Multibaseline SAR Interferometry from Complex Data

Multibaseline interferometric synthetic aperture radar (In-SAR) systems are used to obtain the height profile of the observed ground scene with high accuracy. The techniques that are commonly used exploit only the interferometric phase information, and are based on maximum likelihood (ML) estimation. Due to the difficulty of expressing the multi-baseline likelihood function in closed form, they adopt the statistical independence approximation of the interferometric phases. In this paper, we investigate the effect of this approximation, and analyze the performance of two statistical methods exploiting complex SAR images, which takes into account the mutual correlation among all the interferometric images. These two methods are based on a local and a contextual approach. The local one is based on an ML estimator, and processes each pixel independently from the neighboring ones. The contextual approach adopts a Bayesian estimator in order to regularize and improve the height reconstruction. The presented results show that the inclusion of SAR image correlation information allows to improve the height reconstruction accuracy. The proposed method also provides an estimation of speckle-reduced image intensity.