Compressive Sensing and Sparsity: Theory and Applications in Tomography

  • Alessandro Perelli (Invited speaker)

Activity: Participating in or organising an event typesParticipation in workshop, seminar, course


Compressive Computed Tomography Image Reconstruction with Denoising Message Passing Algorithms Abstract: The compressive reconstruction of images from a limited number of projections is becoming more appealing to reduce the X-ray radiation dose in medical Computed Tomography (CT) or to detect explosives and related materials using dual energy CT scanners but still reconstruction algorithms do not achieve high diagnostic performances. In this talk we present a preconditioned variant of the denoising approximate message passing (DAMP) algorithm aiming to extend its applicability from the theoretical domain of i.i.d. random sensing matrices to deterministic ones using preconditioning and from sparse signal models to generically structured ones. DAMP is an iterative algorithm that at each iteration linearly estimates the conditional probability of the image given the measurements and employs a non-linear denoising function which implicitly imposes an image prior, i.e. incorporates many different structured signal models. The ability to use the different denoisers (e.g. wavelets, total variation, BM3D) gives DAMP the flexibility not exploited in other reconstruction algorithms. The proposed DAMP approach has been tested on simulated CT baggage and the reconstruction results with reduced number of views are promising compared to traditional filtered back projection and to a state-of-the-art model based iterative reconstruction method.
Period12 Nov 201513 Nov 2015
Event typeWorkshop
LocationManchester, United KingdomShow on map


  • Computed Tomography
  • Inverse Problems
  • Approximate Message Passing