A multichannel maximum likelihood estimation approach to radar imaging

This approach for extracting the two-dimensional (2-D) scattering center model is based on the combination of the cosine-modulated filter bank satisfying the perfect reconstruction condition (PR-CMFB) and maximum likelihood estimation. The analysis banks are first used to implement the multichannel decomposition of scattering signal in frequency-aspect domain. The synthesis banks are then used to recombine the decomposed signals and to conveniently obtain several separable subregional signals with high energy and without loss of information. Therefore, the maximum likelihood estimator (MLE) can be used in each subregion independently, making the approach very efficient and possessing superior noise performance. The simulations show that such combination works well and is very promising for radar imaging