A sparse representation for efficient bandwidth utilization in wireless surveillance networks

Abstract

We propose a low-cost low-rank-based framework for the operation of wireless surveillance systems. The proposed framework has two modes of operations: an initialization offline mode, in which low-rank terms of few initial frames are recovered using RPCA. Then these recovered low-rank terms are transmitted over the wireless network to the receiver. In the real-time mode of operation, sparse terms of the captured frames are calculated using FRPCA, then transmitted to the receiver. Transmission of only the sparse terms greatly saves the used bandwidth and hence the cost of the transmission process.