Fast implementation of SAR imaging using sparse ML methods

High-resolution sparse spectral estimation techniques have recently been shown to offer significant performance gains as compared to most conventional estimation approaches, although such methods typically suffer the drawback of being computationally cumbersome. In this paper, we seek to alleviate this drawback somewhat, examining computationally efficient implementations of the recent iterative sparse maximum likelihood-based approaches (SMLA), exploiting the inherent rich structure of these estimators. The derived implementations reduce the resulting computational complexity with at least one order of magnitude, while still yielding exact implementations. The effectiveness of the discussed techniques are illustrated using experimental examples.