INFDok - Dokumentenvolltextserver des Fachbereichs Informatik

Kurzfassung auf Englisch:

Functional programming languages, like Lisp or Racket are known to be general purpose languages with a steep learning curve and wide range of applications. They can be used interactively to solve problems and have inspired other comparably new languages with respect to functional extensions (e.g. Python or Swift). However, at least the interpreted functional languages, have a poor reputation when it comes to execution speed. In this paper, we will prove that this reputation is either wrong or at least not applicable for the field of computer vision and the language Racket. We build upon the VIGRACKET module which combines the best of the compiled and interactive worlds with respect to common tasks in computer vision, Racket and the VIGRA C++ library. We will present the current state of the module and analyze the execution speed for the most important operation types. After the analysis we present ways to increase the processing speed and optimize the module accordingly. To demonstrate the use of the optimized library, we connect Racket to the Microsoft Kinect Sensor and present two case studies: a blue-screen simulation, and an interactive natural pointing device interface.