A study on the implementation of networked image-sensing by using smart image sensors

Large-scale image sensing over the extended area and time is very effective to obtain global information, which can be attained by the local information from the spatially-distributed sensors for long-term period. However, on the other hand, how to overcome the bandwidth problem caused by huge data volume is one of the crucial issues to implement a practical system. From the viewpoint of communication bandwidth, we consider that it would be Very cost-effective if the necessary data only can be acquired and transmitted at necessary time. In this paper, an experimental system using a group of Spatially-Variant Sampling (SVS) smart image sensors is presented. This system is targeted to reduce the data volume at image acquisition level by utilizing selective spatial/temporal resolution reduction schemes, taking into considerations of the priority of particular regions of the scenes. The main advantage of this system is the pixel-level handling of images on sensor nodes and a host which contributes to the better efficiency and more flexibility in data management, because this system has no need for encoding/decoding of the image data, which is normally very computationally-intensive when the large number of image sensors are involved.