Abstract. For any given set of angles θ 0 < . . . < θ n in [0, π), we show that a set of n+2 2Radon projections, consisting of k parallel X-ray beams in each direction θ k , k = 0, . . . , n, determines uniquely algebraic polynomials of degree n in two variables.
Introduction.Most of the methods for approximate reconstruction of a univariate function f are based on sampling values of f at a finite number of points, and the tools used are usually those of interpolation. This is a natural approach to approximation of univariate functions since a table of function values is a standard type of information about f that comes as output in practical problems and processes described by functions in one variable, and in addition, the Lagrange interpolation problem by polynomials is always solvable. In the multivariate case, such an approach encounters serious difficulties. For example, it is well known that pointwise interpolation by multivariate polynomials is no more possible for every choice of nodes. Moreover, there are a lot of practical problems in which information about the relevant function comes as a set of functionals different from point evaluations. In tomography, electronic microscopy, and technics, the data often consists of values of linear integrals over segments. In many situations, a table of mean values of a function of d variables on (d − 1)-dimensional hyperplanes is considered to be the most natural type of data for multivariate functions. Hakopian's famous interpolation formula [6] (see also [7]) is an important reason to take this approach.Hakopian proved that for any given n + 2 distinct points X 0 , . . . , X n+1 on the boundary of a convex body D (say, a disk), the set of integrals of f over all the linear segments [X i , X j ] determines uniquely every polyno-