Rotation invariant texture classification algorithm based on DT-CWT and SVM

Abstract

A rotation invariant texture classification algorithm based on dualtree complex wavelet transform (DT-CWT) and support vector machines (SVM) is proposed. First, the texture image is transformed by Radon transform to convert the rotation to translation, the rotation invariant feature vector is composed of the energies of the subbands acquired by DT-CWT which is shift invariant to the transformed texture image, the SVM algorithm is used to the texture classification at last. This algorithm is compared with the classifier of probabilistic neural network (PNN) and other rotation invariant texture classification algorithm, the experiment results show that it can improve the classification rate effectively. © Springer-Verlag Berlin Heidelberg 2007.