Hyperspectral Image (HSI) Processing based on Local Data Distribution Analysis in LFDA Algorithm

Lina Yang, Huiwu Luo, and Yuan Yan Tang


Intelligent Signal Processing, Applications, Local Linear discriminant analysis, Dimension reduction, Hyperspectral image (HSI)


A pixel in Hyperspectral Image (HSI) can be regarded as a signal, which is represented by a vector with high dimension. In this paper, an intelligent signal processing method is used to treat HSI. Dimension reduction, as a preprocessing step, plays a significant role in the procedure of HSI classification. Based on the typical behavior of Local Fisher Discriminant Analysis (LFDA), an improved LFDA which is called ILFDA is proposed in this paper. According to the affinity matrix in the local linear discriminant analysis, the local variance and prior probability are adopted to describe the local data distribution. This method consider not only the sample data distribution characteristics, but also the discriminant analysis in HSI classification. We carry out experiments on a real HSI database, and the results of the overall accuracy and kappa coefficient indicate that the proposed method is more effective than conventional dimension reduction methods.

Important Links:

Go Back