Abstract: Time resolution is one of important properties of ICT techniques.It is a effective methods of impoving speed of reconstruction to adopt parallel algorithm;Based on Beowulf cluster，we have presented a parallel implementation of filter back-projection and presented an accelerated algorithm based on Intel Pentium SIMD technologies.On cluster flat constructed with four P4/2.9 G PC we have maken experiment of reconstruction using four groups of data collected by industrial CT collection.The results of experiment indicate that the reconstruct speed can be improved 4-6 times by using SIMD technique，accelerated 1.5-3 times by using cluster parallel technique，and accelerated 8-10 times by using these methods synthetically.

Key words: industrial CT, image reconstruction, filtering back-projection, SIMD, cluster computing

摘要： 工业CT图像的重建速度是工业CT产品的一个重要指标。使用并行算法是提高重建速度的一个行之有效的方法。提出了基于Beowulf集群系统的滤波反投影算法的并行实现方法;并且提出了基于Intel 奔腾SIMD技术的加速算法。在用4台P4/2.9 G微机构建的集群系统平台上对工业CT采集的4个不同的断层投影数据进行重建实验，实验数据表明使用SIMD技术可以得到4-6倍的加速，使用集群并行技术的算法能够得到1.5-3倍的加速，综合应用这两项技术可以得到8-10倍的加速。

关键词: 工业CT, 图像重建, 滤波反投影, SIMD, 集群计算