The Hough transform is an algorithm used in image processing to detect special shapes on an image. Originally it was developed for detecting lines, but using generalised Hough transform it is possible to find arbitrary objects described with their models. The algorithm maps the desired properties to a parameter space, where it tries to find the best fitting points by a voting procedure. The results are transformed back to the original images domain.
OpenCL is a framework for writing programs that execute across heterogeneous platforms consisting of central processing units, graphics processing units and other processors. By using OpenCL any program can access the resources of the platforms and they can be used for general purpose computing.
The thesis aims to investigate the possibilities of improving Hough transform's performance by using the parallelization capabilities of OpenCL. The designed algorithm will be implemented in C++ and OpenCL languages, it will be benchmarked and the results will be evaluated.