Synthetic Aperture Radar Speckle Research Articles

Ground Target Classification in Noisy SAR Images Using Convolutional Neural Networks

Speckle noise is an inherent but annoying property in the synthetic aperture radar (SAR) imaging. In this paper we investigate the influence of speckle on the classical convolutional neural network (CNN) for SAR target classification. Then a dual stage coupled CNN architecture, named despeckling and classification coupled CNNs (DCC-CNNs), is proposed to distinguish multiple categories of ground targets in SAR images with strong and varying speckle. It first applies the despeckling sub-network for noise reduction. After that, residual speckle features as well as target information would be learned by the classification sub-network in order to solve the noise robustness problem of CNN. Besides, a new quantitative measure is developed for the quality assessment of SAR target images. It takes into account structural properties of the speckled SAR image of the target of interest and consistency with visual perception. Finally, a series of comparative experiments and discussions are carried out to validate the proposed assessment criterion and DCC-CNNs. Using synthetic SAR images based on the public MSTAR datasets, results show that the overall classification accuracy for ten ground target classes could be higher than 82% at a variety of speckle noise levels.

Adaptive non-local level-set model for despeckling and deblurring of synthetic aperture radar imagery

ABSTRACTIn this article, we modify Mumford–Shah level-set model to handle speckles and blur in synthetic aperture radar (SAR) imagery. The proposed model is formulated using a non-local regularization framework. Hence, the model duly cares about local gradient oscillations (corresponding to the fine details/textures) during the evolution process. It is assumed that the speckle intensity is gamma distributed, while designing a maximum a posteriori estimator of the functional. The parameters of the gamma distribution (i.e. scale and shape) are estimated using a maximum likelihood estimator. The regularization parameter of the model is evaluated adaptively using these (estimated) parameters at each iteration. The split-Bregman iterative scheme is employed to improve the convergence rate of the model. The proposed and the state-of-the-art despeckling models are experimentally verified and compared using a large number of speckled and blurred SAR images. Statistical quantifiers are used to numerically evaluate the performance of various models under consideration.

A Hybrid Method of SAR Speckle Reduction Based on Geometric-Structural Block and Adaptive Neighborhood

Given the improvement of synthetic aperture radar (SAR) imaging technologies, the resolution of SAR image is largely improved and the variation of backscatter amplitude should be considered in SAR image processing. In this paper, considering the spatial geometric properties of SAR image in gray pixel space and the sample selection in the estimation of true signal, local directional property of each pixel is explored with the help of SAR sketching method, and two specially designed filters are integrated for adaptive speckle reduction of SAR images. Specifically, based on the sketch map of a SAR image, the orientation of the sketch point lying at each sketch segment is assigned to the corresponding pixel, and thus all pixels of the SAR image are classified as the directional pixels and the nondirectional pixels. For the directional pixels, given the significant directionality of its neighborhood, a geometric-structural block (GB) is built to center on it and GB-wised nonlocal means filter is designed to estimate the true values of all pixels contained in the GB. Moreover, using the local orientation, the whole image is adopted as the searching range to search the similar GBs. For the nondirectional pixels, based on the locally estimated equivalent number of looks, a novel pixel-based metric is proposed to determine the local adaptive neighborhood (AN) with which an AN-based filter is developed to estimate its true value. Besides, since some nondirectional pixels are contained in GBs, a Bayesian-based fusion strategy is designed for the fusion of their estimated values. In the experiments, three synthetic speckled images and five real SAR images [obtained with different resolutions (e.g., 3, 1, and 0.1 m) and different bands (e.g., X-band, C-band, and Ka-band)] are used for evaluation and analysis. Owing to the usage of local spatial geometric property and the combination of two different filters, the proposed method shows a reasonable performance among the comparison methods, in terms of the speckle reduction and the details’ preservation.

Using BayesShrink, BiShrink, Weighted BayesShrink, and Weighted BiShrink in NSST and SWT for Despeckling SAR Images

Synthetic aperture radar (SAR) images are inherently degraded by multiplicative speckle noise where thresholding-based methods in the transform domain are appropriate. Being sparse, the coefficients in the transformed domain play a key role in the performance of any thresholding methods. It has been shown that the coefficients of nonsubsampled shearlet transform (NSST) are sparser than those of stationary wavelet transform (SWT) for either clean or noisy images. Therefore, it is expected that thresholding-based methods in NSST outperform those in the SWT domain. In this paper, BayesShrink, BiShrink, weighted BayesShrink, and weighted BiShrink in NSST and SWT domains are compared in terms of subjective and objective image assessment. As BayesShrink try to find the optimum threshold for every subband, BiShrink uses coefficients, name “parent,” to clean up coefficients called “child,” and the weighted methods consider the coefficients’ noise efficiency, which implies that subbands in the transform domain may be affected by noise differently. Two models for considering the parent in the NSST domain are proposed. In addition, for both BayesShrink and BiShrink, considering the weighting factor (coefficients noise efficiency) would improve the performance of the corresponding methods as well. Experimental results show that the weighted-BiShrink despeckling approach in the NSST domain gives an outstanding performance when tested with both artificially speckled images and real SAR images.

An Integrated Method for Simulation of Synthetic Aperture Radar (SAR) Raw Data in Moving Target Detection

SAR (Synthetic Aperture Radar) raw data simulation has a significant role in the evaluation of newly-proposed methods for the estimation of moving target parameters. The evaluation of methods in different cases emphasizes the importance of the need to have fast simulators. Using reverse SAR imaging methods for the raw data simulations has achieved good results in the simulation of the static targets, but for the simulation of moving targets these methods have a few shortcomings. In this paper, we propose a method to simulate a speckled scene with moving targets in the hybrid domain. First, the scene is simulated, including speckle, which is statistically in accordance with real SAR image behavior. Then, a reverse imaging algorithm (inverse chirp scaling) was applied on the scene to generate the SAR raw data. The moving target simulation was also done in the time-domain as the next step. Finally, the results from two prior steps were superposed to generate the SAR raw data with moving targets. All steps of the proposed method were evaluated separately. The speckle procedure was evaluated by comparing the speckled SAR image before the simulation and the image of the SAR simulated raw data. The results show similar variations in real and imaginary parts of these data. The correlation between the reflectivity map and the SAR images after the simulation was calculated and the obtained correlation coefficient was about 0.95 for different images. The final data were further analyzed for the displacement of moving targets’ positions. The results show similar displacement between moving target SAR raw data with a background and without a background.

G0-WISHART DISTRIBUTION BASED CLASSIFICATION FROM POLARIMETRIC SAR IMAGES

Abstract. Enormous scientific and technical developments have been carried out to further improve the remote sensing for decades, particularly Polarimetric Synthetic Aperture Radar(PolSAR) technique, so classification method based on PolSAR images has getted much more attention from scholars and related department around the world. The multilook polarmetric G0-Wishart model is a more flexible model which describe homogeneous, heterogeneous and extremely heterogeneous regions in the image. Moreover, the polarmetric G0-Wishart distribution dose not include the modified Bessel function of the second kind. It is a kind of simple statistical distribution model with less parameter. To prove its feasibility, a process of classification has been tested with the full-polarized Synthetic Aperture Radar (SAR) image by the method. First, apply multilook polarimetric SAR data process and speckle filter to reduce speckle influence for classification result. Initially classify the image into sixteen classes by H/A/α decomposition. Using the ICM algorithm to classify feature based on the G0-Wshart distance. Qualitative and quantitative results show that the proposed method can classify polaimetric SAR data effectively and efficiently.

MuLoG, or How to Apply Gaussian Denoisers to Multi-Channel SAR Speckle Reduction?

Speckle reduction is a longstanding topic in synthetic aperture radar (SAR) imaging. Since most current and planned SAR imaging satellites operate in polarimetric, interferometric, or tomographic modes, SAR images are multi-channel and speckle reduction techniques must jointly process all channels to recover polarimetric and interferometric information. The distinctive nature of SAR signal (complex-valued, corrupted by multiplicative fluctuations) calls for the development of specialized methods for speckle reduction. Image denoising is a very active topic in image processing with a wide variety of approaches and many denoising algorithms available, almost always designed for additive Gaussian noise suppression. This paper proposes a general scheme, called MuLoG (MUlti-channel LOgarithm with Gaussian denoising), to include such Gaussian denoisers within a multi-channel SAR speckle reduction technique. A new family of speckle reduction algorithms can thus be obtained, benefiting from the ongoing progress in Gaussian denoising, and offering several speckle reduction results often displaying method-specific artifacts that can be dismissed by comparison between results.

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.

Pixel Classification of SAR ice images using ANFIS-PSO Classifier

Synthetic Aperture Radar (SAR) is playing a vital role in taking extremely high resolution radar images. It is greatly used to monitor the ice covered ocean regions. Sea monitoring is important for various purposes which includes global climate systems and ship navigation. Classification on the ice infested area gives important features which will be further useful for various monitoring process around the ice regions. Main objective of this paper is to classify the SAR ice image that helps in identifying the regions around the ice infested areas. In this paper three stages are considered in classification of SAR ice images. It starts with preprocessing in which the speckled SAR ice images are denoised using various speckle removal filters; comparison is made on all these filters to find the best filter in speckle removal. Second stage includes segmentation in which different regions are segmented using K-means and watershed segmentation algorithms; comparison is made between these two algorithms to find the best in segmenting SAR ice images. The last stage includes pixel based classification which identifies and classifies the segmented regions using various supervised learning classifiers. The algorithms includes Back propagation neural networks (BPN), Fuzzy Classifier, Adaptive Neuro Fuzzy Inference Classifier (ANFIS) classifier and proposed ANFIS with Particle Swarm Optimization (PSO) classifier; comparison is made on all these classifiers to propose which classifier is best suitable for classifying the SAR ice image. Various evaluation metrics are performed separately at all these three stages.

Speckle-reducing scale-invariant feature transform match for synthetic aperture radar image registration

The anisotropic scale space (ASS) is often used to enhance the performance of a scale-invariant feature transform (SIFT) algorithm in the registration of synthetic aperture radar (SAR) images. The existing ASS-based methods usually suffer from unstable keypoints and false matches, since the anisotropic diffusion filtering has limitations in reducing the speckle noise from SAR images while building the ASS image representation. We proposed a speckle reducing SIFT match method to obtain stable keypoints and acquire precise matches for the SAR image registration. First, the keypoints are detected in a speckle reducing anisotropic scale space constructed by the speckle reducing anisotropic diffusion, so that speckle noise is greatly reduced and prominent structures of the images are preserved, consequently the stable keypoints can be derived. Next, the probabilistic relaxation labeling approach is employed to establish the matches of the keypoints then the correct match rate of the keypoints is significantly increased. Experiments conducted on simulated speckled images and real SAR images demonstrate the effectiveness of the proposed method.

A novel approach for object-based change image generation using multitemporal high-resolution SAR images

ABSTRACTObject-based change detection offers a unique approach for high-resolution images to capture meaningful detailed change information while suppressing noise in change detection results. In this approach, mean intensities of objects are commonly used as a feature and images comparison is carried out based on simple mathematical operations such as ratioing. The strong intensity variations within an object – a consequence of high spatial resolution – combined with synthetic aperture radar (SAR) image speckle degrade the accuracy of object mean intensity estimate, and consequently, affect the quality of the estimated object-based change image. A change quantification approach that takes into account the characteristics of high-resolution SAR images, that is, SAR speckle and the strong intensity variation, is proposed. By descending to the pixel level, a new representation of change data (i.e. the change signal) is proposed. With this representation, change quantification boils down to measuring the roughness of the change signal. Two techniques to assess the intensity of change at the object-level, based on Fourier and wavelet transforms (WT) of the change signal, are proposed. Their main advantages lie in their ability to capture the dominant change behaviour of the object, while being insusceptible to irrelevant disturbances. The proposed approach is evaluated using two multitemporal data sets of TerraSAR-X images acquired over Beijing and Shanghai. The qualitative and quantitative analyses of the results demonstrate the superior discrimination power of the proposed change variables compared with the object-based modified ratio (MR) and the absolute log ratio (LR) images.

Local Contrast Based Adaptive SAR Speckle Filter

In this paper, we propose an adaptive filtering technique for Synthetic Aperture Radar (SAR) images. A new windowing technique is introduced where the total window is divided into five equal sized overlapping sub-windows. The pixel to be filtered is a part of each of these sub-windows. A weighted mean of all sub-windows is computed for the pixel under consideration. The weights are accounted from a measure of heterogeneity calculated for each sub-windows. The filter is able to adapt automatically and adjust the speckle suppression strength based on local statistics. This allows the filter to preserve edges while strongly suppressing speckle over homogeneous areas. The proposed filter was compared with some well known SAR filtering techniques in terms of speckle suppression and edge preservation ability. Several experiments were performed on datasets acquired from both air-borne and space-borne SAR platforms. Some well known indices were used for quantitative comparison with other filters. Among the filters compared, the proposed filter shows good speckle suppression ability while still exhibiting reasonable edge preservation ability.

A Novel Asymmetrical Probability Density Function for Modeling Log-Ratio SAR Images

Multitemporal synthetic aperture radar (SAR) images are often employed for change detection use due to the stable and periodic monitoring of the ground by SAR sensors. Because of the multiplicative nature of speckle noise in SAR data, the comparison between multitemporal images is often carried out according to a standard log-ratio operator. Therefore, in this letter, the probability density function (pdf) of the log-ratio SAR image is studied. Assuming that the SAR speckle is gamma distributed with unit mean, we present the analytic distribution form of the log-ratio image. The distribution is capable of capturing the nonsymmetrical feature of the histogram of the log-ratio data. At the same time, we provide a moment-based method for estimating the parameters of the pdf using empirical data. In the experiments, we use chi-square goodness-of-fit test and one-sample Kolmogorov–Smirnov test to evaluate the performance of the approach. The Gaussian distribution and the generalized Gaussian distribution are also employed for comparison. Experiment on two real SAR data sets demonstrates the effectiveness of the proposed method.

Assessment of SAR speckle filters in the context of object-based image analysis

ABSTRACTThe initial step in most object-based classification methodologies is the application of a segmentation algorithm to define objects. In the context of synthetic aperture radar (SAR) image analysis, the presence of speckle noise might hamper the segmentation quality. The aim of this study is to assess the segmentation performance of SAR images when no filter or different filters are applied before segmentation. In particular, the performance of the mean-shift segmentation algorithm combined with different adaptive and non-adaptive filters is assessed based on both synthetic and natural SAR images. Studied filters include the non-adaptive Boxcar filter and four adaptive filters: the well-known Refined Lee filter and three recently proposed non-local filters differing, in particular, in their dissimilarity criteria: the Hellinger and the Kullback–Leibler filters are based on stochastic distances, whereas the NL-SAR filter is based on the generalized likelihood ratio. Two measures were used for quality assessment: -index and -index. Over-segmentation was assessed by the -index, the ratio of the resulting number of segments to the number of connected components of the ground-truth classes. The accuracy of the best possible classification given on the segmentation result was assessed with ground truth information by maximizing the -index. A Monte Carlo experiment conducted on synthetic images shows that the quality measures significantly differ for the applied filters. Our results indicate that the use of an adaptive filter improves the performance of the segmentation. In particular, the combination of the mean-shift segmentation algorithm with the NL-SAR filter gives the best results and the resulting process is less sensitive to variations in the mean-shift operational parameters than when applying other filters or no filter. The results obtained may help improve the reliability of land-cover classification analyses based on an object-based approach on SAR data.

Tracking of Coastal Swell Fields in SAR Images for Sea Depth Retrieval: Application to ALOS L-Band Data

Swell propagation in shallow water sets specific relationships between water depth and swell parameters like swell wavelength and period. These relationships allow coastal bathymetry to be retrieved if swell parameters are measured. Synthetic aperture radar (SAR) is able to image swell waves, and spectral analysis of SAR images is a well-known approach for measuring swell parameters. However, owing to nonlinearities in SAR imaging, speckle and noise, spectral analysis can result in significant bathymetric errors. The paper individuates the conditions in which linear imaging is achieved and presents an algorithm able to preserve the accuracy of the calculated bathymetry against image speckle and noise. The proposed approach includes: 1) image resizing and filtering before spectral analysis; 2) limitations to the domain of the spectral analysis; and 3) spatial smoothing of the estimated parameters. The algorithm is tested on L-band ALOS PALSAR data collected over coastal regions in the Gulf of Naples, Italy, showing that swell can be properly tracked from open sea to shoreline. Dense coverage and submetric accuracy is thus achieved.

Unbiased-average minimum biased diffusion speckle denoising approach for synthetic aperture radar images

Means of synthetic aperture radar (SAR) images represent the radiation densities of scenes, and the preservation of means is significant in speckle denoising for the application of SAR images. We provide an improved scheme of the minimum biased diffusion (MinBAD) algorithm for speckle denoising using partial differential equations. Considering the characteristics of SAR speckle and the radiation accuracy for postprocessing needs, several improvements such as normalization, homomorphic transformation, and average-preserving processing are introduced into the MinBAD algorithm. Besides the equivalent number of looks and edge preserving index, a new index, radiation accuracy error, is defined to evaluate the denoising effect. Experimental results for both artificial images and real SAR images are used to validate the performance of the proposed unbiased-average MinBAD speckle reducing approach.

Polarimetric SAR Speckle Filtering and the Extended Sigma Filter

The advancement of synthetic aperture radar (SAR) technology with high-resolution and quad-polarization data demands better and efficient polarimetric SAR (PolSAR) speckle-filtering algorithms. Two requirements on PolSAR speckle filtering are proposed: 1) speckle filtering should be applied to distributed media only, and strong hard targets should be kept unfiltered; and 2) scattering mechanism preservation should be taken into consideration, in addition to speckle reduction. The purpose of this paper is twofold: 1) to propose an effective algorithm that is an extension of the improved sigma filter developed for single-polarization SAR; and 2) to investigate speckle characteristics and the need for speckle filtering for very high resolution (decimeter) PolSAR data. The proposed filter was specifically developed to account for the aforementioned two requirements. Its effectiveness is demonstrated with Jet Propulsion Laboratory airborne synthetic aperture radar data, and comparisons are made with a boxcar filter, the refined Lee filter, and a Wishart-based nonlocal filter. For very high resolution PolSAR systems, such as the German Aerospace Center F-SAR and Japanese Pi-SAR2, with decimeter spatial resolution, we found that the complex Wishart distribution is still valid to describe PolSAR speckle characteristics of distributed media and that speckle filtering may be needed depending on the size of objects to be analyzed. F-SAR X-band data with 25-cm resolution is used for illustration.

Change detection in synthetic aperture radar images based on non-local means with ratio similarity measurement

Image change detection is of widespread interest due to a large number of applications in diverse disciplines. In this study, a novel change detection approach for synthetic aperture radar (SAR) images based on a non-local means algorithm is proposed. A non-local means technique is introduced to generate a difference image by using complete information from a pair of observed images. To take the characteristics of SAR images into account, a new ratio-based relativity measurement between two speckled SAR image patches based on a ratio distance is proposed. Theoretical analysis indicates that the ratio distance is valid for SAR images. The probability density function of the ratio distance is deduced to map the distance into a relativity value. Furthermore, the ratio distance and the probability density function are both parameter-free. The new non-local means technique is successfully applied to extend the classical mean-ratio detector for SAR image detection. Experimental results on real SAR images show that the proposed approach is robust to speckle noise and effective for the detection of change information between multitemporal SAR images.

Local Maximal Homogeneous Region Search for SAR Speckle Reduction With Sketch-Based Geometrical Kernel Function

With the flourish of the nonlocal mean method, the neighborwise similarity metric is widely applied in speckle reduction for its robust performance on the search of similar samples. In this metric, an isotropic kernel function is usually chosen to aggregate the corresponding pixels' distance between two neighborhoods. It means that the kernel function is considered as the explanation of the local spatial relationship at each pixel. However, for anisotropic features (such as edges and lines), a strong relationship exists along their directions rather than across them, so the isotropic kernel is not suitable to explain the spatial relationship around these features. Meanwhile, due to the inherent speckle in synthetic aperture radar (SAR) images, the discrimination and exploration of the geometrical properties of anisotropic features are important for the construction of adaptive kernel function. In this paper, the sketch map which is a representation of the sketch information of SAR images is extracted as the criterion for designing the kernel function. Meanwhile, due to the properties of symmetric and maximal self-similarity, a modified ratio distance is proposed and used jointly with the constructed kernel function as a similarity metric. Then, under the local stationary assumption, the local maximal homogeneous region of each pixel is searched by using the region growing method with the proposed metric. Moreover, maximal likelihood rule is used within the region for the estimation of true value. From the experiments on the synthetic and real SAR images, a promising performance in terms of speckle reduction and preservation of the details is achieved by our proposed method.

A KPCA texture feature model for efficient segmentation of RADARSAT-2 SAR sea ice imagery

Sea ice information obtained from synthetic aperture radar (SAR) images is crucial for ensuring safe marine navigation and supporting climate change studies in polar regions. We propose a kernel principal component analysis (KPCA) local texture feature model for efficient sea ice segmentation. The proposed KPCA texture feature model is significant for several reasons. First, it takes into account the multiplicative nature of SAR speckle noise. The resulting KPCA features therefore assume independent and identically distributed (i.i.d.) Gaussian-like noise, which satisfies the assumptions inherent in classical statistical models such as the k-means algorithm. Second, the KPCA features are compact and statistically independent and therefore capable of reducing data redundancy. Third, the KPCA texture features are discriminative. Finally, the extraction of KPCA features requires less computation. Based on the KPCA model, a segmentation scheme is implemented in four steps. First, local patch-based texture descriptors are extracted from the image. Second, KPCA is performed on the patch-based texture descriptors to obtain compact and discriminative texture features with Gaussian-like noise characteristics. Third, the k-means algorithm is performed on extracted KPCA features to segment the image. Finally, a neighbourhood-based majority voting scheme is employed to determine the final label of each pixel. We compared the proposed method with several other popular sea ice segmentation approaches. The results demonstrated that our method is robust to speckle noise level, is accurate and fast and may thus better support the operational segmentation of sea ice in SAR imagery.