Pascal triangle transform approach to the calculation of 3D moments

Abstract Three-dimensional moments have been widely used in computer vision, but until now obtaining 3D moments had always needed much computation, which has not been resolved well. In this paper, we propose a fast and simple algorithm for calculating 3D moments. First, the Pascal triangle transform (PTT) is used to calculate monomials with one variable. Second, a powerful systolic structure for implementing the PTT with additions only is proposed. Third, the calculation of monomials is extended to those with three variables. Finally, sequential and parallel algorithms that need no multiplications are provided for calculating 3D moments. Compared with known methods, the new method reduces the computational complexity significantly.