Exploiting global redundancy in big surveillance video data for efficient coding

In the era of big data, the gap between the fast increasing size of the surveillance video data and the relatively stable video compression rate has become prominent. Existing data compression methods based on the known redundancies are not able to eliminate the dominant redundancy embedded in the big surveillance video data (BSVD) of multiple sources in large time span. In this study, a new approach based on global redundancy of the BSVD has been explored. We first analyze the compositions of the global redundancy based on characteristics of BSVD. A coding framework is proposed to eliminate global redundancy. Simulated experiments have been performed to examine the performance of the approach in comparison with the high efficiency video coding. The experimental result showed that the proposed approach can reach a compression rate of 1/400 for a huge dataset of surveillance videos.