Advances in Hyperspectral Image Processing Techniques

6Progressive Band Selection Processing for Hyperspectral Image Classification

Chunyan Yu¹ Meiping Song¹ and Chein‐I Chang²

¹Center for Hyperspectral Imaging in Remote Sensing (CHIRS), Information and Technology College, Dalian Maritime University, Dalian, China

²Remote Sensing Signal and Image Processing Laboratory, Department of Computer Science and Electrical Engineering, University of Maryland Baltimore County, Baltimore, MD, USA

6.1 Introduction

Hyperspectral image classification (HSIC) has received considerable interest in recent years, just to name a few [1–15]. A general approach is to use a spectral classifier such as support vector machine (SVM) to perform spectral classification coupled with a spatial domain‐based technique to capture spatial contextual information such as Markov random field (MRF) model [5], extended morphological profiles (EMP) [2], and edge preserving filter (EPF) [7] to take care of both spectral and spatial correlation. SVM is originally designed to perform binary classification. So, when it is used to perform multi‐class classification, two common approaches are considered, one‐against‐one and one‐against‐rest [16]. No matter which one is adopted for multi‐class classification, there are always inherent issues in misclassification that need to be addressed [16, pp. 182–184]. Most importantly, such multi‐class classification is usually carried out in a one‐shot operation in the sense that all multiple classes are classified by a classifier simultaneously ...