Department of Electronics and Communication Engineering, National Institute of Technology (NIT), Tiruchirappalli, India; Department of Electrical and Computer Engineering, University of Saskatchewan, Saskatoon, Canada.
Comput Med Imaging Graph. 2013 Oct-Dec;37(7-8):419-29. doi: 10.1016/j.compmedimag.2013.08.006. Epub 2013 Sep 7.
Micro-computed tomography (micro-CT) plays an important role in pre-clinical imaging. The radiation from micro-CT can result in excess radiation exposure to the specimen under test, hence the reduction of radiation from micro-CT is essential. The proposed research focused on analyzing and testing an alternating direction augmented Lagrangian (ADAL) algorithm to recover images from random projections using total variation (TV) regularization. The use of TV regularization in compressed sensing problems makes the recovered image quality sharper by preserving the edges or boundaries more accurately. In this work TV regularization problem is addressed by ADAL which is a variant of the classic augmented Lagrangian method for structured optimization. The per-iteration computational complexity of the algorithm is two fast Fourier transforms, two matrix vector multiplications and a linear time shrinkage operation. Comparison of experimental results indicate that the proposed algorithm is stable, efficient and competitive with the existing algorithms for solving TV regularization problems.
微计算机断层扫描（micro-CT）在临床前成像中发挥着重要作用。micro-CT 的辐射会导致受试样本的辐射过量，因此减少 micro-CT 的辐射至关重要。本研究集中分析和测试交替方向增广拉格朗日（ADAL）算法，以使用全变差（TV）正则化从随机投影中恢复图像。在压缩感知问题中使用 TV 正则化可以通过更准确地保留边缘或边界来使恢复的图像质量更清晰。在这项工作中，ADAL 解决了 TV 正则化问题，ADAL 是一种用于结构优化的经典增广拉格朗日方法的变体。该算法的每次迭代计算复杂度为两个快速傅里叶变换、两个矩阵向量乘法和一个线性时间收缩操作。实验结果的比较表明，所提出的算法是稳定的、高效的，并与解决 TV 正则化问题的现有算法具有竞争力。
