on the Unsupervised Convolutional Sparse Auto-Encoder for Hyperspectral Remote Sensing Imagery

“…Meanwhile, with the development of the sensors, the hyperspectral imaging techniques can also provide abundant detail and structural spatial information (Grahn et al, 2007, Camps-Valls et al, 2014, Landgrebe et al, 2003, Zhao et al, 2015a, Zhao et al, 2015b. The high spectral resolution and high spatial resolution properties enable the hyperspectral imagery data to become very useful and widely applicable in agriculture, surveillance, astronomy, mineralogy, and environment science areas (Chang et al, 2013, Fauvel et al, 2013, Feng et al, 2016, Jiao et al, 2015. Among the various application areas, the most common utilization of the hyperspectral imagery data is the ground object classification.…”