Home / Research / Hyperspectral imagery /

A split-and-merge approach for hyperspectral band selection

The problem of band selection is of great importance to handle the curse of dimensionality for hyperspectral image applications (e.g., classification). This paper proposes an unsupervised band selection approach based on a split-and-merge concept. This new approach provides relevant spectral subbands by splitting the adjacent bands without violating the physical meaning of the spectral data. Next, it merges highly correlated bands and subbands to reduce the dimensionality of the hyperspectral image. Experiments on three public datasets and comparison with state-of-the-art approaches show the efficiency of the proposed approach.
PLMM-based unmixing results
Fig. 1. OA, kappa, and number of bands for each method on Indian Pines and Pavia datasets.

The band selection algorithm is described in the paper published in IEEE Geosci. Remote Sens. Letters in 2017:

The band selection algorithm is available as a MATLAB code:

sitemeter stats