Interactive Media Systems, TU Vienna

A Testing Environment for Video-Based Multiview Computer Vision Research

By Efstathios Stavrakis and Margrit Gelautz


We describe the methodology, design considerations and practical implementation of an environment used in developing and testing video-based, multiview computer vision algorithms. The environment we have built is composed by a variety of heterogeneous hardware devices for acquisition and display, connected together into a coherent system by object-oriented software components. Much attention has been paid into the development of a Vision-Device Abstraction Layer (VDAL) that handles the operation of hardware image acquisition devices. The software components are divided into User Interface and Low-Level computer vision constituents. These are combined into more complex entities, Modules, that perform a complete computer vision process. Our environment is capable of providing output on disk, in real-time on common monitors, but can also take advantage of large stereo displays. It is designed with modularity, extensibility and reusability in mind. The main contribution of this paper is to describe the setup and efficient implementation of a coherent environment for video-based multiview applications and algorithms’ development. We provide our experience of how a variety of different hardware and software has been glued together to form the basis of our computer vision processing environment.


Reference currently not available.