Modification of correlation kernels in SVM, KPCA and KCCA in texture classification

Modified versions of the correlation kernels in the kernel methods, e.g., SVMs, kPCA and kCCA are presented, which are based on the L_p norm and max norm as well as the blindness of the odd-order autocorrelations to sinusoidal or symmetrically distributed signals. The poor generalization of the higher-order correlation kernels and the inferior performance of the correlation kernels of odd-orders to even-orders are improved with the modifications. The performance of the modified correlation kernels is evaluated and compared in texture classification experiments.