Correlation-Based Band Selection for Hyperspectral Image Classification

TL;DR

The paper tackles high dimensionality and band redundancy in hyperspectral imaging by introducing a correlation-based band selection method. It computes pairwise band correlations using the Pearson metric $r_{XY}$ and defines the Average Band Correlation ${ABC}_i = \frac{1}{N-1} \sum_{j \neq i} | r_{B_i,B_j} |$, selecting bands with ${ABC}_i$ below a threshold (0.65) to retain diverse information. The reduced band sets are classified with a support vector machine (SVM) and evaluated on the PA and SA benchmark datasets, achieving competitive overall accuracy compared with PCA-based dimensionality reduction and a similarity-based unsupervised method. The approach is simple and efficient, enabling faster processing while preserving classification performance, though threshold selection remains an area for future work and refinement.

Abstract

Hyperspectral images offer extensive spectral information about ground objects across multiple spectral bands. However, the large volume of data can pose challenges during processing. Typically, adjacent bands in hyperspectral data are highly correlated, leading to the use of only a few selected bands for various applications. In this work, we present a correlation-based band selection approach for hyperspectral image classification. Our approach calculates the average correlation between bands using correlation coefficients to identify the relationships among different bands. Afterward, we select a subset of bands by analyzing the average correlation and applying a threshold-based method. This allows us to isolate and retain bands that exhibit lower inter-band dependencies, ensuring that the selected bands provide diverse and non-redundant information. We evaluate our proposed approach on two standard benchmark datasets: Pavia University (PA) and Salinas Valley (SA), focusing on image classification tasks. The experimental results demonstrate that our method performs competitively with other standard band selection approaches.

Figures (1)

• Figure 1: Comparison of Ground Truth, PCA, SB and Proposed method for SA and PA datasets. (a) and (e) are Ground Truth maps, (b) and (f) are Classification maps using PCA method, (c) and (g) are Classification maps using SB method and, (d) and (h) are Classification maps using Proposed method on SA and PA datasets, respectively.