Reconstruction of CT Images Using Iterative Least-Squares Methods with Nonnegative Constraint
Open Access
- 15 March 2019
- journal article
- Published by Research Institute of Signal Processing, Japan in Journal of Signal Processing
- Vol. 23 (2) , 41-48
Abstract
The basic mathematical problem in computed tomography (CT) is defined as finding a nonnegative solution to a rectangular linear system with a sparse coefficient vector and matrix. The coefficient vector and matrix respectively correspond to acquired projection data and a projection operator based on a discrete Radon transform. As an approach to find a nonnegative solution corresponding to a tomographic image, a continuous method using simultaneous ordinary differential equations has been proposed. The continuous method can produce high-quality images even from an insufficient number of projection data. However, it requires a huge computational cost to obtain a high-quality image because of numerical integration. In this paper, to reduce the image-reconstruction time, we proposed two iterative methods without numerical integration that enable us to reconstruct high-quality CT images without negative pixels. The methods were produced on the basis of a hierarchical alternating least-squares algorithm that is known as a fast solver for nonnegative matrix factorization and conjugate gradient methods that are effective for a linear system. Through numerical experiments, we also discussed the performances of the proposed methods in terms of both the quality of obtained images and the time required to obtain high-quality images.Keywords
This publication has 14 references indexed in Scilit:
- Global convergence of a modified HALS algorithm for nonnegative matrix factorization2015
- Algorithms for nonnegative matrix and tensor factorizations: a unified view based on block coordinate descent frameworkJournal of Global Optimization, 2013
- Common Lyapunov Function Based on Kullback–Leibler Divergence for a Switched Nonlinear SystemMathematical Problems in Engineering, 2011
- Continuous-time image reconstruction using differential equations for computed tomographyCommunications in Nonlinear Science and Numerical Simulation, 2010
- Fast Local Algorithms for Large Scale Nonnegative Matrix and Tensor FactorizationsIEICE Transactions on Fundamentals of Electronics, Communications and Computer Sciences, 2009
- Hierarchical ALS Algorithms for Nonnegative Matrix and 3D Tensor Factorization2007
- Learning the parts of objects by non-negative matrix factorizationNature, 1999
- The MATLAB ODE SuiteSIAM Journal on Scientific Computing, 1997
- Positive matrix factorization: A non‐negative factor model with optimal utilization of error estimates of data valuesEnvironmetrics, 1994
- Methods of conjugate gradients for solving linear systemsJournal of Research of the National Bureau of Standards, 1952