Three novel invariant moments based on radon and polar harmonic transforms

This paper presents three novel invariant moments on the basis of radon and Polar Harmonic Transforms (PHTs). Gray images have been studied, as they can avoid the error of binarizating. Radon transform is introduced with great robustness to noise. PHTs are Orthogonal and numerically stable and their kernels are simpler compared with most other moments. Thus they are applied to the radon image. Then we define three new invariant moments, and three invariant feature vectors. The Libsvm classifier is employed to implement classification in the experiments. The theoretical analysis and experimental results indicate that the classification accuracy of our method is higher and with lower time cost than that of other moments. They are also robust to noise.