Fractal-Based Local Range Slope Estimation from Single SAR Image with Applications to SAR Despeckling and Topographic Mapping

Abstract

In this paper, we propose a range slope estimation procedure from single synthetic aperture radar (SAR) images with both methodological and applicative innovations. The retrieval algorithm is based on an analytical linearized direct model, which relates the SAR intensity data to the range local slopes and encompasses both a surface model and an electromagnetic scattering model. Scene topography is described via fractal geometry, whereas the Small Perturbation Method is adopted to represent the scattering behavior of the surface. The range slope map is then used to estimate the surface topography and the local incidence angle map. For topographic mapping applications, also referred to as shape from shading, a regularization procedure is derived to recover the azimuth local slope and reduce distortions. Then we present a new intriguing application of the inversion procedure in the field of SAR despeckling. Proposed techniques and high-level products are tested in a wide series of experiments, where the algorithms are applied to both simulated (canonical) and actual SAR images. It is proved that the proposed range slope retrieval technique can (1) provide an estimate of the surface shape, with overall better performance w.r.t. typical models used in this field and (2) be useful in advanced despeckling techniques.

What Can Multifractal Analysis Tell Us about Hyperspectral Imagery?

“…Over time, applications have expanded from a simple description of linear features to surface feature analysis and research on information stored in the form of images [3]. Remote sensing researchers have applied the concept to aerial and satellite imagery analysis, including radar data processing for irrigation mapping [4], surface modelling [5], land cover mapping [6], estimation of building density [7] and segmentation [8], among others.…”

A Convenient Analytical Framework for Electromagnetic Scattering From Composite Targets

“…To this end, the modeling rationale can be decomposed into two main steps: (1) the acquisition process, that is, the set of procedures applied to measurements to obtain the final observable, for example, a SAR image, a scatterometric, or a delay Doppler map and (2) the physical mechanisms that contribute to the measurement performed by the sensor. Concerning the first issue, accurate and well‐established imaging models, which describe the data processing chain, have been developed in the literature (see Clarizia et al, ; Di Martino et al, ; Franceschetti & Lanari, ; Zavorotny & Voronovich, ). Such models provide a relationship between the sensor observable and the (RCS) of the illuminated scene.…”

“…Here we focus on the scattering models, since the RCS is demonstrated to be proportional to the final observable (Clarizia et al, ; Di Martino et al, ). Closed‐form expressions of RCS may greatly ease the retrieval procedure in inversion algorithms and the automatic processing by computer programs.…”