Compressive Computed Tomography Image Reconstruction with Denoising Message Passing Algorithms

Alessandro Perelli, Michael Davies

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

In this work the reconstruction of images from a limited number of projections is considered in order to achieve X-ray radiation dose reduction in Computed Tomography (CT) while achieving high diagnostic performances. The feasibility of message passing Compressive Sensing (CS) imaging algorithms to CT image reconstruction has been studied, aiming to extend the algorithm from its theoretical domain of i.i.d. random matrices. We have proposed a denoising-based Turbo CS algorithm (D-Turbo) and we have extended the application of the denoising approximate message passing (D-AMP) algorithm to partial Radon Projection data, exploiting a generic denoiser in a CS reconstruction algorithm. The proposed CS message passing approaches have been tested on simulated CT data using the BM3D denoiser yielding an improvement in the reconstruction quality compared to existing direct and iterative methods. The results show the effectiveness of using a generic denoiser Turbo CS or message passing algorithm to few projections CT reconstruction.
Original languageEnglish
Title of host publicationProceedings of Workshop on Signal Processing with Adaptive Sparse Structured Representations 2015
Publication statusPublished - 2015
Event6th Workshop on Signal Processing with Adaptive Sparse Structured Representations, July 6-9, 2015, 2015 - Cambridge, United Kingdom
Duration: 6 Jul 20159 Jul 2015

Conference

Conference6th Workshop on Signal Processing with Adaptive Sparse Structured Representations, July 6-9, 2015, 2015
Country/TerritoryUnited Kingdom
CityCambridge
Period6/07/159/07/15

Keywords

  • Computed tomography
  • Approximate message passing

Fingerprint

Dive into the research topics of 'Compressive Computed Tomography Image Reconstruction with Denoising Message Passing Algorithms'. Together they form a unique fingerprint.

Cite this