Multi-Pose Fusion for Sparse-View CT Reconstruction Using Consensus Equilibrium

TL;DR

This paper introduces Multi-Pose Fusion, a novel algorithm leveraging consensus equilibrium to jointly reconstruct CT images from multiple object poses, improving over single-pose methods.

Contribution

It presents a new multi-pose reconstruction algorithm using consensus equilibrium, extending plug-and-play methods for multi-rotation-axis CT data.

Findings

Multi-Pose Fusion outperforms single-pose reconstruction in simulated experiments.

The method effectively integrates data from multiple poses for improved image quality.

Demonstrates the potential for enhanced CT imaging in multi-rotation scenarios.

Abstract

CT imaging works by reconstructing an object of interest from a collection of projections. Traditional methods such as filtered-back projection (FBP) work on projection images acquired around a fixed rotation axis. However, for some CT problems, it is desirable to perform a joint reconstruction from projection data acquired from multiple rotation axes. In this paper, we present Multi-Pose Fusion, a novel algorithm that performs a joint tomographic reconstruction from CT scans acquired from multiple poses of a single object, where each pose has a distinct rotation axis. Our approach uses multi-agent consensus equilibrium (MACE), an extension of plug-and-play, as a framework for integrating projection data from different poses. We apply our method on simulated data and demonstrate that Multi-Pose Fusion can achieve a better reconstruction result than single pose reconstruction.