Reconstruction of compressed video via non-convex minimization
Chao Ji,
Jinshou Tian,
Liang Sheng,
Kai He,
Liwei Xin,
Xin Yan,
Yanhua Xue,
Minrui Zhang,
Ping Chen,
Xing Wang
Affiliations
Chao Ji
Key Laboratory of Ultra-fast Photoelectric Diagnostics Technology, Xi’an Institute of Optics and Precision Mechanics (XIOPM), Chinese Academy of Sciences (CAS), Xi’an, Shaanxi 710119, China
Jinshou Tian
University of Chinese Academy of Sciences, Beijing 100049, China
Liang Sheng
The Northwest Institute of Nuclear Technology, Xi’an 710024, China
Kai He
University of Chinese Academy of Sciences, Beijing 100049, China
Liwei Xin
Key Laboratory of Ultra-fast Photoelectric Diagnostics Technology, Xi’an Institute of Optics and Precision Mechanics (XIOPM), Chinese Academy of Sciences (CAS), Xi’an, Shaanxi 710119, China
Xin Yan
University of Chinese Academy of Sciences, Beijing 100049, China
Yanhua Xue
University of Chinese Academy of Sciences, Beijing 100049, China
Minrui Zhang
University of Chinese Academy of Sciences, Beijing 100049, China
Ping Chen
Key Laboratory of Ultra-fast Photoelectric Diagnostics Technology, Xi’an Institute of Optics and Precision Mechanics (XIOPM), Chinese Academy of Sciences (CAS), Xi’an, Shaanxi 710119, China
Xing Wang
University of Chinese Academy of Sciences, Beijing 100049, China
This paper studies the sparsity prior to compressed video reconstruction algorithms. An effective non-convex 3DTPV regularization (0 < p < 1) is proposed for sparsity promotion. Based on the augmented Lagrangian reconstruction algorithm, this paper analyzes and compares three non-convex proximity operators for the ℓp-norm function, and numerous simulation results confirmed that the 3DTPV regularization can gain higher video reconstruction quality than the existing convex regularization and is more competitive than the existing video reconstruction algorithms.
