A Survey of Bottom-Up Visual Saliency Methods in Wireless Multimedia Sensor Networks

The implementation of visual saliency in wireless multimedia sensor networks (WMSNs) can be useful for a large application scope. Autonomous detection and tracking based on human gaze behavior as well as efficient bandwidth utilization for sensing devices in WMSNs can be achieved through such a method. In this paper, a review on two major categories of visual saliency model is presented. Available models are analyzed in terms of resource efficiency. After an evaluation is performed, it is found that a purely computational pixel-based saliency model is most suitable to be used in WMSNs due to its low computational complexity and memory requirements, as well as efficient detection of conspicuous objects.