Description

3D X-Ray Computed Tomography (3DXCT) is a well-established method in medicine which recently gained importance and acceptance in the inspection and analysis of industrial components. It allows for imaging specimens in three dimensions to detect and measure outer structures as well as hidden errors inside. Thus 3DXCT offers a comprehensive overview of a specimen, which is generated as follows: The specimen is rotated on the rotary plate, taking a projection image at each angular position. The reconstruction algorithm uses the stack of projection images generated from a full rotation to reconstruct the 3D volumetric dataset.
Our goal is to do this reconstruction process on the GPU and to implement a fast and efficient method using high end graphic cards hardware. Concerning computation speed GPUs are outperforming top of the line CPUs and every (half a) year a new generation of GPUs is published doubling the performance of its predecessors. However GPUs also face drawbacks to overcome, e.g. their limited memory. But as GPUs are getting increasingly flexible in terms of programming and also the equipped memory is sufficient for individual tasks, the interest in porting common algorithms to fast and comparably lower cost GPU hardware is increasing.

Task

Your task will be the development of a novel GPU based reconstruction method of industrial CT data. You will learn how to utilize and program modern GPU hardware and you will implement a tool to handle and reconstruct industrial CT data. You will work in a creative environment and you are welcome to contribute your own ideas, rather than only implementing an algorithm or technique from a paper.

Requirements

Basic knowledge in computer graphics and visualization, C++ programming, experience in GPU programming is advantageous.

Additional Information

• 3DCT
• NVIDIA
• FH OOE – Wels Campus
• Principles of Computerized Imaging