Automatic endmember extraction using pixel purity index for hyperspectral imagery

作者：Zhou Qianlan; Zhang Jing^*; Tian Qi; Zhuo Li; Geng Wenhao

来源：22nd International Conference on MultiMedia Modeling, MMM 2016, 2016-01-04 to 2016-01-06.

DOI：10.1007/978-3-319-27674-8_19

• Endmember extraction
• Endmember extraction algorithms
• Hyper-spectral imageries
• Hyper-spectral images
• Minimum noise fraction
• Noise covariance matrix
• Pixel purity index
• Projection vectors

Pixel Purity Index (PPI) is one of effective endmember extraction algorithms, which is a processing technique designed to determine which pixels are the most spectrally unique or pure. This paper proposes an automatic endmember extraction using pixel purity index for hyperspectral imagery. In computing the PPI, projection vectors are generated by applying the Givens rotation firstly. Then, pixels are projected onto the projection vectors. Next, the pixels located at the extreme positions are recorded. At last, the PPI score can be obtained. In endmember extraction, the number of endmembers is determined by using the Noise Subspace Projection (NSP) method. Hyperspectral image dimension is reduced by improving the Noise Covariance Matrix (NCM) estimation for Minimum Noise Fraction (MNF) transformation. The endmembers can be extracted with the improved pixel purity index. Compared with traditional APPI algorithm, the experimental results show that the proposed algorithm can obtain more endmembers as well as improve the accuracy of endmembers.

• 出版日期2016
• 单位北京大学