基于Deflation技术的预调制Restarted GMRES算法

Preconditioned Restarted GMRES Algorithm Based on Deflation Technique

电学层析成像的图像重建需要对逆问题进行求解,而求解过程中存在着非线性、欠定性以及病态性严重等难题,使得图像重建可能不收敛,或者致使收敛,但获得的图像分辨率较低。针对现有的一些图像重建算法,提出基于Deflation技术的预调制Restarted GMRES算法,在原有full GMRES算法基础上,提高了收敛速度以及图像成像分辨率,并通过仿真实验证明。

The image reconstruction of Electrical Tomography needs to solve inverse problem which is non-linear and ill-conditioned. This may make the image reconstruetion diverge or even if converge, the resolution is very low. In this paper, several commonly used image reconstruction algorithms are introduced. Also, based on the （;MRES algorithm ,this paper proposes a preconditioned restarted GMRES algorithm based on deflation technique to speed up convergence and improve imaging quality. Simulations and experiments show that it is effective to be applied in industrial two-phase measurement.