Exploiting the tree‐structured compressive sensing of wavelet coefficients via block sparse Bayesian learning. Issue 16 (1st August 2018)

• Record Type: Journal Article
• Authors: Qin, Le; Tan, Jiaju; Wang, Zhen; Wang, Guoli; Guo, Xuemei
• Is Part Of: Electronics letters. Volume 54:Issue 16(2018)
• Page Start: 975
• Page End: 976
• Publication Date: 2018-08-01
• Subjects: matrix algebra -- learning (artificial intelligence) -- compressed sensing -- trees (mathematics) -- Bayes methods -- image coding -- signal reconstruction; wavelet coefficients -- block sparse Bayesian learning -- tree structure -- recovering signals -- nonoverlapped block structure -- sparse model -- bSBL‐based algorithm -- intra‐block correlations -- compressive sensing algorithms; Electronics -- Periodicals; 621.381
• DOI: 10.1049/el.2018.0224
• Languages: English
• ISSNs: 0013-5194

Abstract

Abstract : In this Letter, the authors propose a novel framework based on block sparse Bayesian learning (bSBL) for exploiting the tree structure on wavelet coefficients in the process of recovering signals. A Boolean matrix is designed to transfer the tree structure of wavelet coefficients to a non‐overlapped block structure. In this block‐structured sparse model, the bSBL‐based algorithm is used to learn the intra‐block correlations and to derive the updating rule of model parameters. Experimental results show that for both 1D and 2D signals their proposed algorithm has superior performances compared with other model‐based compressive sensing algorithms.