Estimation of polarization orientation angle (POA) is of general interest in polarimetric SAR (PolSAR) image processing because it is useful to remove the orientation-induced polarimetric distortion but also to infer physical target parameters. Under the assumption of scatter reflection symmetry, POA can be readily estimated from a circular polarization basis formulation or equivalently from zeroing the Pauli-basis correlation term ${{\boldsymbol{T}}}_{23}$. This solution does not account for a possible presence of helical components; alternatively, we developed a general formulation for POA estimation using Touzi parameterization that explicitly contains a helical angle. When applied to PolSAR imagery of rough terrain surface, those two estimators generally give different results. Based on the POA derived from coregistered digital elevation model data, we find the former estimator can underestimate the POA for rough terrain surface. In this article, we further investigate the underlying cause. Although these two POA estimators differ in whether assuming zero helicity or not, the inherent helicity on the rough surface scattering is negligible; instead, the difference in the estimated POA is mainly attributed to the degree of mixing of a secondary scattering mechanism. This mixing impact can be significant even for rough terrain with only light vegetation cover, suggesting a practical need for inclusion of alternative estimators.

Model-based polarimetric decompositions are often used to generate scene classifications from polarimetric synthetic aperture radar (POLSAR) imagery. The classification quality largely depends on whether the adopted polarimetric scattering mechanisms match the major in-scene scattering mechanisms. However, in-scene scattering variations still cause potential mismatches, resulting in unphysical or inaccurate decomposition results. The robustness of model-based polarimetric decompositions remains a general concern. In this article, we address the robustness of model-based polarimetric decomposition to variations of the in-scene scattering mechanisms using simulated datasets. The known simulated scattering mechanisms provide the needed “ground truth” against which we quantitatively evaluate the robustness of polarimetric decompositions. Accurate retrieval of known scattering mechanisms from simulated polarimetric data led us to develop a robust model-based decomposition. We propose a new approach to solve model-based decomposition by employing an $L_{1}$ -regularized optimization procedure, which automatically selects a set of optimal polarimetric scattering mechanisms and guarantees nonnegative powers for the selected scattering mechanisms. We illustrate this $L_{1}$ model decomposition by employing both simulated datasets and actual POLSAR imagery. Our new $L_{1}$ -regularized approach to POLSAR model-based decomposition that mitigates observed biases seen in earlier decompositions provides robust scattering mechanism estimates and eliminates unphysical, negative scattering powers.

Polarimetric SAR (PolSAR) imagery offers an enhanced capability to reveal the salient scattering properties of scene content. PolSAR-based target decomposition has been widely used to show different apparent scattering mechanisms for various target classes, empowering a direct yet powerful technique for SAR imagery analysis. Among those common targets, modeling the random volume scattering from vegetation is one of the most important. Generally, one models vegetation as a cloud of randomly oriented thin cylinders, mainly intended for twigs and branches. At high radar frequencies, PolSAR imagery shows a strong response from leaves in the vegetation canopy. In this letter, we derive the polarimetric scattering theory for general random volume scatterers, including both thin cylinders and thin disks as limiting cases for leaf response. Adding the proposed random thin disk model explains the observed scattering difference between deciduous forest and coniferous forest, which we then incorporate into a new model-based PolSAR target decomposition scheme.

Model-based polarimetric decompositions are often used to generate scene classifications from polarimetric SAR imagery. The original Freeman-Durden model has been modified and improved upon multiple times over the past 2 decades. However, quantitative, in-depth analyses of these incoherent model-based decompositions have lagged in comparison. Here we assess model-based polarimetric decomposition techniques for robustness to variations of the in-scene scattering mechanisms using simulated data sets. The simulated scattering mechanisms, e.g. volume scattering model, polarimetric signal-to-noise, simple multi-bounce models, etc., are known and thus provide "ground truth." We will illustrate our results by applying select model-based decompositions to both simulated and actual polarimetric SAR imagery.

Bistatic synthetic aperture radar (BiSAR) offers flexible imaging configurations to form SAR images at various sensing geometries. However, polarimetric observations with BiSAR are also subject to greater geometrical dependency compared to those with monostatic SAR. The polarimetric analysis practical to retrieve geophysical target parameters is currently lacking on the BiSAR imagery. In this paper, we investigate its dependence on the bistatic geometry using both isotropic and anisotropic targets. It shows that for general targets the bistatic geometry convolves with target shape, orientation and polarization angles in all polarimetric parameters.

In this paper, we address polarimetric radar scattering from steep terrain and its relation to polarization orientation angle (POA) by adopting a tilted surface model based on Bragg scattering. We found that, as the azimuthal slope increases, |VV| decreases at a faster rate than |HH|, they become equal, when POA is ±45°, and |HH|>|VV| afterwards. Contrasting to the general perception, the cross-pol |HV| does not always increase with azimuth slope, but reaches a maximum then decreases to zero. In addition, we investigate the effect of soil moisture on PolSAR scattering characteristics of steep terrain and the effect of vegetation over surface on POA estimation. The latter is demonstrated with NASA/JPL TOPSAR L-band PolSAR data and C-band InSAR data.

Pixels in SAR imagery can be grouped as those of coherent scatterers and those of distributed scatterers. Identifying coherent scatterers from distributed scatterers is of a general interest because the former are often associated with man-made targets whereas the latter are often associated with natural clutter. An ideal coherent scatter would exhibit scattering stability with respect to time, frequency, aperture angle, and polarization state. At high resolution, coherent scatterer returns may not satisfy the premise of circularly complex speckle. In this paper, we compare and evaluate the coherent scatterer detection methods based on the spectral stability among multiple sub-bands and one based on the non-circularity metric. It shows evaluating non-circularity offers a favorable option for identifying coherent scatters and accordingly demonstrates the need to be cautious on the circular speckle premise in high resolution images.

Polarization orientation angle (POA) is an important parameter of polarimetric radar scattering from slopes in mountainous region. In this paper, we address polarimetric scattering from steep slopes and its relation to POA by adopting a tilted surface model based on Bragg scattering. We have found that, as the azimuthal slope increase, |VV| decreases at a faster rate than |HH|, they become equal, when POA is +-45 degrees, and |HH|>|VV| afterwards. And the cross-pol |HV| does not always increase with azimuth slope, but reaches a maximum then decreases to zero. In addition, we investigate the effect of soil moisture on PolSAR scattering characteristics of steep terrain and the effect of vegetation over surface on POA estimation. The latter is demonstrated with NASA/JPL TOPSAR L-band PolSAR data and C-band InSAR data.

Model-based polarimetric decompositions have long been applied to classify polarimetric SAR imagery. This long and successful history has, at times, been interrupted by nagging issues: polarimetric orientation angle effects, negative scattering powers, simplistic scattering models, etc. We test model-based polarimetric decomposition techniques for robustness to variations of the underlying in-scene scattering mechanisms. We determine the accuracy and robustness of the decomposition results using simulated data sets. The simulation inputs, e.g. volume scattering model, polarimetric signal-to-noise, simple multi-bounce models, etc., are known and thus provide “ground truth” for comparison to model-based decomposition results. We will illustrate our results by applying a select set of model-based decompositions to both simulated and actual PolSAR imagery.

Speckle filtering is an indispensable operation in synthetic aperture radar (SAR) image processing but one which inevitably reduces image resolution. In order to preserve the intrinsic target features, adaptive speckle filters have been developed using weighted averages commensurate with the similarity of the target statistics. The target statistics are commonly derived from a prefiltering step which suffers from residue speckle contamination and feature smearing. In this paper, we adopted finite mixture models to characterize the observed in-scene variation and proposed a rigorous and progressive mixture regression method to better estimate the target statistics. The mixture model once fitted is able to capture the statistical properties of the highly textured and heterogeneous target variation, which is often observed in high-resolution SAR images. A nonlocal mean method is used for robust similarity evaluation of the local variation patterns between small image patches. The goal is to develop an improved polarimetric SAR (PolSAR) speckle filter that can accomplish a solid balance between speckle suppression and feature preservation. With the proposed filter, distinct scattering mechanisms and small-scale target features are retained from the start, even with single-look complex PolSAR observations. We test the algorithm using simulated data and single-look high-resolution PolSAR images: one acquired by DLR's F-SAR system and one by DLR's E-SAR system.

In high-resolution polarimetric synthetic aperture radar (SAR) images, texture variation increasingly dominates the scene of those selected targets, creating a major challenge in speckle modeling and target classification. When texture is incorporated into speckle models, it is often assumed that pixels are independent from each other. In this paper, we allow dependent neighboring pixels and evaluate the texture as a Markov random field (MRF). The estimated local dependence potentially reveals the spatial patterns of targets which are used to augment polarimetric SAR image classification through a MRF-mixture model. We evaluated the MRF-mixture model to describe textured speckle and performed a classification experiment on a high-resolution image. It shows an improvement when heterogeneity and spatial patterns are clearly present.

In this paper, the recent advancement of PolSAR filtering is reviewed. Speckle filtering is a necessary procedure for most of PolSAR applications, to reduce speckle noise level and to form incoherent scattering by averaging coherency or covariance matrices of neighboring pixels. In this paper, we addressed several important but often ignored issues associated with PolSAR speckle filtering, and the recent very high resolution SLC PolSAR data demands special consideration in speckle filtering. The conclusion, we have reached, is that the best filter does not exist, because the desired filtering is determined by application requirements and personnel preference. A filter may be good for one application, but could be undesirable for other applications. In other words, speckle filtering is not an exact science, and the perfect speckle filtering cannot be achieved.

We propose to test model-based polarimetric decomposition techniques for robustness to variations in the underlying in-scene scattering mechanisms. The accuracy and robustness of the decomposition results are determined from simulated data sets. The simulation input parameters, e.g. volume scattering model, polarimetric signal-to-noise, etc., are known and thus provide the “ground truth” for comparison to the model results. The observed correlations between simulation and model results allow estimates of model accuracy and model robustness.

In this paper, we investigate the double bounce radar backscattering from building walls not aligned in the azimuth direction (i.e., slanted double bounce scattering) based on scattering models. We apply the Fresnel equation for reflections from the building wall, and then the bistatic small perturbation model developed by Ulaby for ground scattering. Our analysis indicates that scattering matrix is not symmetrical for the path from radar to wall to ground and back to radar, but adding the reverse path from radar to ground to wall and back to radar makes the scattering matrix symmetrical. The effect on the polarization orientation estimation for an urban scene using L-band E-SAR will be analyzed.

At high resolution, synthetic aperture radar (SAR) speckle tends to be non-Gaussian distributed and diversely textured. Many parametric speckle distributions have been developed to fit specific in-scene content. In contrast, mixture models offer an empirical approximation with the potential to fit arbitrary variations. In this letter, we investigate the feasibility and the efficiency of using finite mixture models of an identical parametric kernel to characterize the wide range of high-resolution speckle. We evaluate and compare the capability of mixture fitting with gamma, $\mathcal{K}$, and $\mathcal{G}^{0}$ kernels against various scene types. Despite the characterization disparity among these base kernels, we show that using any of them in a mixture setting rapidly improves speckle modeling. Finite gamma mixtures, even with a simple kernel form, are applicable to high-resolution SAR imagery for consistent description of complex textured speckle variations.