Radiometric Correction and Feature Extraction of Molecular Hyperspectral Imaging Data

Some molecular hyperspectral images of retina sections were collected. Due to the infection of lamp, a spectral curve extracted directly from the original hyperspectral data can not truly present biochemical character. The main preprocessing step of the hyperspectral data is radiometric correction. The paper provides the gray correction coefficient algorithm to eliminating the influence. Because hyperspectral data cube includes a great deal of single band image, data redundancy is very serious. The paper cites that PCA(Principal Component Analysis) algorithm can validly extract feature information and eliminate data redundancy and achieve dimensionality reduction.