Compressed uncoded screen content video transmission in bandwidth-constrained wireless networks

Abstract

Recently, uncoded transmission schemes based on SoftCast have aroused great interest of the academia. However, the existing work mainly focuses on the natural camera-captured video. Moreover, the bandwidth compression case of these schemes has not been well studied so far, which usually occurs in bandwidth-constrained wireless environments. In this paper, we focus on these issues by proposing a Compressed Uncoded Video Transmission System (CUVTS) based on Compressed Sensing (CS) toward screen content video applications. The main work of CUVTS is leveraging CS to compress the coefficient blocks with continuous compression ratios, which is more efficient than the “block dropping” strategy in SoftCast. Motion-Compensated Temporal Filtering (MCTF) and Discrete Wavelet Transform (DWT) are firstly applied to remove the data redundancy and promote the data compressibility. Then block-based CS is applied to the coefficient blocks with diverse compression ratios determined by the bandwidth allocation component with block importance considered. Finally, optimal power allocation problem is formulated and solved to minimize the transmission distortion. In various channel SNR and bandwidth budget cases, simulation results show that CUVTS outperforms SoftCast with average of 0.79dB up to 8.33dB gain in Peak Signal-to-Noise Ratio (PSNR).