Loading...

Efficient Iterative Sparse Recovery Techniques

Azghani, Masoumeh | 2014

1859 Viewed
  1. Type of Document: Ph.D. Dissertation
  2. Language: Farsi
  3. Document No: 46548 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Marvasti, Farokh
  7. Abstract:
  8. In this thesis, we aim to explore the recovery of sparse signals from their compressive or random samples. At first, the Compressed Sensing (CS) recovery is considered and an iterative method with adaptive thresholding has been suggested which has superior performance compared to its counterparts in both reconstruction quality and simplicity. Then, random sampling, a special kind of compressive sensing, is investigated which is practically more efficient to be implemented than the compressive sampling scheme. A number of random sampling recovery techniques are offered based on sparsity which has very low computational complexity in a way that largedimensional signals can efficiently be compressed and sampled using these techniques.Our simulations on the images confirm this point.
    Another problem investigated in this thesis is the sparsity-based microwave medical imaging of cancer tumors. In this application, the CS measurement matrix is imposed by the problem and we do not have any control on it. Three sparse recovery techniques have been suggested for this application. The proposed methods are based on iterative adaptive thresholding, reweighted least squares minimization, and smoothed iterative adaptive thresholding. Moreover, the convergence proof of the first two have been provided in the thesis.The last application considered here is the reconstruction of the video frames sampled by CS. We have proposed a sparse recovery method which reconstructs the video frames using the multi-hypothesis technique. The suggested method outperforms the other schemes in both accuracy and simplicity
  9. Keywords:
  10. Sparse Recovery ; Random Sampling ; Compressive Sensing ; Microwave Imaging ; Compressed Video Sensing

 Digital Object List

 Bookmark

...see more