Applying KPCA to improving information compression and speckle reduction for multifrequency polarimetric SAR image

To our knowledge, there does not exist any paper in the open literature about making use of KPCA (kernel principal component analysis) for improving information compression and speckle reduction for multifrequency polarimetric SAR (synthetic aperture radar) image. We now present our research results on such an application. In the full paper, we explain our research results in some detail; in this abstract, we just add some pertinent remarks to listing the two topics of explanation. The first topic is: KPCA method. In this topic, we mention that KPCA is the nonlinear generalization of linear principal component analysis (PCA) using a kernel trick, which utilizes the Mercer kernel function to calculate the dot product of feature space. The second topic is: information compression and speckle reduction based on KPCA. In this topic, we derive Eq. (10) in the full paper to apply KPCA to directly processing the intensity or amplitude of multipolarimetric SAR images. The first few principal component images thus obtained compress information, reduce speckle and strengthen details. Finally we take the NASA/JPL multipolarimetric SAR images of P, L, and C band quadpolarizations as illustrative images to experiment on our research. The experimental results show preliminarily that our KPCA method can extract and compress the information of original images more effectively than linear PCA and only involves the calculation of eigenvalues of a kernel matrix.