Stability estimates for the regularized inversion of the truncated Hilbert transform
Let \(a_1<a_2<a_3<a_4\) be real numbers, and \(H_T:L^2(a_2,a_4)\rightarrow L^2(a_1,a_3)\) be the truncated Hilbert transform. Employing the spectral theory of the truncated Hilbert transform, the authors study the following inverse problem: Let \(f_{ex}\) be an unknown object of interest, for which a noisy measurement \(g\) is given such that \(\|g-H_Tf_{ex}\|_{L^2(a_1,a_3)}\leq\delta,\) for some noise level \(\delta>0\). Is it possible to construct \(f_{ex}\) in some interval of interest \((a_2,a_3-\mu),\) for some small \(\mu>0\)? Under several conditions, especially the boundedness of \(\|f_{ex}\|_{L^2(a_2,a_4)},\) it is proved that any algorithm to compute a solution in the set of admissible solutions is a regularization method for the reconstruction on \((a_2,a_3-\mu).\) Then, an algorithm to find \(f_{ex}\) is introduced based on a truncated singular value decomposition. Stability, convergence, error analysis as well as Tikhonov regularization are presented, with Hölder-type continuity. Similar results are derived if the condition of the boundedness of \(\|f_{ex}\|_{L^2(a_2,a_4)}\) is replaced by knowing an upper bound for the variation of \(f_{ex}\) in addition to reasonable boundary conditions on \(f_{ex}\). The technique of the paper is applied to derive logarithmic-type stability for the algorithm for solutions throughout the entire interval \((a_2,a_4),\) which is a known result as the authors indicate.
This page was built for publication: Stability estimates for the regularized inversion of the truncated Hilbert transform
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6829806)
