Use of neighborhood unhomogeneity to detect the edge of hyperspectral spatial stray light region

Since hyperspectral images contain fine spectral information of the target objects, there is continuous demand for the detailed analysis about such images. A somewhat neglected but critically important challenge regarding hyperspectral image is the presence of "spatial stray light" - a phenomenon caused by scattering of photons from target objects. In this study, neighborhood unhomogeneity method is used to detect the edge of spatial stray light region in hyperspectral images. Analyses were conducted on reflectance datasets of standard card, potato leaf, and corn kernels from three inbred lines. By using neighborhood unhomogeneity method, the edge of spatial stray light region could be detected obviously and experimental results show the effectiveness of the proposed method. © 2014 Elsevier GmbH.