Towards OpenVL: Improving Real-Time Performance of Computer Vision Applications

Abstract

Meeting constraints for real-time performance is a main issue for computer vision, especially for embedded computer vision systems. This chapter presents our progress on our open vision library (OpenVL), a novel software architecture to address efficiency through facilitating hardware acceleration, reusability, and scalability for computer vision systems. A logical image understanding pipeline is introduced to allow parallel processing. We also discuss progress on our middleware—vision library utility toolkit (VLUT)—that enables applications to operate transparently over a heterogeneous collection of hardware implementations. OpenVL works as a state machine,with an event-driven mechanismto provide users with application-level interaction. Various explicit or implicit synchronization and communication methods are supported among distributed processes in the logical pipelines. The intent of OpenVL is to allow users to quickly and easily recover useful information from multiple scenes, in a cross-platform, cross-language manner across various software environments and hardware platforms. To validate the critical underlying concepts of OpenVL, a human tracking system and a local positioning system are implemented and described. The novel architecture separates the specification of algorithmic details from the underlying implementation, allowing for different components to be implemented on an embedded system without recompiling code.