Reconstruction of CT Images Based on Cross-Entropy Method

This paper describes the system models that are developed for multi-phase flow CT through the use of the Geant4 simulating software toolkit. Models with five, seven and nine radiation sources are reconstructed. In order to improve the resolution of the CT images reconstructed using conventional algorithms, a cross-entropy minimization technique together with an information extension method is proposed. A minimum cross-entropy (MCE) method - SMART is presented. The choices of weighting factors and prior information are discussed. This method improves the spatial resolution and reduces artifacts effectively. Both simulation and experimental results show that the SMART method improves not only the quality of reconstructed images but also the real time performance.