Progressive band selection processing for hyperspectral image classification

Chunyan Yu, Meiping Song, Chein I. Chang

Research output: Chapter in Book/Report/Conference proceedingChapter


This chapter introduces a new band selection (BS) approach to hyperspectral image classification, called progressive band selection processing of hyperspectral image classification (PBSP-HSIC), which performs image classification stage-by-stage progressively in the sense that each stage performs HSIC according to a specifically selected band subset. Its idea is first to introduce a criterion, referred to as class classification priority (CCP), to measure the priority of each class to be classified. These calculated CCP probabilities are then used to construct a p-ary Huffman coding tree (HCT) to navigate classification to be performed layer-by-layer where each layer is specified by a classification stage that selects a particular band subset to perform classification. Finally, classification is performed layer-by-layer along the HCT from top to down progressively. Interestingly, such PBSP-HSIC offers a rare view of how different classes being classified in progressive stages, which has never been explored in the past. The experimental results also show that PBSP-HSIC performs better than HSIC using full bands.

Original languageEnglish
Title of host publicationAdvances in Hyperspectral Image Processing Techniques
Number of pages25
ISBN (Print)9781119687788
Publication statusPublished - 2022 Nov 11

All Science Journal Classification (ASJC) codes

  • General Engineering


Dive into the research topics of 'Progressive band selection processing for hyperspectral image classification'. Together they form a unique fingerprint.

Cite this