Comparison of Iterative Deconvolution Algorithms for the Mapping of Acoustic Sources

Abstract

The DAMAS algorithm is compared with the DAMAS2 method and 
\ntwo other Fourier-based deconvolution approaches.
\nOne is the Richardson-Lucy method, 
\nwhich is widely used for the deconvolution of
\nastronomical images.
\nThe second is a gradient-type NNLS approach, where
\nspectral procedures are implemented to accelerate the computations.
\nIt is described how the DAMAS2 and the Fourier-based NNLS 
\nalgorithms can be embedded in
\nan outer iteration loop to take the variation of the 
\npoint-spread function into account.
\nAll methods are tested with synthetic data.
\nAt first an example with a simple linear array and a small opening angle
\nis considered, where the variation of the point-spread function
\nin the source region is negligible.
\nIn this test case the results of the DAMAS2, the Fourier-based NNLS,
\nthe Richardson-Lucy, and the original DAMAS algorithm are compared.
\nIt is shown that all algorithms
\nmore or less introduce oscillations in the reconstructed source
\ndistribution.
\nThe effect of background noise on the results is investigated.
\nA second test case with a planar array and
\na large opening angle is presented, to demonstrate the influence
\nof a strong variation of the point-spread function.
\nIt is shown that the methods which neglect the variation of the point-spread
\nfunction lead to distorted results, while
\nthe embedded Fourier-based methods and the DAMAS algorithm 
\ngive a significantly better reconstruction of the source distribution.
\n