Interpolation of harmonic functions based on Radon projections (Q2510400)

Reconstructing a function of \(d\) variables using a table of its mean values on \((d-1)\)-dimensional hyperplanes is not only a problem of intrinsic interest within approximation theory, but also one of primary importance in applications such as computer tomography. Given the Radon projections of a harmonic function \(u\) along \(2n+1\) chords of the unit circle, this paper studies the reconstruction of \(u\) in the unit disk from the \((2n+1)\)-dimensional space of real bivariate harmonic polynomials of total degree at most \(n\). First, the authors prove a Marr-type formula for the Radon projections of the harmonic basis polynomials in terms of Chebyshev polynomials of second kind. This is shown to imply the existence of a unique solution to the above harmonic reconstruction problem in case all \(2n+1\) chords are chosen at a constant distance \(t\) from the origin, where \(t\) is not a zero of the first \(n\) Chebyshev polynomials of second kind. In case the chords form a regular convex \((2n+1)\)-sided polygon, a detailed analysis obtains error estimates for the harmonic interpolation scheme on the unit circle and in the unit disk in both \(L^2\) and maximum norms. Also, the condition number of the associated interpolation matrix is proved to be bounded by a small constant independent of the degree \(n\). The numerical experiments presented confirm the accuracy and stability predicted by the theoretical results.