Hyperspectral Band Selection via Optimal Combination Strategy

Band selection is one of the main methods reducing number dimensions in a hyperspectral image. Recently, various have been proposed to address this issue. However, these usually obtain band subset perspective locally optimal solution. To achieve an solution with global perspective, paper developed novel method for via combination strategy (OCS). The contributions are as follows: (1) subspace partitioning approach which can accurately points subspace. This ensures that similar bands be divided into same subspace; (2) two candidate representative large amount information and high similarity chosen from each subspace, fully represent all (3) designed acquire subset, achieves perspective. results on four public datasets illustrate satisfactory performance against other methods.