Correlation based video processing in video sensor networks

While conventional sensor networks have revealed their vast potentials, simple and special purposed data doubtlessly limits their expansion into information hungry domains. Video sensor networks present an unprecedented opportunity to annex rich information with distributed sensor networks. However, directional sensing range and high data rate unique to video sensors also demands novel solutions, especially for reducing data amount and individual sensor's workload. Toward this end, this paper proposes a method for cooperative video processing in video sensor networks bused on sensor correlations. By pairing up highly correlated nodes and divide sensing tasks among them, each sensor only needs to cover a fraction of the targeted area. Compared to the non-cooperative approach, video sensor's image processing and transmission workload is significantly reduced. We design efficient algorithms to calculate the correlations, partition sensing areas, and fuse partial images. A set of experiments are also performed to prove the concept.