Abstract: A paper-cutting recognition algorithm was proposed based on Multiresolution Fourier-Mellin transform. The algorithm does Fourier-Mellin transform on paper-cutting, then based on the invariance of Fourier-Mcllin transform, calculates the energies of the subbands acquired by wavelet transform, which is composed by mean square deviation and mean value of apiece subband. These energies regard as the invariant feature vectors. Experiments indicate that this feature vector not only has the translation, rotation and scale invariance, but also satisfactorily achieves the excessive geometrical deformation pattern recognition.

Key words: Multiresolution Fourier-Mellin transform, Feature extraction, SVM, Pattern recognition