Quantitative invertibility and approximation for the truncated Hilbert and Riesz transforms (Q2280517): Difference between revisions

The article concerns the task of reconstructing a given function on a bounded Lipschitz domain \(\Omega_1\subset\mathbb{R}^n\) from its truncated Hilbert or truncated Riesz transform or a perturbation thereof, given on a bounded Lipschitz domain \(\Omega_2\subset\mathbb{R}^n\) with \(\bar\Omega_1\cap\bar \Omega_2=\emptyset\). This problem is strongly ill-posed in \(L^2\), so a stable inversion is not possible. The work provides conditional stability estimates for the case that the input data belongs to \(H^1(\Omega_1)\), implying in particular that the problem is stable if the input data does not oscillate too much. Moreover, an estimate of the \(L^2(\Omega_1)^n\) norm of an \(\varepsilon\)-optimal reconstruction is established, where by \(\varepsilon\)-optimal reconstruction we mean a function \(f=f(h)\in L^2(\Omega_1)^n\), \(h\in L^2(\Omega_2)\), such that \(\lVert T(f)-h\rVert_{L^2(\Omega_2)}\leq\varepsilon\). Here, \(T(f)\) stands for the truncated Hilbert transform or the sum \(\sum_{j=1}^n R_j f_j\), where \(R_j\) is the \(j\)-th Riesz transform. The viewpoint of this work on the truncated transforms is that of harmonic extension. This allows to invoke PDE tools and produces results that are more general than those previously available. The fact that previous contributions based on singular values often yield better constants in the estimates is clearly mentioned.