Abstract
The nonlocal approach, proposed originally for additive white Gaussian noise image filtering, has rapidly gained popularity in many applicative fields and for a large variety of tasks. It has proven especially successful for the restoration of Synthetic Aperture Radar (SAR) images: single-look and multi-look amplitude images, multi-temporal stacks, polarimetric data. Recently, powerful nonlocal filters have been proposed also for Interferometric SAR (InSAR) data, with excellent results. Nonetheless, a severe decay of performance has been observed in regions characterized by a uniform phase gradient, which are relatively common in InSAR images, as they correspond to constant slope terrains. This inconvenience is ultimately due to the rare patch effect, the lack of suitable predictors for the target patch. In this paper, to address this problem, we propose the use of offset-compensated similarity measures in nonlocal filtering. With this approach, the set of candidate predictors is augmented by including patches that differ from the target only for a constant phase offset, which is automatically estimated and compensated. We develop offset-compensated versions of both basic nonlocal means and InSAR-Block-Matching 3D (BM3D), a state-of-the-art InSAR phase filter. Experiments on simulated images and real-world TanDEM-X SAR interferometric pairs prove the effectiveness of the proposed method.
Highlights
Synthetic Aperture Radar Interferometry (InSAR) is one of the most powerful tools to gather information on Earth’s topography and its deformation in time
In [30,39] we have proposed Interferometric SAR (InSAR)-BM3D, a nonlocal method for InSAR phase filtering inspired to the nonlocal Block-Matching 3D (BM3D) algorithm [40], which blends the nonlocal approach with transform-domain shrinkage and Wiener filtering
We focus on the offset-compensated version of InSAR-BM3D and analyze its performance on several simulated and real-world test images
Summary
Synthetic Aperture Radar Interferometry (InSAR) is one of the most powerful tools to gather information on Earth’s topography and its deformation in time It exploits two or more SAR images of the same region, acquired with some spatial and/or temporal diversity. SAR images are corrupted by intense noise, which makes it difficult to extract reliable information from them This applies in particular to interferometric data which are affected by spatial and temporal decorrelation effects, associated with scattering changes in the signal backscattered at the different antennas and at the different acquisition times. The underlying assumption is that close pixels are generally similar to one another This certainly holds in stationary areas of the image, where, local filters work very well, reducing the noise variance without introducing any bias in the estimate. In non-stationary areas, like at the boundary between different regions, or in the presence of compact objects, close pixels are not necessarily similar, and local averages introduce significant biases in the estimate, which translates into a resolution loss
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
