Variational regularization of the weighted conical Radon transform

TL;DR

This paper introduces a novel variational regularization method for the weighted conical Radon transform, demonstrating improved reconstruction quality in imaging applications like emission tomography with Compton cameras.

Contribution

It establishes uniqueness of inversion for the weighted conical Radon transform with general weights and develops a stable variational regularization framework with numerical algorithms.

Findings

TV-regularization outperforms quadratic Tikhonov in reconstruction quality.

Numerical schemes based on Chambolle-Pock algorithm effectively implement the regularization.

Uniqueness of inversion is proven for the weighted conical Radon transform.

Abstract

Recovering a function from integrals over conical surfaces recently got significant interest. It is relevant for emission tomography with Compton cameras and other imaging applications. In this paper, we consider the weighted conical Radon transform with vertices on the sphere. Opposed to previous works on conical Radon transform, we allow a general weight depending on the distance of the integration point from the vertex. As first main result, we show uniqueness of inversion for that transform. To stably invert the weighted conical Radon transform, we use general convex variational regularization. We present numerical minimization schemes based on the Chambolle-Pock primal dual algorithm. Within this framework, we compare various regularization terms, including non-negativity constraints, $H^1$-regularization and total variation regularization. Compared to standard quadratic Tikhonov regularization, TV-regularization is demonstrated to significantly increase the reconstruction quality from conical Radon data.