期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2014
卷号：70
期号：1
出版社：Journal of Theoretical and Applied
摘要：The current computer vision-based systems (CVS) are becoming computationally demanding due to the improvement of their functionalities that is difficult to be achieved with single-core frameworks. Such deficiencies of the single-core frameworks have led to the introduction multi-core frameworks to meet the required performance of their functionalities. However, in order to obtain good performance for CVS on multi-core frameworks, it is crucial to utilise parallelism tools efficiently. These parallelism tools need to be utilised on hotspots (most time-consuming functions in algorithm) in order to minimise development time and to reduce application development costs. This is a challenging task and requires an in-depth investigation of multi-core frameworks. This research work investigates the utilisations of multi-core frameworks capability for a real-time object tracking application problem using a parallel software tool known as Intel� Parallel Studio XE tool. In the investigation, two established multi-core frameworks, namely, Threading Building Blocks (TBB) and Open Multi-Processing (OpenMP) were implemented at identified hotspot functions of the tracking algorithm. The performances of these two multi-core frameworks were then evaluated and compared based on computed speedup, efficiency and scalability. The results from this investigation demonstrated that the processing time of real-time object tracking was improved by using hotspots identification. In addition to that, multi-core frameworks could make the tracking algorithm explicitly faster when compared to single-core frameworks and OpenMP outperformed TBB.
关键词：Multi-core Frameworks; Parallel Programming; Image Processing; Real-time Object Tracking; OpenMP; Threading Building Blocks (TBB); Intel� Parallel Studio XE
