Noise Reduction of Undersampled ISAR Images by Entropy-Based Regularization

摘要

The research described in this report was performed during the 1996 fiscal yearas part of an effort to improve radar classification and identification capabilities for noncooperative airborne targets. Effective radar based imaging techniques should be able to employ the information contained in both the amplitude and the phase of the scattered field data. Moreover, since practical radar systems collect restricted and noisy data, image reconstwction algorithms should allow for the inclusion of a priori information about the target and should attempt to mitigate the effects of noise. All three issues-complex-valued data, prior information, and noise mitigation-are often not adequately treated by conventional entropy-based methods (and are mutually exclusive in many other approaches). We present a Bayesian regularization method that employs a cross-entropy functional and does not require the measured data to positive and real-valued. In fact, this technique is not restricted to a particular type of data and we present reconstruction examples for several different imaging problems. The basic model of this approach is similar to that used in usual maximum a posteriori analysis and allows for a more general relationship between the image and its 'configuration entropy' than that usually employed.