Least squares-based filter for remote sensing image noise reduction

Abstract

The Vondrák filter is a unique technique for smoothing data. The filter aims to achieve a balance between the fidelity and the smoothness of the filtered results. It can therefore preserve the original attributes of the observational data while, at the same time, smooth out the noise. We reformulate the 1-D Vondrák filter that has been widely used in data processing in fields such as astronomy and geophysics and then extend it into two dimensions. The method of conjugate gradients is used to solve the least squares optimization problem. The proposed 2-D filter is a powerful tool for enhancing the quality of various geoscience and remote sensing data such as satellite images. Various tests with simulated and real synthetic aperture radar interferograms show that the new filter is very effective in removing the noise.