Synthetic Aperture Radar Speckle Noise Research Articles

Dual Consistency Alignment Based Self-Supervised Learning for SAR Target Recognition With Speckle Noise Resistance

Deep learning based on Convolutional Neural Networks (CNN) has been widely applied in Synthetic Aperture Radar (SAR) target recognition and made significant progress. However, due to the physical effects of the equipment used to collect images, various degrees of speckle noise will be introduced into SAR images. Traditional CNN based SAR target recognition methods are premised on the same noise intensity in the training and testing set, which is contrary to the target recognition in practice. To alleviate this problem, we propose a novel speckle noise resistant framework for SAR target recognition, called Dual Consistency Alignment based Self-Supervised Learning(DCA-SSL). Firstly, original SAR images are randomly added to speckle noise with different thresholds through multiplicative noise, after which contrastive pre-training is performed on unlabeled data. During this period, we combine instance pseudo-label consistency alignment and feature consistency alignment to align multiple threshold speckle noise views with original views under the same targets. Finally, the pre-trained model is migrated to the downstream SAR speckle noise target recognition task. In this article, speckle noise modeling is conducted based on MSTAR data testing set, and experiment results reveal that this method can adapt to different intensities of speckle noise, is robust to modeled SAR image recognition and maintains a high recognition rate even in small sample learning. Our code is available at: <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/Jordan-Liao/DCA-SSL</uri> .

CFOG-like image registration algorithm based on 3D-structural feature descriptor for suburban optical and SAR

Owing to nonlinear radiation differences between optical and synthetic aperture radar (SAR) images, and the inevitable SAR speckle noise, the registration of optical and SAR images is challenging work. This paper proposes a novel pixel-wise 3D-structural feature descriptor based on channel-features-of-oriented-gradients (CFOG), thus, it inherits the fast and robust advantages of CFOG. First, the Block-Harris and grid points feature extraction method is used to obtain sufficient uniformly distributed keypoints from the reference images. Then, the new feature descriptor based on the combination of first-order oriented gradients and second-order gradients (FOSG) is proposed. Specifically, two different operators, the Sobel and the ratio of exponentially weighted averages are utilized to calculate the gradient for the optical and SAR images, respectively. On this basis, the first-order oriented gradients and second-order gradients are calculated by a 3D-Gaussian filter and then normalized along the Z-direction to form the FOSG by superposition. Finally, template matching based on the sum of squared differences similarity metric is implemented in the frequency domain. Our method is experimentally compared with five state-of-the-art methods using six pairs of suburban optical and SAR images. The results show that the proposed algorithm is robust to nonlinear radiation differences and SAR speckle noise, and achieves high accuracy.

A Rotation-Invariant Optical and SAR Image Registration Algorithm Based on Deep and Gaussian Features

Traditional feature matching methods of optical and synthetic aperture radar (SAR) used gradient are sensitive to non-linear radiation distortions (NRD) and the rotation between two images. To address this problem, this study presents a novel approach to solving the rigid body rotation problem by a two-step process. The first step proposes a deep learning neural network named RotNET to predict the rotation relationship between two images. The second step uses a local feature descriptor based on the Gaussian pyramid named Gaussian pyramid features of oriented gradients (GPOG) to match two images. The RotNET uses a neural network to analyze the gradient histogram of the two images to derive the rotation relationship between optical and SAR images. Subsequently, GPOG is depicted a keypoint by using the histogram of Gaussian pyramid to make one-cell block structure which is simpler and more stable than HOG structure-based descriptors. Finally, this paper designs experiments to prove that the gradient histogram of the optical and SAR images can reflect the rotation relationship and the RotNET can correctly predict them. The similarity map test and the image registration results obtained on experiments show that GPOG descriptor is robust to SAR speckle noise and NRD.

A Robust Algorithm Based on Phase Congruency for Optical and SAR Image Registration in Suburban Areas

Automatic registration of optical and synthetic aperture radar (SAR) images is a challenging task due to the influence of SAR speckle noise and nonlinear radiometric differences. This study proposes a robust algorithm based on phase congruency to register optical and SAR images (ROS-PC). It consists of a uniform Harris feature detection method based on multi-moment of the phase congruency map (UMPC-Harris) and a local feature descriptor based on the histogram of phase congruency orientation on multi-scale max amplitude index maps (HOSMI). The UMPC-Harris detects corners and edge points based on a voting strategy, the multi-moment of phase congruency maps, and an overlapping block strategy, which is used to detect stable and uniformly distributed keypoints. Subsequently, HOSMI is derived for a keypoint by utilizing the histogram of phase congruency orientation on multi-scale max amplitude index maps, which effectively increases the discriminability and robustness of the final descriptor. Finally, experimental results obtained using simulated images show that the UMPC-Harris detector has a superior repeatability rate. The image registration results obtained on test images show that the ROS-PC is robust against SAR speckle noise and nonlinear radiometric differences. The ROS-PC can tolerate some rotational and scale changes.

A Bayesian Joint Decorrelation and Despeckling of SAR Imagery

Despeckling of synthetic aperture radar (SAR) is a known research challenge. A novel solution to this problem has been developed and evaluated via an iterative maximum a posterior estimation incorporating a Bayesian joint decorrelation and despeckling based on a correlation model. This model realistically explores the physical correlation process of SAR speckle noise and is determined automatically via Bayesian estimation in the log-Fourier domain. A patchwise computation is used to account for the spatial nonstationarity associated with SAR image data. The proposed approach is compared to the existing despeckling techniques using both simulated and real SAR data, and the experimental results demonstrate the improvement in preserving the structural details while suppressing speckle noise.

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.

Edge-detection-based filter for SAR speckle noise reduction

Some adaptive filters, such as the Kuan, Lee, minimum mean square error (MMSE) and Frost filters, have been tested on synthetic aperture radar (SAR) data without considering the level of homogeneity in the pixels. Therefore, they degrade the spatial resolution of images and smooth details, while also decreasing the speckle noise level. There are other filters, such as the enhanced Lee and gamma maximum a posteriori (MAP), that utilize the level of homogeneity, but they cannot adequately suppress speckle noise. In addition to these weaknesses, pixels surrounding a point scatterer are also treated as point scatterers due to inadequacy of the method based on evaluating the coefficient of variation for differentiating between them and the point scatterer. We have developed a new method based on the assessment of similarity of homogeneity levels in the image, incorporating edge-detection filters to identify meaningful features and an algorithm to filter the pixels surrounding point scatterers. This method, called the UNSW (University of New South Wales) adaptive filter (UAF), was compared to nine filters using different quantitative and qualitative methods. The results show the ability of the UAF to simultaneously reduce speckle and preserve details as well as its ability to filter more pixels. The effect of increasing the damping factor on speckle noise reduction performance has also been assessed using this method.

Multipolarization SAR Data for Surface Feature Delineation and Forest Vegetation Characterization

This paper presents the utility of multipolarization Synthetic Aperture Radar (SAR) data for surface feature delineation and forest vegetation characterization. Three channels of ratioed data (VV/HH, VH/HH, and VH/VV) are generated from the HH, VV, and VH polarization data (V = vertical, H = horizontal). The ratioed data are linearly stretched to yield a digital number within a range of 0 to 255. The techniques for reducing SAR speckle noise and for measuring the degree of separation are discussed. For surface feature delineation, the results indicate that cross polarization as well as cross polarization ratioed data better delineate those surface features that are difficult to separate by like polarization data. The results suggest using a median value filtering technique to reduce within-plot data fluctuation to increase the separability measure. For forest vegetation characterization, the results indicate that multipolarization SAR data may be used to estimate forest properties such as total-tree biomass, basal area, and tree height.