GPU-Accelerated Computer Vision on the Linux Platform

The GPUCV library accelerates computer vision algorithms using the huge computational power of modern graphics processing units (GPUs). This paper presents a series of alterations to the GPUCV library that enable it to be used natively on the Linux platform and so bring the performance advantages provided by GPUs to computer vision applications running on Linux. The considerations and decisions involved in this process are presented and discussed, and the performance characteristics of the GPUCV library on this new platform are compared to the original Microsoft Windows platform. Finally, ideas for future work on increasing the performance and ease-of-use of GPU accelerated computer vision development are put forward.