An efficient algorithm for fast computation of orthogonal Fourier-Mellin moments

Orthogonal Fourier-Mellin moments have better feature extraction capabilities and are more robust to image noise than the classical Zernike moments. However, orthogonal Fourier-Mellin moments have not been widely used as features in pattern recognition due to the computational complexity of the orthogonal Fourier-Mellin radial polynomials. This paper analyzes the deficiencies of the existing methods, and introduces an efficient recursive algorithm to compute the orthogonal Fourier-Mellin moments. The algorithm consists of a recurrence relation for Mellin orthogonal radial polynomials, which derived from the Jacobi polynomials for fast computation of orthogonal Fourier-Mellin moments. An experiment using binary image is designed to test the performance of the algorithm. The experimental result demonstrates that the computational speed of orthogonal Fourier-Mellin moments has been adequately improved over the present methods.