An Improved Distributed Compressive Video Sensing Based on Adaptive Sparse Basis

This paper proposes an improved Distributed Compressive Video Sensing (DCVS) framework based on adaptive sparse basis, which integrates the recently emerging Distributed Video Coding (DVC) and Compressive Sensing (CS) theory. The proposed framework incorporates a low-complexity encoder and shifts most computation burden to the decoder-side. At the encoder, the video frames are sampled independently. However, the decoder recovers each block in a frame jointly using its side information (SI) and the state-of-the-art sparse basis generated by a few temporal neighboring blocks in previously reconstructed preceding and/or following key-frames. Experimental results show that the proposed framework outperforms not only the intra-encoding and intra-decoding scheme but also the DVCS scheme with wavelet transform based sparse basis.