Research On Compressive Sensing Application In Video Compression

Compressive Sensing (CS) based on the revelation that a sparse signal can be reconstructed by a small number of linear projections. By employing CS, the data can be compressed at a rate significantly below the Nyquist rate and perform as if it were possible to directly acquire the important information about the signals. This enables dramatically reduced measurement time and dramatically reduced resources in practice.Distributed video coding (DVC) has recently been proposed to reduce the complexity of the encoder, whereas it suffers from the sampling cost of huge amount of image data. To relax such sampling burden, this paper develops a novel sub-sampling distributed video coding by utilizing compressive sensing technique. Due to the inherent sparsity in video sources, the video frames are compressively sampled at the encoder. On the other hand, by exploiting the correlation between CS measurements and side information and by performing sparsity recovery, the video frames are recovered at the decoder. When compared with the traditional fully-sampling equivalence, the new framework enjoys the reduction of transmission rate, the reduction of implementation complexity and the robustness to channel losses, which are verified in the simulations.Distributed compressive sensing (DCS) extends CS to the application of recovering a sparse signal ensemble. Consider a number of correlated signals with each individually sparse in a certain basis. In DCS, by exploiting both inter-signal and intra-signal correlation structures, these signals are measured via a CS technique independently and reconstructed jointly at a collection point in the receiver. Numerical results show that compared with separate CS recovery, DCS could achieve a further reduction of measurements.