Denoising and Resource Allocation in Uncoded Video Transmission

  • Hao Cui ,
  • Ruiqin Xiong ,
  • ,
  • Zhihai Song ,
  • Feng Wu

IEEE Journal of Selected Topics in Signal Processing | , Vol 9: pp. 102-112

Publication

The recently revitalized uncoded transmission has shown great potential in handling channel variations and user heterogeneities for wireless video communications. In this paper, we propose to adopt motion-compensated temporal filtering (MCTF) at the sender and denoising techniques at the receiver to fully exploit the temporal and spatial redundancy in video signals and therefore significantly improve the efficiency of uncoded video transmission. Although there are readily applicable MCTF and denoising techniques, integrating them into the uncoded video transmission framework is nontrivial, especially in the case of denoising. To address these challenges, we first propose spatial-domain transmission and cascaded denoising to harness the potential of denoising techniques. Second, we develop novel resource allocation algorithms based on variable-size L-shaped chunk division, which are not only compatible with spatial-domain transmission, but also achieves better energy efficiency than existing schemes based on fixed-size rectangular chunks. Our experimental results show that when transmitting 720p videos, our scheme achieves up to a 3.3 dB gain in video PSNR over the state-of-the-art uncoded transmission scheme SoftCast and up to a 5.3 dB gain over a digital scheme based on robust rate adaptation and H.264 scalable video coding (SVC).