Al-Rifaie Mohammad Majid, Blackwell Tim
Department of Computing, Goldsmiths, University of London, London, SE14 6NW UK.
Evol Intell. 2016;9(3):67-79. doi: 10.1007/s12065-016-0140-7. Epub 2016 Aug 8.
This paper extends particle aggregate reconstruction technique (PART), a reconstruction algorithm for binary tomography based on the movement of particles. PART supposes that pixel values are particles, and that particles diffuse through the image, staying together in regions of uniform pixel value known as aggregates. In this work, a variation of this algorithm is proposed and a focus is placed on reducing the number of projections and whether this impacts the reconstruction of images. The algorithm is tested on three phantoms of varying sizes and numbers of forward projections and compared to filtered back projection, a random search algorithm and to SART, a standard algebraic reconstruction method. It is shown that the proposed algorithm outperforms the aforementioned algorithms on small numbers of projections. This potentially makes the algorithm attractive in scenarios where collecting less projection data are inevitable.
本文扩展了粒子聚集体重建技术（PART），这是一种基于粒子运动的二值断层扫描重建算法。PART假设像素值为粒子，粒子在图像中扩散，并在均匀像素值区域聚集在一起，这些区域称为聚集体。在这项工作中，提出了该算法的一种变体，并重点关注减少投影数量以及这是否会影响图像重建。该算法在三个不同尺寸和前向投影数量的体模上进行了测试，并与滤波反投影、随机搜索算法以及标准代数重建方法SART进行了比较。结果表明，所提出的算法在少量投影情况下优于上述算法。这可能使该算法在不可避免地收集较少投影数据的场景中具有吸引力。
