绝对差值排序的全变分低剂量CT重建算法

Total Variation Algorithm Based on Rank-Ordered Absolute Differences for Low-lose CT Reconstruction

针对传统全变分(Total Variation,TV)低剂量CT(Computed Tomography,CT)算法在重建图像时出现阶梯效应和模糊图像边缘的问题,提出了基于绝对差值排序(Rank-Ordered Absolute Differences,ROAD)和中值先验(Median Prior,MP)模型的TV低剂量CT重建算法.首先采用ROAD对传统TV模型中的扩散函数进行改进,然后将改进后的TV模型与MP模型结合得到新的惩罚项,最后将该惩罚项应用于基于惩罚加权最小二乘重建算法从而构造新的目标函数.实验结果表明,新算法不仅可以抑制阶梯伪影的产生,还能够很好地保留图像边缘细节信息.

A new total variation algorithm based od and the median prior （MP） model for lowon the rank-ordered absolute differences （ROAD） methlose CT reconstruction was proposed to overcome the drawback of the traditional total variation （TV） algorithm, which can result in the staircase effect and blur the image edge for the low-dose computed tomography （CT）. The ROAD was applied to improve the diffusion function of the traditional TV model at first. Then, combining with this TV model, a new penalty item was formulated with the improved MP model. At last, this new item was applied to estab- lish the new objective {unction based on the penalized weighted least square reconstruction Mgorithm. The experiment results illustrate that the new algorithm can reducing staircase artifacts, while can well preserve image details and edges.