An elastic net-based hybrid hypothesis method for compressed video sensing

• Published in: Multimedia tools and applications Vol. 74; no. 6; pp. 2085 - 2108
• Main Authors: Chen, Jian, Chen, Yunzheng, Qin, Dong, Kuo, Yonghong
• Format: Journal Article
• Language: English
• Published: Boston Springer US 01.03.2015; Springer Nature B.V
• Subjects: Algorithms; Analysis; Coding standards; Compressed; Computer Communication Networks; Computer Science; Data Structures and Information Theory; Detection; Digital video; Hypotheses; Multimedia; Multimedia computer applications; Multimedia Information Systems; Networks; Reconstruction; Regression; Sampling; Sensors; Signal processing; Sparsity; Special Purpose and Application-Based Systems; Studies; Video; Wireless networks; Wireless multimedia sensor networks; Elastic net; Compressed sensing; Hypothesis prediction; Distributed video coding
• ISSN: 1380-7501; 1573-7721
• DOI: 10.1007/s11042-013-1743-y

Compressed Sensing, an emerging framework for signal processing, can be used in image and video application, especially when available resources at the transmitter side are limited, such as Wireless Multimedia Sensor Networks. For a low-cost and low-power demand, we consider the plain compressive sampling and low sampling rates and propose a Compressed Video Sensing scheme. As a result, most burden of video processing can be shifted to the decoder which employs a hybrid hypothesis prediction method in reconstruction. The Elastic net-based multi-hypothesis mode, one part of the prediction method, combines the multi-hypothesis prediction and the elastic net regression together. And in the process of decoding, either this mode or the single-hypothesis one is implemented based on the threshold which is selected from [1e-11, 1). Both of the prediction modes are carried out in the measurement domain and a residual reconstruction as the final step is executed to accomplish the recovery. According to the performance presented by the simulation results, the proposed multi-hypothesis mode provides a better reconstruction quality than the other multi-hypothesis ones and the proposed scheme outperforms the observed state-of-the-art schemes for compressed-sensing video reconstruction at low sampling rates.