A fast offset estimation approach for interferometric synthetic aperture radar (InSAR) image pair subpixel registration is proposed for cases of relatively gentle topography and/or short baseline. A coarse-to-fine registration strategy is taken. The pixel-level offset is estimated in the coarse registration step by a fast feature-based estimation, which uses the speeded up robust feature operator and fast least trimmed squares (Fast-LTS) estimator to accelerate the feature extraction and parameter estimation. A fine registration is performed subsequently. The conventional normalized cross-correlation algorithm (NCCA) searches for the optimal subpixel offset by oversampling either the coarse cross correlation or the InSAR image patch pair. The offset estimation accuracy is restricted by the oversampling rate, and the computational burden is heavy when high accuracy is demanded. In this letter, we transform the oversampling and correlation searching process of NCCA into a nonlinear optimization problem, which takes the maximization of the coherent cross correlation as the objective function; by solving it, the subpixel offset can be fast and exactly obtained without any image oversampling. The final registration parameters are inverted by Fast-LTS fitting of a series of subpixel tie point correspondences which can be constructed after applying the approach to several image patch pairs. RadarSat-2 data are used to test the approach, and the results show that it performs very well not only on the speed but also on the accuracy.

Huynen decomposition (HD) as the first formalized target decomposition has not been widely accepted. The preference for symmetry and regularity restricts not only its application but also its unification with other target dichotomies. The nonuniqueness issue then arises because we may have different dichotomies of radar targets, but we have no idea on how to select them. In this paper, a unified Huynen dichotomy is developed by extending HD for a full preference for symmetry and regularity, nonsymmetry, irregularity, and their couplings. It covers all of the existing dichotomies and provides a unified selection mechanism for them. Scattering preference is identified as a main feature of target dichotomy, and its concise description is devised by relating each dichotomy to a canonical scattering. A scattering degree of preference (SDoP) parameter is defined to measure the preference of each dichotomy. In virtue of an adaptive combination and permutation of SDoPs, a scattering pyramid description of the mixed scattering is developed, which has better discrimination of target than entropy/alpha. An SDoP/alpha classification is further proposed by statistical modeling of the unified dichotomy, which is a competent alternative to entropy/alpha. The excellent performance of unified dichotomy makes us believe that the existing concerns on HD are well treated and the Huynen-Cloude controversy, in a sense, may be ended.

Scattering similarity was first proposed by Yang et al. to measure the similarity between two single scatterers. It was extended by Chen et al. to measure the similarity between a mixed scatterer and a single scatterer. This letter develops a random similarity parameter to further measure the similarity between two mixed scatterers. The parameter not only covers Yang's and Chen's similarities by providing a general scattering similarity measurement, but also is useful for scattering randomness description by enabling a fast alternative and a competent complementary to the entropy parameter. A novel model-based characterization scheme of mixed scatterer is then proposed by parallel combining the random similarities between the mixed scatterer and three canonical mixed volume scatterers. By further fusing with the SPAN, the scheme can characterize both the texture and the scattering information regarding a target. Comparative experiment with Chen's approach on L-band ESAR Oberpfaffenhofen data demonstrates its excellent discrimination of radar targets.

An adaptive classification is developed as a hybrid of the eigenvector-based and the model-based target decompositions for polarimetric synthetic aperture radar (PolSAR) data. The classification adopts the canonical scattering models that widely used in model-based decompositions to provide an improvement for the well-known H/α classification. First, a correspondence principle is adopted to adaptively identify the matched canonical models. The selected models are parallelly combined based on the scattering similarity for a fine depiction of the scattering mechanism then. Twelve classes are finally obtained, and each one carries a unique symbol to show a specific scattering. The classification does not depend on a particular data set, avoids the hard partitioning, and solves the obscures in H/α. Comparison on the real PolSAR data sets with H/α and the existing scattering similarity-based classification validates the better discrimination.

A fast and competent alternative to the widely used Cloude–Pottier entropy/alpha ($H/ \alpha $) classification is developed for the rapid response application of polarimetric synthetic aperture radar (PolSAR) data. Random similarity which measures both the scattering similarity and randomness of polarimetric scatterers is used to enable an $H/ \alpha $ -like classification in terms of two key parameters, i.e., the similarity-based angle $\alpha _{s}$ and entropy $H_{s}$, as the alternatives to the Cloude–Pottier angle $\alpha $ and entropy H, respectively. Parameters $\alpha _{s}$ and $H_{s}$ maintain the same physical information as parameters $\alpha $ and H, so the existing knowledge regarding $\alpha $ and H can be naturally extended to them. Angle $\alpha _{s}$ measures scattering mechanism and is ranged within the same interval [0°, 90°] as $\alpha $ while entropy $H_{s}$ measures scattering randomness which is also a logarithm within the interval [0, 1] similar to H. The pixelwise eigendecomposition in the calculation of $\alpha $ and H is avoided for $\alpha _{s}$ and $H_{s}$, and the resulted efficiency improvement is, thus, considerable. By rigorously modeling the $\alpha _{s}\hbox{--}\alpha $ and the $H_{s}\hbox{--}H$ relationship to illustrate the competence of the $H_{s}\hbox{--}\alpha _{s}$ combination in discrimination of target and to identify the searching ranges for the boundary determination, an $H_{s}/ \alpha _{s}$ classification is then devised with the boundaries of the eight effective classes being determined by an optimization to minimize the misclassification and further integrated on different PolSAR images to remove the possible bias from dataset for general applicability. Comparative experiment on both space-borne and airborne PolSAR datasets with $H/ \alpha $ indicates that $H_{s}/ \alpha _{s}$ can achieve very consistent roll-invariant target discrimination as $H/ \alpha $ (overall accuracy ≥95%, kappa coefficient ≥0.95) but with averagely 150 times higher efficiency although the LAPACK-based eigenanalysis tool has been used to accelerate the eigendecomposition for $H/ \alpha $. Preliminary result from the adaptive model-based classification reveals that the $H_{s}$-involved boundaries in $H_{s}/ \alpha _{s}$ are independent of a particular PolSAR dataset.