• DocumentCode
    778244
  • Title

    Stable recovery of sparse overcomplete representations in the presence of noise

  • Author

    Donoho, David L. ; Elad, Michael ; Temlyakov, Vladimir N.

  • Author_Institution
    Dept. of Stat., Stanford Univ., CA, USA
  • Volume
    52
  • Issue
    1
  • fYear
    2006
  • Firstpage
    6
  • Lastpage
    18
  • Abstract
    Overcomplete representations are attracting interest in signal processing theory, particularly due to their potential to generate sparse representations of signals. However, in general, the problem of finding sparse representations must be unstable in the presence of noise. This paper establishes the possibility of stable recovery under a combination of sufficient sparsity and favorable structure of the overcomplete system. Considering an ideal underlying signal that has a sufficiently sparse representation, it is assumed that only a noisy version of it can be observed. Assuming further that the overcomplete system is incoherent, it is shown that the optimally sparse approximation to the noisy data differs from the optimally sparse decomposition of the ideal noiseless signal by at most a constant multiple of the noise level. As this optimal-sparsity method requires heavy (combinatorial) computational effort, approximation algorithms are considered. It is shown that similar stability is also available using the basis and the matching pursuit algorithms. Furthermore, it is shown that these methods result in sparse approximation of the noisy data that contains only terms also appearing in the unique sparsest representation of the ideal noiseless sparse signal.
  • Keywords
    approximation theory; iterative methods; signal denoising; signal representation; time-frequency analysis; Kruskal rank; greedy approximation algorithm; incoherent dictionary; matching pursuit; noisy data; optimal sparse decomposition; signal processing theory; sparse overcomplete representation; stable recovery; stepwise regression; superresolution signal; Dictionaries; Linear algebra; Matching pursuit algorithms; Noise generators; Noise level; Signal processing; Signal processing algorithms; Signal representations; Stability; Vectors; Basis pursuit; Kruskal rank; greedy approximation; incoherent dictionary; matching pursuit; overcomplete representation; sparse representation; stability; stepwise regression; superresolution;
  • fLanguage
    English
  • Journal_Title
    Information Theory, IEEE Transactions on
  • Publisher
    ieee
  • ISSN
    0018-9448
  • Type

    jour

  • DOI
    10.1109/TIT.2005.860430
  • Filename
    1564423