Soft Mobile Video Broadcast Based On Compressive Sensing

In recent years, with the popularization of mobile terminal, more and morepeople like to use the wireless network to watch the video. In this case,couldthe existing wireless communication technology and wireless video encodermeet the increasing demand of user? The wireless communication needs thescalability and robustness to satisfy the different users. The video encoder is thepursuit of the compression efficiency and reconstructed video quality. It cansave bandwidth when broadcasting a video signal to multiple users, but thereare three different points:1) unlike wired video transmission, different usershave different channel conditions.2) In wireless video transmission thesignal is easily affected by the noise.3) It is not effective in the videocompressibility. In the traditional wireless video coding, the scalable videocoding can provide several basic layer and enhancement layer. User can choosethe number of layers to decode according to their channel condition. However,the scalable video coding may produce cliff effect.With the development of wireless video technology, people proposed softbroadcast, two of them are Softcast and DCAST. The basic idea of Softcast isthe linear relationship between transmission signal and pixel values. Thisensures that users can decode different quality of video according to theirchannel condition. All operations of Softcast are linear, it use a new codecwhich combines the video compression and error protection. Softcast uses2D-DCT transform to remove the redundancy in one frame and uses3D-DCTto remove the redundancy between frames, but its compression efficiency isnot high. DCAST, using distributed source coding technology and softbroadcast technology, add the traditional inter prediction technology, butthe compression rate and video reconstruction quality can be better. In thiswork, we propose a soft video broadcast based on compressive sensing.Compressed sensing is applied to generate measurements of equal importancefrom a video such that a receiver with a better channel will naturally have moreinformation at hands to reconstruct the content without penalizing others, so it can achieve the broadcasting also can improve the compression rate. We utilizethe nonlocal self-similarity of intra-frame and the similarity of video frames toimprove the quality of reconstructed video. We compare our results withSoftcast and DCAST, and find that compression efficiency and videoreconstruction quality are better.