Radar Image Processing Research Articles

PSDSD-A Superpixel Generating Method Based on Pixel Saliency Difference and Spatial Distance for SAR Images.

Superpixel methods are widely used in the processing of synthetic aperture radar (SAR) images. In recent years, a number of superpixel algorithms for SAR images have been proposed, and have achieved acceptable results despite the inherent speckle noise of SAR images. However, it is still difficult for existing algorithms to obtain satisfactory results in the inhomogeneous edge and texture areas. To overcome those problems, we propose a superpixel generating method based on pixel saliency difference and spatial distance for SAR images in this article. Firstly, a saliency map is calculated based on the Gaussian kernel function weighted local contrast measure, which can not only effectively suppress the speckle noise, but also enhance the fuzzy edges and regions with intensity inhomogeneity. Secondly, superpixels are generated by the local k-means clustering method based on the proposed distance measure, which can efficiently sort pixels to different clusters. In this step, the distance measure is calculated by combining the saliency difference and spatial distance with a proposed adaptive local compactness parameter. Thirdly, post-processing is utilized to clean up small segments. The evaluation experiments on the simulated SAR image demonstrate that our proposed method dramatically outperforms four state-of-the-art methods in terms of boundary recall, under-segmentation error, and achievable segmentation accuracy under almost all of the experimental parameters at a moderate segment speed. The experiments on real-world SAR images of different sceneries validate the superiority of our method. The superpixel results of the proposed method adhere well to the contour of targets, and correctly reflect the boundaries of texture details for the inhomogeneous regions.

МЕТОДИ И АЛГОРИТМИ ЗА ОБРАБОТКА НА АЕРОФОТО И РАДИОЛОКАЦИОННИ ИЗОБРАЖЕНИЯ, УЛЕСНЯВАЩИ АНАЛИЗА НА ИНФОРМАЦИЯТА (Methods and Algorithms for Aerial and Radar Image Processing, that Facilitating Analysis of Information)

Bulgarian Abstract: В монографията са разгледани методи и алгоритми за обработка на аерофото и радиолокационни изображения улесняващи анализа на информацията, които успешно са приложени при решаване на задачи от областта на радиофизиката, оптиката и радиолокацията. Монографията се състои от три части. В първата част са представени особеностите на човешкото визуално възприятие и методите за преобразуване на изображението в цифров вид. Във втората част са математическите методи и алгоритми за преобразуване и възстановяване на изображения, които могат да бъдат приложени върху оптични и радарни изображения. А в третата част е показана ефективността на методите и крайните резултати получени в следствие прилагането на описаните математически методи и алгоритми илюстрирани с много примери. English Abstract: Methods and algorithms for aerial and radar image processing, that facilitating analysis of information are considered in this monograph and they have been successfully applied in solving problems in radiophysics, optics and radiolocation. The monograph consists of three parts. In the first part, the features of the human visual perception and the digital image formation are discussed. In the second part, different mathematical models and algorithms for image transformation and restoration, which could be applied to photo and radar images, are considered. In the third part, effectiveness of considered methods is shown and the study results of application of different methods and algorithms are illustrated by many examples.

Processing of radar images containing objects with significantly different radar cross-section in the onboard remote sensing complex for search and rescue operations in the Arctic Region

The method of adaptive nonlinear matching of dynamic ranges of radar signal and imaging device is proposed. The method is intended for use in the onboard remote sensing complexes during formation of radar images of the earth's surface. The method is aimed at increasing the visibility of small objects on the background of the underlying surface in order to simplify the procedures for their detection and identification. The computational complexity of the proposed algorithm makes it possible to use it in real time.

Marine Radar Image Processing of Compressed Sensing Based on Arbitrary Block Statistical Histogram and Dynamic Dictionary

In compressed sensing (CS), the absolute value of the inner product of signal and atom (or dictionary) is used to select atom (or dictionary); the atoms (or dictionaries) with larger absolute values of inner product are selected to decompose and reconstruct signals. This characteristic usually makes it impossible to distinguish the edges of objects (or targets) in some complex cases, especially when the density of objects is very high. For example: let signal <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$x_{1} =[{\cdots 00110110\cdots }]^{T}$ </tex-math></inline-formula> , <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$x_{2} =[{\cdots 001110100\cdots }]^{T}$ </tex-math></inline-formula> ; <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\psi _{1} $ </tex-math></inline-formula> , <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\psi _{2} $ </tex-math></inline-formula> , <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\psi _{3} $ </tex-math></inline-formula> and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\psi _{4} $ </tex-math></inline-formula> denote several sparse basis matrices (i.e., sparse dictionaries); <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\psi _{1i} =[{\cdots 00000100\cdots }]^{T}$ </tex-math></inline-formula> , <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\psi _{2j} =[{\cdots 00110000\cdots }]^{T}$ </tex-math></inline-formula> , <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\psi _{3k} =[{\cdots 00111000\cdots }]^{T}$ </tex-math></inline-formula> , and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\psi _{4l} =[{\cdots 00111110\cdots }]^{T}$ </tex-math></inline-formula> are the sparse bases (i.e., atoms) from these sparse basis matrices (i.e., dictionaries), respectively. For signal <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$x_{1} $ </tex-math></inline-formula> , <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\psi _{4l} $ </tex-math></inline-formula> (or <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\psi _{4}$ </tex-math></inline-formula> ) is selected instead of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\psi _{2j} $ </tex-math></inline-formula> (or <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\psi _{2}$ </tex-math></inline-formula> ); likewise, for <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$x_{2} $ </tex-math></inline-formula> , <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\psi _{4l} $ </tex-math></inline-formula> (or <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\psi _{4}$ </tex-math></inline-formula> ) will be selected instead of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\psi _{3j} $ </tex-math></inline-formula> (or <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\psi _{3}$ </tex-math></inline-formula> ) and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\psi _{1i} $ </tex-math></inline-formula> (or <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\psi _{1}$ </tex-math></inline-formula> ). This causes the edge of the object to be unrecognizable. Another example is in <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$l_{p} $ </tex-math></inline-formula> regularization, it is not necessarily the optimal case: the smaller the value of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\| {\Psi _{d}^{-1} x} \|_{0} $ </tex-math></inline-formula> , the better the sparse basis matrix (i.e., dictionary), or the larger the weighing parameter. At present, the inner product is used to process almost all signals in CS. Choosing the dictionary that matches the signal best rather than the dictionary with the maximum inner product value can reconstruct the signal more accurately. In order to overcome the above shortcoming, we propose a fully automatic radar image processing algorithm of CS based on arbitrary block statistical histogram and dynamic dictionary. We use arbitrary block statistical histogram to calculate the non-zero block numbers of different sizes, which can better choose the appropriate sparse base (or dictionary) for the signal. Furthermore, the better measurement vector y can be obtained, and the signal can be reconstructed more accurately than the state-of-the-art methods at the receiving end. To realize the proposed method, we construct objective functions and flow charts for noiseless and noisy signals, respectively. In our method, for noiseless signal, discarding the sub-images that do not contain objects can reduce running time; for noisy signal, according to the theory of wavelet, choosing the appropriate wavelet (i.e., wavelet basis) can usually suppress noise. Our proposed algorithm can overcome the above shortcoming of CS in using the absolute value of the inner product, reduce running time, suppressnoise, and improve the signal-to-noise ratio (SNR). The simulation results show that our method is superior to the state-of-the-art methods.

A New Synthetic Aperture Radar (SAR) Imaging Method Combining Match Filter Imaging and Image Edge Enhancement.

In general, synthetic aperture radar (SAR) imaging and image processing are two sequential steps in SAR image processing. Due to the large size of SAR images, most image processing algorithms require image segmentation before processing. However, the existence of speckle noise in SAR images, as well as poor contrast and the uneven distribution of gray values in the same target, make SAR images difficult to segment. In order to facilitate the subsequent processing of SAR images, this paper proposes a new method that combines the back-projection algorithm (BPA) and a first-order gradient operator to enhance the edges of SAR images to overcome image segmentation problems. For complex-valued signals, the gradient operator was applied directly to the imaging process. The experimental results of simulated images and real images validate our proposed method. For the simulated scene, the supervised image segmentation evaluation indexes of our method have more than 1.18%, 11.2% and 11.72% improvement on probabilistic Rand index (PRI), variability index (VI), and global consistency error (GCE). The proposed imaging method will make SAR image segmentation and related applications easier.

Unsupervised classification for polarimetric SAR images based on the improved CFSFDP algorithm

ABSTRACTIn polarimetric synthetic aperture radar (PolSAR) image processing, the number of classes is an important factor for PolSAR image classification. Therefore, how to accurately estimate the number of PolSAR image classes is an important issue. In this article, we propose a novel unsupervised classification method which can accurately estimate the number of classes for PolSAR images. First, the PolSAR image is initialized into many small clusters by using the complementary information from Yamaguchi decomposition and distribution characteristics of data. Second, the improved clustering by fast search and find of density peaks, named as improved CFSFDP algorithm, is introduced to select the appropriate category number. Finally, to improve the representation of each category, the PolSAR data set is classified by an iterative fine-tuning process based on a complex K-Wishart function. The performance of the proposed classification approach is presented and analysed on three real data sets. The experimental results show that the proposed classification method can accurately estimate the category number and enhance the classification accuracy in comparison with other traditional methods. It is also shown that the data distribution characteristic has the additional information beyond the target scattering decomposition, and this information is important for the initialization.

Software Tools for Statistical Analysis of Some Precipitation Characteristics

The paper presents the design and implementation of software tools for statistical analysis of the real data based on the assumptions that empirical distribution can be approximated by generalized negative binomial (GNB) or generalized gamma (GG) families. Models based on GG distributions are widely applied in such practical problems as processing of synthetic-aperture radar images and speech signals, hydrological analysis and optical communications. In this paper, the GNB distributions are considered as a mixed Poisson law with the mixing GG distribution. This family could provide better fit with the different statistical data than classical negative binomial distributions that have been successfully used for analysis of precipitation events earlier. The parameter estimation is implemented using a functional approach, so approximations by different types of distributions are compared in sense of different metrics. The results of application of the implemented software tools are demonstrated on the example of the Potsdam precipitation events.

A Novel Weighted Doppler Centroid Estimation Approach Based on Electromagnetic Scattering Model for Multichannel in Azimuth HRWS SAR System

Similar to the conventional squint mode synthetic aperture radar (SAR) imaging processing, the estimation of Doppler centroid is one key problem for the low-squint-mode (within the range of [−5°, 5°]) multichannel in an azimuth high-resolution and wide-swath (HRWS) SAR system. In this paper, based on the electromagnetic scattering model, a novel weighted Doppler centroid estimation approach is proposed for the multichannel in the azimuth HRWS (MC-HRWS) SAR system. First, Maxwell’s equations are employed to derive the echo model for the multichannel SAR system. Then, an improved backscattering model is presented based on the small perturbation method and the available echo model, which is adopted to produce the weights for Doppler centroid estimation. More importantly, in order to improve the precision of Doppler centroid estimation, a weighted ambiguity-free Doppler centroid estimation approach is proposed, where the range-variant characteristic of Doppler centroid is employed to resolve the ambiguity number of Doppler centroid, and the echoes from different range bins with different signal-to-noise ratios are considered. In addition, the Cramer–Rao low bound of the estimated Doppler centroid is also discussed in this paper. The effectiveness of the proposed Doppler centroid estimation approach is verified via simulated and real measured low-squint-mode MC-HRWS SAR data.

Mixed Mode Radar Coincidence Imaging With Hybrid Excitation Radar Array

A novel mixed mode radar coincidence imaging (MMRCI) system, which combines the coherent radar detection and incoherent radar coincidence imaging function during the radar coincidence imaging (RCI) process, is presented in this paper. The target location function of the MMRCI system gives the target position information synchronously with imaging tests, and this makes the MMRCI work independently with low target position estimation error. In the MMRCI system, the resolution of target distance estimation is proportional to the whole bandwidth assisted by the double-frequency linear frequency modulation (DFLFM) signal when the occupied bandwidth of the DFLFM signal is less than one-tenth of the whole bandwidth. The target position estimation error caused by the target motion can be reduced effectively so that the image quality of the moving target can be guaranteed in a longer detection time. The RCI is reconstructed using the incoherent part of the transmitting signal. A general relationship between the spatial correlation of the random radiation field in the imaging plane and the deployments of linear random source is analyzed in the paper. The imaging efficiency is ensured through the grading imaging method based on the imaging requirement. The effectiveness of the MMRCI method is validated via a set of simulations and experiments, including superresolution RCI of both stationary and moving targets in various scenes using the convex optimization algorithms.

Fast matrix inversion and determinant computation for Polarimetric Synthetic Aperture Radar

This paper introduces a fast algorithm for simultaneous inversion and determinant computation of small sized matrices in the context of fully Polarimetric Synthetic Aperture Radar (PolSAR) image processing and analysis. The proposed fast algorithm is based on the computation of the adjoint matrix and the symmetry of the input matrix. The algorithm is implemented in a general purpose graphical processing unit (GPGPU) and compared to the usual approach based on Cholesky factorization. The assessment with simulated observations and data from an actual PolSAR sensor show a speedup factor of about two when compared to the usual Cholesky factorization. Moreover, the expressions provided here can be implemented in any platform.

An Ice-Drift Estimation Algorithm Using Radar and Ship Motion Measurements

This paper presents a novel automatic real-time remote sensing algorithm that uses radar images and global positioning satellite system measurements to estimate the ice-drift velocity vector in a region around a free-floating and potentially moving vessel. It is motivated by the low image frequency of satellite systems together with the inconvenience of deploying and retrieving ice trackers (beacons) on the ice. The algorithm combines radar image processing with two Kalman filters to produce the estimated local drift vector decoupled from the ship motion. The proposed design is verified using a full-scale data set from an ice management operation north of Svalbard in 2015. It is found that the performance of the algorithm is comparable with that of trackers on the ice.

Generic framework for vessel detection and tracking based on distributed marine radar image data

Situation awareness is understood as a key requirement for safe and secure shipping at sea. The primary sensor for maritime situation assessment is still the radar, with the AIS being introduced as supplemental service only. In this article, we present a framework to assess the current situation picture based on marine radar image processing. Essentially, the framework comprises a centralized IMM–JPDA multi-target tracker in combination with a fully automated scheme for track management, i.e., target acquisition and track depletion. This tracker is conditioned on measurements extracted from radar images. To gain a more robust and complete situation picture, we are exploiting the aspect angle diversity of multiple marine radars, by fusing them a priori to the tracking process. Due to the generic structure of the proposed framework, different techniques for radar image processing can be implemented and compared, namely the BLOB detector and SExtractor. The overall framework performance in terms of multi-target state estimation will be compared for both methods based on a dedicated measurement campaign in the Baltic Sea with multiple static and mobile targets given.

Development of Adaptive Threshold and Data Smoothening Algorithm for GPR Imaging

There are many approaches available to separate the background and foreground in image processing applications. Currently, researchers are focusing on wavelet De-noising, curvelet threshold, Edge Histogram Descriptor threshold, Otsu thresholding, recursive thresholding and adaptive progressive thresholding. In fixed and predictable background conditions, above techniques separate background and foreground efficiently. In a common scenario, background reference is blind due to soil surface moisture content and its non-linearity. There are many methodologies proposed from time to time by researchers to solve this blind reference background separation. But challenges still now remain, because there are two major problems in ground penetrating radar imaging such as targets like ground enhances the false alarm and non-metallic target detection, where the threshold decision is a critical task. In this paper, a novel real time blind adaptive threshold algorithm is proposed for ground penetrating radar image processing. The blind threshold was decided to use normal random variable variance and image data variance. Further, the image was smoothened by random variance ratio to image data variance. Experimental results showed satisfactory results for the background separation and smoothening the targeted image data with the proposed algorithm.

Bidimensional empirical mode decomposition method for image processing in sensing system

Recent developments in image processing play a significant role in processing the images clearly. Undoubtedly, identifying the target against clutter is another challenging task in Ground Penetrating Radar (GPR) image processing. To successfully distinguish targets from surrounding clutter, a novel Bidimensional Empirical Mode Decomposition (BEMD) system is proposed. The overall process can be broken down into four steps. First, images are decomposed using BEMD arithmetic to extract the Intrinsic Mode Functions (IMFs). Then, these IMFs could be examined using gradual single frequency signals to bring out and highlight the physical meaning of instantaneous amplitudes and instantaneous frequencies. To find the best-fitting hyperbolas, the object position of the vertex is estimated using maximum point estimation method and velocity is estimated using a minimum entropy method. Finally, the location of reflection hyperbolas can be extracted clearly. The simulation and experimental images show the effectiveness of this method in identifying targets.

Image processing and analysis of radar and lidar data: new discoveries in Verona southern lowland (Italy)

ABSTRACTThis contribution proposes an evaluation of lidar and radar data processing and its potential in revealing archaeological features within a level plain environment, the southern lowland of Verona (Italy), focusing on evidences dating back to the Bronze Age. Many archaeological sites in the research area, including some of the most outstanding settlements of Terramare Culture, were identified or at least examined through aerial photo observation. Even if in several occasions modern agricultural activities contributed to the discoveries, bringing to the surface artifacts and scrapes of buried layers, this kind of impact has also been progressively deteriorating the archaeological record, hence the proto-historic landscape is now discernible through evanescent marks which cannot be always detected using customary optical sensors. Lidar and radar data analysis has then been considered as an alternative, non-invasive method of investigation on such a vast area.

A Characterization of the Domain of Beta-Divergence and Its Connection to Bregman Variational Model

In image and signal processing, the beta-divergence is well known as a similarity measure between two positive objects. However, it is unclear whether or not the distance-like structure of beta-divergence is preserved, if we extend the domain of the beta-divergence to the negative region. In this article, we study the domain of the beta-divergence and its connection to the Bregman-divergence associated with the convex function of Legendre type. In fact, we show that the domain of beta-divergence (and the corresponding Bregman-divergence) include negative region under the mild condition on the beta value. Additionally, through the relation between the beta-divergence and the Bregman-divergence, we can reformulate various variational models appearing in image processing problems into a unified framework, namely the Bregman variational model. This model has a strong advantage compared to the beta-divergence-based model due to the dual structure of the Bregman-divergence. As an example, we demonstrate how we can build up a convex reformulated variational model with a negative domain for the classic nonconvex problem, which usually appears in synthetic aperture radar image processing problems.

A Sparse SAR Imaging Method Based on Multiple Measurement Vectors Model

In recent decades, compressive sensing (CS) is a popular theory for studying the inverse problem, and has been widely used in synthetic aperture radar (SAR) image processing. However, the computation complexity of CS-based methods limits its wide applications in SAR imaging. In this paper, we propose a novel sparse SAR imaging method using the Multiple Measurement Vectors model to reduce the computation cost and enhance the imaging result. Based on using the structure information and the matched filter processing, the new CS-SAR imaging method can be applied to high-quality and high-resolution imaging under sub-Nyquist rate sampling with the advantages of saving the computational cost substantially both in time and memory. The results of simulations and real SAR data experiments suggest that the proposed method can realize SAR imaging effectively and efficiently.

Fast Detection of Oil Spills and Ships Using SAR Images

In this paper, we show the capabilities of a new maritime control system based on the processing of COSMO-SkyMed Synthetic Aperture Radar (SAR) images. This system aims at fast detection of ships that may be responsible for illegal oil dumping. In particular, a novel detection algorithm based on the joint use of the significance parameter, wavelet correlator and a two-dimensional Constant False Alarm Rate (2D-CFAR) is designed. Results show the effectiveness of such algorithms, which can be used by the maritime authorities to have a faster although still reliable response. The proposed algorithm, together with the short revisit time of the COSMO-SkyMed constellation, can help with tracking the scenario evolution from one acquisition to the next.

APPLICATION OF ANISOTROPIC FILTRATION IN PROCESSING OF W-BAND RADAR IMAGES

The article deals with the application of the anisotropic filtering algorithm for preliminary processing of W-band radar images. A comparative analysis of the parameters of the algorithm (sets of apertures) has been performed and conclusions have been drawn on the prospects for its further investigation in the conditions of the set tasks.

Application of Mixture Regression for Improved Polarimetric SAR Speckle Filtering

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.