On the performance improvement of microwave imaging using MDL principle

The Multiple Signal Classification (MUSIC) algorithm is a well-known method for target (scatterer) localization in microwave imaging. An important issue for this target localization scheme is to determine the number of scatterers. Generally, the number of scatterers can be deduced from the multiplicity of the largest eigenvalues of a prespecified correlation matrix. In both low SNRs and real data, this method is an invalid procedure for targets enumeration, thus the performance of the MUSIC algorithm will be degraded under these circumstances. To improve the performance of microwave imaging-based MUSIC algorithm, in this paper, we utilize the Minimum Description Length (MDL) estimator, a well-known principle in the information theory, to estimate the number of scatterers. In addition, we introduce the improved MDL estimator for scatterers enumeration in real data. For this estimator, we determine the penalty term coefficient of the MDL estimator using computer simulations. Numerical results show that the MDL and the improved MDL schemes have high accuracy in the scatterers enumeration and improve the performance of microwave imaging for both simulated and real data.