A scheme for distributed compressed video sensing based on hypothesis set optimization techniques

Multi-hypothesis prediction technique can greatly take advantage of the correlation between the video frames to obtain a high quality performance. In this paper, we propose a scheme for distributed compressive video sensing based on hypothesis set optimization techniques which further enhances the reconstruction quality and reconstruction speed of video compared with existing programs. The innovation in this paper includes four parts: (1) superb hypotheses selection-based hybrid hypothesis prediction technique, which selects the superb hypotheses from the original hypothesis set corresponding to the block to be reconstructed in the video sequence to form a new set, and then implements the hybrid hypothesis prediction (HHP) with the new one; (2) hypothesis set update-based hybrid hypothesis prediction technique, which selects the high quality hypotheses and derives new hypotheses by interpolating, and then replaces the noisy hypotheses with the new ones; (3) advanced hybrid hypothesis prediction technique, which improves the judgment formula of HHP model through averaging the Euclidean distances to each measurement to realize the goal of the adaptive judgment of the HHP model in various sampling rates; (4) adaptive weighted elastic net (AWEN) technique, which combines norm, $$\ell _1$$ℓ1, $$\ell _2$$ℓ2 and then weights both of them with the distance vector to form AWEN penalty term. The simulation results show that our proposal outperforms the start-of-the-art schemes without using the hypothesis set optimization techniques.