Hyperspectral Image Classification Based on SPCA and Domain Transform Recursive Filtering
YU Duo, HUANG Yongdong
1.Institute of Image Processing and Understanding, North Minzu University, Yinchuan 750021, China
2.Center for Mathematics and Information Science, Dalian Minzu University, Dalian, Liaoning 116600, China
A new hyperspectral image classification method is proposed based on Segmented Principal Component Analysis（SPCA） and Domain Transform Recursive Filtering（DTRF）. First, the SPCA method is used to reduce the dimension of hyperspectral image and extract the first principal component of each band subset. Then, DTRF with different parameters deals with the first principal component of each band subset to form a stacked edge-preserving filter map. And the Principal Component Analysis（PCA） is used to fuse the features of the stack-preserving filter map. At last, the Basic Thresholding Classifier（BTC） classifies the fused principal components. Simulation experiments show that the proposed method can improve the classification accuracy, and the overall accuracy, average accuracy and Kappa coefficient are higher than the some existing methods.
