Adaptive image sequence reduction in surveillance using region enhancement block compressive sensing

Due to huge amount of visual surveillance data, network congestion and latency is serious problem for surveillance applications. To reduce the data amount, data compression must be done before transmission. Compressive sensing is a new theory that sample and compress signals at the same time with little cost of bandwidth and energy for transmission. While conventional compressive sensing requires huge memory to compress whole image at a time, block compressive sensing is introduced which compress image in block manners. Based on that, region enhancement block compressive sensing is proposed to adaptive compress image sequence and enhance the quality of region of interest with reduced data transmission. In proposed scheme, a target detecting algorithm is used to select the important blocks and more number of measurements is allocated on them. Thus the transmission amount is adaptively adjusted with the target in or out. The proposed scheme is evaluated in practical surveillance data and the result demonstrates that it adjusts the measurement ratio as needed and the quality of region of interest is enhanced at reduced measurement ratio.