The emergence of visual big data is a doubleedged sword to mobile communications. On one hand, the massive scale of visual data transmission brings a huge challenge to the RAN. On the other, the abundant information in visual big data may improve the spectrum efficiency and robustness of visual communications. In this article, we propose a DaC-RAN architecture for visual communications, which integrates the ideas of SDN and C-RAN.

We propose to separate the control and data planes in the conventional infrastructure, and integrate a new data plane specifically designed for visual communications into the virtual base station. We demonstrate the proposed DaC-RAN architecture through a practical visual communication system based on pseudo-analog transmission and CSbased data sampling and reconstruction. The correlation information retrieved from visual big data is utilized as prior knowledge in CS decoding. Preliminary simulation evaluations show that significant gain is achieved in spectrum efficiency over the conventional method.