基于PICCS的非局部TGV能谱CT重建算法

Nonlocal TGV Rreconstruction Algorithm based on PICCS for Spectral CT

能谱CT将宽谱划分为窄谱,导致通道内光子数目明显减少,加大了噪声影响,故从噪声投影中重建出高质量图像是能谱CT的一个研究热点.传统全变分(total variational,TV)容易造成重建图像中出现块状伪影等问题,总广义全变分(total generalized variation,TGV)算法可以逼近任意阶函数,再结合非局部均值算法的思想,同时考虑到不同能谱通道下重建图像的相关性,将高质量全能谱重建图像作为先验图像指导能谱CT重建,提出了基于先验图像约束压缩感知(prior image constrained compressed sensing,PICCS)的非局部TGV重建算法.实验结果表明,所提算法在抑制噪声的同时能够有效复原图像细节及边缘信息,且收敛速度快.

Energy spectrum CT divides the broad spectrum into narrow spectrum,resulting in a significant reduction in the number of photons in the channeland increases the influence of noise.Therefore,it is imperative to study the reconstruction of high-quality CT images from noise projection.Traditional total variational method is easy to cause the reconstruction images appear in the block artifact,in this paper,combine the ideas of nonlocal means algorithm and total generalized variation(TGV)which could approximate any order functions,using correlation of the reconstructed image under different spectral channels,using high quality full spectrum reconstruction image as a guide for prior image spectrum CT reconstruction,is presented based on the prior image constrained compressed sensing(PICCS)nonlocal TGV reconstruction algorithm.Experimental results show that the proposed algorithm can effectively recover image detail information while suppressing noise and has better convergence.