In this paper, a general algorithm for the computation of the Fourier coefficients of 2π-periodic (continuous) functions is developed based on Dirichlet characters, Gauss sums and the generalized Möbius transform. It permits the direct extraction of the Fourier cosine and sine coefficients. Three special cases of our algorithm are presented. A VLSI architecture is presented and the error estimates are given. §1 Introduction and notationThe arithmetic Fourier transform (AFT) offers a convenient method, based on the construction of weighted averages, to calculate the Fourier coefficients of a complex-valued periodic function. It was discovered by Bruns [3] in 1903. Similar algorithms were studied by Tufts and Sadasiv [15] for the calculation of the Fourier coefficients of even periodic functions. This method was extended in [11] to calculate the Fourier coefficients of both the even and odd components of any periodic function.In [8] and [9], Knockaert presented the theory of generalized Möbius transform and gave a general formulation, which was developed by the authors [4]. Meanwhile, the 1-D AFT derivation in [15] was directly extended to 2-D function (see [10], [12], [16]). Kelley [7] and coworkers proposed a row-column algorithm in which the 2-D Fourier transform is decomposed into 1-D AFT. Atlas [2] developed the concept of two-dimensional AFT using the Bruns' method. In [5], Bruns's version of 1-D AFT is extended to the 2-D case. Schiff [14] and Hsu [6] used the Möbius inversion formulae to compute the inverse Z-transform. Furthermore, some VLSI architectures for the AFT are developed in [7] and [13].Has AFT been exhaustive?