Rotation invariant texture classification algorithm based on Curvelet transform and SVM

Abstract

A rotation invariant texture classification algorithm based on curvelet transform and support vector machines (SVM) is proposed. The multidirectional and multiscale curvelet transform can offer more texture information and its energies are more compact as well as the SVM can achieve better classification. Compute the energies of the subbands acquired by curvelet transform to texture image first, then extract the rotation invariant feature vectors of isotropic, anisotropic and circular shift. The SVM algorithm is used to the texture classification at last. This method is compared with other rotation invariant texture classification algorithm, the experiment results show that it can improve the classification rate effectively.