A video complexity index for cluster computing

Abstract

Entertainment applications that work with images and videos processing use the Cluster computing to decrease their rendering time. A complexity level can be experimentally determined according to the time spent to render a video. This paper study the time spent for rendering 3D videos on a Beowulf cluster computing with PovRay software on a GNU/Linux environment. For another hand, the video complexity is determined by a novel metric named Video Complexity Index (VCI) that considers both, the spatial and temporal video characteristics, the metric is tested with videos of different complexity. Experimental results with the cluster number increasing demonstrate that VCI metric successfully classifies the video complexity according to the number of nodes used to render them, considering the processing time consumed for a cluster. So, VCI metric can be very useful to find the minimal number of nodes depending of the 3D video complexity