Multifrequency Polarimetric SAR Image Despeckling by Iterative Nonlocal Means Based on a Space-Frequency Information Joint Covariance Matrix

Abstract

This paper presents an iterative nonlocal means (NLM) filtering method under the Bayesian framework to deal with the issue of multifrequency fully polarimetric synthetic aperture radar (PolSAR) image despeckling. Differing from most of the PolSAR filters designed for single-frequency data, the proposed NLM method is developed based on a space-frequency information joint covariance matrix, which can not only utilize multifrequency polarimetric information but also exploit the correlation between any two pixels in an image patch. Furthermore, with the aim of accelerating the filtering procedure and better retaining image details, an effective preselection step is employed. The filtering results obtained with both a simulated dataset and real multifrequency PolSAR datasets acquired by the AIRSAR system confirm the good performance of the proposed method in both reducing speckle and retaining details, when compared with some of the state-of-the-art despeckling algorithms.