一种TV模型下基于ADMM的加速算法

An accelerating algorithm based on ADMM under TV model

如何利用不充分的投影数据进行精确的图像重建是当前电子计算机断层扫描(Computed Tomography,CT)领域的研究热点,受辐射剂量、人体自身条件、重建时间等诸多因素的限制,投影数据不会总是完整的,而医学采样过程中,如果能减少采样数目,无疑大大减少了X射线对病患和医生的伤害,随着压缩感知(Compressed sensing,CS)理论的提出,稀疏角度重建已经得到了较为广泛的应用。全变分(Total Variation,TV)算法是稀疏角度重建的一种有效方法,文章在TV算法和ADMM算法的基础上,提出了一种新的加速收敛算法,该算法先进行ADMM-TV算法迭代,然后在每轮迭代后对图像进行调整,有效加快收敛速度。相关的实验结果表明,该加速算法能够在相同的迭代次数下获得精确度更高的图像,在加速收敛方面效果良好。

Now,the accurate image reconstruction from few-view projections is a hot point in the study of Computed Tomography(CT).The projections is not always full caused of the limit of many factors,such as radiation dose, the body’s own condition s and so on.if we can reduce the number of sampling in medicine, there is no doubt that it will greatly reduces the injury of Xray to the patient and doctor.with the propose in Compressed sensing,the sparse optimization has recently been wide applied.Total Variation(TV) algorithm is an effective way to sparse reconstruction, based on che TV algorithm and ADMM algorithm,a new accelerated convergence algorithm is proposed, the algorithm firstly iterates the ADMM-TV algorithm, and then adjusts the image after each iteration to accelerate the convergence speed effectively.The experimental results show that the algorithm can obtain more accurate images with the same number of iterations, and is effective in accelerating convergence.