Multi-phase joint reconstruction framework for multi-view video compression using block-based compressive sensing

The video captured by different visual sensor in a visual sensor network is first compressed using the block-based compressive sensing algorithm. All the videos are encoded independently at different sub-rates and transmitted to a host workstation for reconstruction. Then, the proposed multi-phase joint reconstruction framework is applied to improve the reconstruction of lower subrate videos. In this case, frames extracted from higher subrate videos are used to produce some side information, which serve as prediction of the counterpart frames in lower subrate videos. Next, the difference between them at the measurement level is calculated. The difference is then added to the prediction to obtain the final reconstruction. The experimental results show that the proposed framework is able to outperform the other frameworks on average by a margin of 1dB to 2dB at different subrates on various multi-view videos.