Belief Propagation Based Compressed Video Streaming in Wireless Multimedia Sensor Networks

Compressed sensing is a novel technology to acquire and reconstruct sparse signals below the Nyquist rate, and thus has great potential for video streaming in wireless multimedia sensor networks (WMSN) due to its significant reduction of computational complexity. In this paper, we firstly propose a Markov chain model to characterize variations of the video sequences´ temporal correlation. To estimate the transition matrix, we use the measurement residual to reflect the intensity changes and then calculate the transition property. Lastly, a compressed video sensing framework for wireless networks is presented using the belief propagation algorithm and the proposed Markov model. Numerical results show that our proposal could efficiently utilize the temporal correlation in video, and obtains improved reconstruction quality in comparison with conventional methods.