Efficient estimation of 2-dimensional frequencies of sinusoids by the annihilator method and constrained total least squares

The authors address the problem of estimating the 2-dimensional frequencies from a set of double indexed samples consisting of unknown linear combinations with efficiency. These problems arise in high resolution radar/sonar imaging, array signal processing and nuclear magnetic resonance imaging. A new approach is based on the annihilator method and a generalization of the CTLS technique. Simulation results show that this approach can estimate the 2-D frequencies with accuracies approaching the Cramer-Rao bound even when the separation of the sinusoids is a fraction of the discrete Fourier transform resolution bin.<>