Power-Efficient Video Uploading for Crowdsourced Multi-View Video Streaming
Than Than Nu, Takuya Fujihashi, Takashi Watanabe · 2018
Exploiting the inter-camera correlation among the crowdsourced mobile video contributors is desirable as it can improve the coding efficiency and/or video quality. This can be achieved by using a digital video encoder such as H.264/Multi-view Video Coding (MVC). However, highly-complex nature of digitalbased video encoding requires a large amount of power, which is a limited resource of the consumer-grade mobile devices. In addition to video encoding, transmission power is also consumed for video uploading based on the amount of uploading video traffic to satisfy video quality requirement. In this paper, we propose a powerefficient video uploading scheme for crowdsourced multi- view video streaming considering power-saving in both video encoding and video transmission. To this end, we design a cluster-based redirect transmission scheme. In this scheme, video frames of multiple contributors are only encoded by four-dimensional discrete cosine transform (4D-DCT) and modulated by near-analog modulation at a cluster head. By removing digital-based coding and modulation, our scheme achieves the same video quality with the existing scheme under low power requirements. Evaluation results show that the proposed scheme outperforms the digital-based video uploading schemes in terms of video quality and the amount of video traffic to achieve a targeted video quality.