Parameters selection of morphological scale-spacedecomposition for hyperspectral images using tensormodeling

摘要

Dimensionality reduction (DR) using tensor structures in morphological scale-space decomposition (MSSD) forHSI has been investigated in order to incorporate spatial information in DR. We present results of a comprehensiveinvestigation of two issues underlying DR in MSSD. Firstly, information contained in MSSD is reduced usingHOSVD but its nonconvex formulation implicates that in some cases a large number of local solutions can befound. For all experiments, HOSVD always reach an unique global solution in the parameter region suitable topractical applications. Secondly, scale parameters in MSSD are presented in relation to connected componentssize and the influence of scale parameters in DR and subsequent classification is studied.