Industrial X-ray Tomographic Imaging Application of Simultaneous Reconstruction Technique Using GPU

In Computed Tomography (CT), reconstruction of images is a computationally complex process. It is further complicated by implementing iterative reconstruction algorithms for data acquired with high resolution 2D area detectors. For processing such dense CT data with iterative methods, high-performance accelerating tools like General Purpose Graphic Processing Unit (GPGPU) is essential. This paper presents an industrial X-ray tomographic imaging application of Simultaneous Iterative Reconstruction Technique (SIRT) using GPU. A volume of 2k × 2k × 360 pixels has been reconstructed using NVIDIA Quadro P4000 1.2 GHz GPU and time reduction of about 25× has been achieved for GPU-SIRT implementation. Experimental results on image quality and performance of GPU-SIRT for the specific industrial specimen are also presented.