Search

Matching in high-resolution synthetic aperture radar (SAR) images while being more complicated compared to optical images, is especially important due to its numerous applications. The main aim of current research is to determine improvement of SAR image matching process by deploying texture analysis using gray level's co-occurrence matrix (GLCM). Three parts of the pair of TerraSAR-X images are used to implement the methodology. The results show that for some areas with low texture, the conventional image matching algorithm is not able to detect corresponding points, while using other textural features in image matching process leads to improvement in quantity of acceptable matched points.

In the application of remote sensing SAR process in Guizhou karst mountainous region, the fragmented terrain and the complicated crops lead to the unsatisfactory precision, which is caused by using the grey value of SAR image simply in the classification. The original image was processed by the FROST filtering window with 7 × 7. After computing the different window size and direction of GLCM, the different texture characteristic values of GLCM were analyzed. The results show that step size of 1 and moving window with 5 × 5 of GLCM gets the optimum characteristic value. In addition, based on the four GLCM characteristic values of each landuse types, the correlation characteristic is fluctuating and easy to distinguish. To improve the precision of SAR image target recognition ability, as the image texture feature, the correlation characteristic will take part in the classification of the whole study area. But the range of the improvement needs further research.

Image texture classification and segmentation is a main topic in the analysis of many types of images. People usually use the least squares estimation (LSE) for analyzing SAR textures. But we find that the LSE is unstable in practical computation. Therefore, in this paper we present regularization methods for image texture classification and segmentation. Regularization is such a technique which can successfully suppress the instability due to noise or truncation error when computing. Several regularization techniques, including standard regularization (SR), penalized regularization (PR) and total variation based regularization (TVR), are exhibited to reduce instability in texture extraction. Experiment results demonstrate that the regularization methods are superior to LSE and seem to be promising in practical applications.

This paper proposed a texture's feature of gray-level co-occurrence matrix (GLCM) as a calibrator of the sea wave height. Texture in Synthetic Aperture Radar (SAR) images represent the surface roughness of the sea or the wave height. As an archipelago country which has 70% sea of entire region, Indonesia needs and must have early warning system to provide transportation and fisherman safety circumvent the accident. Object of research are the P band full polarization Synthetic Aperture Radar (SAR) image acquired from coastal of East Kalimantan. These polarization are; horizontal-horizontal, horizontal-vertical, vertical-horizontal and vertical-vertical. Extraction of feature is done under observation angle 0° and adjacent pixels is 1. The results showed homogeneity is a consistent and toughest feature in horizontally-vertically polarization and vertical-vertical polarization and this feature can be used as a calibrator.

SAR image processing is an important way to detect marine oil spills. The detection efficiency and accuracy are the most important indicators. In order to detect oil spills more efficiently, a framework based on features and the extreme learning machine method is proposed in this paper. Texture feature is an effective approach for region of interest extraction in image processing, which is consistent with human visual sense. Both gray level co-occurrence matrix and Tamura features are selected to extract features from SAR images, which is different from general methods with denoising and image segmentation procedure. Then, a feature vector is constructed including all the features extracted, and the vector is taken as the input to train an extreme learning machine model. Through the training of many samples of SAR images, a final model for oil spill detection will be accomplished. A lot of detection experiments using this model show that the oil spill detection framework based on feature and extreme learning machine has higher accuracy and efficiency, and it can be carried out in practical application.

In this letter, a simple, yet very powerful local descriptor called local pattern descriptor (LPD) is proposed for synthetic aperture radar (SAR) images classification. The descriptor aims at exploiting the underlying properties of SAR image texture. Specifically, LPD consists of two parts: image quantization and statistical features extraction. The method of image quantization is based on recent local binary pattern. For an SAR image patch in a moving window, after quantization, different patterns can be obtained, which represent the local structures that exist in SAR image. Then, the statistical features extracted from the different patterns are concatenated to construct the LPD. Experiments on the TerraSAR-X image present that the proposed descriptor yields promising results for SAR image classification when compared to other widely used features.

Image splicing is performed for malicious purposes and it is trouble-free due to the availability of simple to use image editing applications. In this paper, we use standard deviation filter to highlight the image details and then apply rotation invariant co-occurrence among adjacent local binary pattern (RIC-LBP) operator to extract the internal statistical information. The standard deviation filter is used in Synthetic Aperture Radar images to detect an object and oil in sea. This filter highlights the abrupt changes in Synthetic Aperture Radar images. In image splicing when two or more images are used to create forged images the abrupt changes occur that are highlighted using standard deviation filter. The RIC-LBP operator provides internal statistics of the image due to its high descriptive ability. We use Columbia database and DSO-1 database for performance evaluation and use Spectral Regression Discriminant Analysis as a classifier. Our method outperforms some of the state of the art methods.

Extracting water body information with SAR images is distinctive and important. Especially for high-resolution SAR images, it's hard to extract water accurately because of Speckle Noises, so a Gabor filtering method related to the image texture is proposed. Firstly, frequency spectrum analysis method was used to identify the wedge and ring characters of image texture, and the period and directional parameters of the texture distribution were computed. Based on these characters, Gabor filter were constructed in order to filter the frequency spectrum of the multi-texture image, and best filter parameters were selected. After the filtering in frequency domain, components of corresponding texture were acquired by transforming the filtered frequency spectrum image to time domain image. The test results of terra SAR image indicate that the water body extraction method is good.

Synthetic aperture radar (SAR) image processing has many applications in the fields of target recognition, mineral detection, weather forecasting, agricultural, and etc. due to its high spatial resolution and imaging technology. However, the process of this type of images is difficult because of the existence of speckle noise. Nowadays, segmentation of textural regions based on designing a Kernel function with proper parameters is a real challenge. In this paper, a new parameter estimation algorithm has been proposed to design an efficient Kernel function for texture-based segmentation of SAR images. In this method, the Curvelet transform is applied to the SAR image only in one step and the inner layer coefficients as texture features are extracted. Then, a kernel function is formed based on the kurtosis value of the Curvelet coefficients energy (KCE). In the next step, the segmentation of different textures is applied by using the estimated KCE Kernel function. Experimental results on both simulated and real SAR images demonstrate that the proposed algorithm is effective for segmentation and description of different textures in SAR images, and it contains less misclassified pixels in comparison with other methods.

The geometric correction method based on SAR image simulation, with advantages of no requirement of ground control points (GCP) and all the processing can be done automatically, has attracted broad attention and application. However, the imprecise input parameters of SAR image simulation, especially the flight angle, will deform the texture of the simulated SAR image, and degrade the matching probability and matching precision of GCPs extracted from it and the real SAR image, and then lose the geometric accuracy. The influence of flight angle error on the image texture, matching probability and matching precision was analyzed in this paper, and a quantificational analysis using the real data was also completed. Results show that GCPs can get a high matching probability and precision when the flight angle error is less than 6°.

Speckle noise of Synthetic aperture radar (SAR) image and the azimuth sensitivity for imaging obstruct the interpretation and applications for SAR images. But it is very much remote sensing source because it can acquire data under all-day and all weather. Texture features reflect the spatial property of ground objects. If texture features change, in general the ground object does so. Therefore, this paper deep analyzes SAR image texture feature and proposes a new multi-temporal SAR image change detection algorithm, which is called by texture features fusion voting (TFFV) algorithm. Finally, the airborne SAR image data tests the method and experimental results indicate that the presented methods are feasible.

In this paper, we propose a SAR image texture segmentation method based on Contourlet transform and Support Vector Machine (SVM). The Contourlet transform provides a multi-resolution and multidirectional representation of images, which can capture the directional texture information into subbands of different scales and directions. Thus, the energy, standard deviation and information entropy defined on Contourlet subbands are used to construct the texture feature vector. The SVM is performed as a nonlinear classifier, and the decision tree strategy is applied to extend the primitive binary SVM to a multiclass SVM. Numerical results show that all three interested textures are successfully segmented.

The Synthetic Aperture Radar (SAR) image quality assessment (IQA) can provide a measurement for SAR jamming effect, which is helpful to improve the jamming pattern. A texture-based SSIM (TSSIM) algorithm is proposed, because of the fact that SAR images have much texture and SSIM algorithm has good performance in optical image quality assessment. TSSIM combines the image gray intensity with textural features to measure the image structural information. The Gray Level Co-occurrence Matrix (GLCM) is adopted to extract the textural features of SAR image. The angle second moment feature is proved the best compared with the other textural features. The results of simulations have shown that TSSIM is effective and can assess the jamming effect more accurately in different regions such as complex texture and simple texture area compared with SSIM and PSNR.

Image texture interpretation is an important aspect of the computer-assisted discrimination of SAR sea ice imagery. Co-occurrence probabilities are the most common approach to solve this problem. However, other texture feature extraction methods exist that have not been fully studied for their ability to interpret SAR sea ice imagery. Gabor filters and Markov random fields (MRF) are two such methods considered. Classification and significance level testing shows that co-occurrence probabilities classify the data with the highest classification rate, with Gabor filters a close second. MRF results significantly lag Gabor and co-occurrence results. However, the MRF features are uncorrelated with respect to co-occurrence and Gabor features. The fused co-occurrence/MRF feature set achieves higher performance.

This paper explores different types of gray level co-occurrence matrix (GLCM) [2] texture features for automated detection of landslides on levees using remotely sensed Synthetic Aperture Radar (SAR). Two approaches of texture analysis are investigated: one based on a rubber band straightening transform (RBST) which has been used extensively in the past in the medical imaging community, and one based on the authors' developed approach of spiral straightening transform (SST). The transforms are used to project a circular region in the image to a rectangular representation where texture feature extraction can be applied. Straightforward linear discriminant analysis, for feature reduction and optimization, and maximum likelihood methods, for classification, are also utilized. The proposed system was tested on L-band SAR data with HH, HV, and VV polarizations collected from NASA's UAVSAR of the Mississippi River levee system between Vicksburg, MS and Clarksdale, MS, USA. The proposed approach is shown to detect all known levee landslides in the test area with a low number of false positives.

This paper proposes the classification framework based on the Bayesian theory with the single polarization multi-temporal synthetic aperture radar (SAR) and an optical data, and incorporates the proposed training sample selection (SS) methods. Within this framework, the combination with gray level co-occurrence matrix (GLCM)-based mean textural measure is investigated. The two procedures of the classification and proposed SS are united, where SS generates accurate and dispersed training samples. Extracted features from multi-temporal SAR data—namely, the average backscattering coefficient, backscatter temporal variability, and long-term coherence and reflectance values from optical data, are integrated with GLCM mean textural data in the framework. Classification results were generated by taking Osaka City, Japan, as the study area. The selected major classes were built-up areas, fields, woodlands, and water bodies. The most suitable data used for classification is the multi-temporal SAR and an optical data fusion with the mean textural, because of the supplement of different data used and the smoothing effect on the images of the texture. Moreover, the higher quality training samples obtained by using the combined SVM and NN-based SS method for training the Bayesian classifier generated the highest classification accuracies in all of tested cases.

In this paper we address the feature selection problem for X-SAR images and further the segmentation of specific chosen classes. After defining a suitable feature space for X-SAR images we select the most significant ones via a supervised machine learning approach: the 1-norm SVM. The selected features will be used for segmentation purposes, in order to segment water areas from the background. We shall see that the most relevant features are based on texture elements. So the segmentation is texture based and achieved with variational calculus and level set methods. The work is mainly focused on urban park X-SAR SpotLight images, where lakes and rivers are often present. The images are collected with the COSMO-SkyMed satellites constellation, equipped with a SAR sensor.

This paper introduces the oil spill monitoring in SAR image by texture analysis and spectral information. In texture analysis, it discusses the parameters of texture extraction, and makes experiment. Then it determines the 17*17 as the windows size, 90 as the angle and 5 as distance. Through neural network classification, these parameters are suitable for oil spill monitoring. It can distinguish with oil and oil-like well. Its accuracy is satisfactory. These parameters are not only used for oil spill monitoring, but also provide a foundation for deeper SAR image classification.

In order to extract available information from SAR image effectively, this paper proposes a method of texture feature extraction based on co-occurrence matrix. Based on the analysis of common attribute values of co-occurrence matrix, the method analyzes the influence of window size and displacement vector on texture feature. By comparing both distribution and mean value of different objects' texture feature, the research gets the best window size and displace vector to calculate co-occurrence matrix; determines practicable texture feature combination for classification, and advances the steps toward selecting and extracting the text feature in detail. The experiment results illustrate that the features extracted by the method are effective in describing objects.

Due to the urgent requirement on SAR image interpretation, a method to extract building footprints from PolSAR image based on polarimetric scattering features and texture features was proposed. Firstly, polarimetric scattering features are extracted by Freeman decomposition and H/A/α decomposition method. Texture features are extracted based on gray level co-occurrence matrix (GLCM). Then, support vector machine (SVM) with Gaussian kernel function is trained to detect building regions by different features sets selected. At last, building polygons extraction is implemented: Line segments are extracted from obtained building regions by radon transform; cycles corresponding to the profiles of buildings in the graph are searched after determining line segment endpoints graph; building footprints are created by eliminating false cycles and connecting component cycles. The approach was demonstrated using airborne PolSAR image. Different features sets selected from experiment data were used to extract building footprints. Comparing the extraction results, the optimum features sets fitting for building footprints were extracted. Results show that the method provides a reliable way to extract building footprints comprehensively using multi-features and edge information. complex scattering characteristics of buildings, accuracy of SAR building extraction is low. In order to monitor the city buildings in all day, all weather and improve the accuracy of building extraction with SAR image, a new building extraction method was put forward in the paper, based on the polarimetric scattering features, texture features and edge information of the airborne PolSAR image. With comprehensive use of polarimetric scattering features, texture features and edge information, a good building extraction effect can be got. A case was studied in the paper to test the method and perfect results were achieved. The research provides a new method to extract the city buildings automatically with PolSAR image. II. ANALYSIS OF BUILDING FEATURES OF POLSAR IMAGE

This paper presents a methodology for automatic co-registration of polarimetric Synthetic Aperture Radar (SAR) images obtained from similar satellites, but with differing view angles. The Scale Invariant Feature Transform (SIFT) algorithm is applied to extract candidate Ground Control Points (GCPs), where two-way keypoint matching eliminates improbable correspondence keypoints. Minimizing the Root Mean Squared Error (RMSE) also removes matching points with large RMSE and refines the location of points in a warp image through a pseudo-affine transformation. In addition, information entropy and spatial dispersion quality constraints allow quantification of the spatial distribution of the GCPs. In terms of the polarimetric SAR registration, co-registration using total power (TP) images can be used to minimize the total registration error for all of the polarization, as the TP combination generates the smallest total RMSE. In addition, geometric calibration can be used to improve the accuracy of co-registration. Moreover, the influence of differences in the texture patterns of polarimetric views on the co-registration performance is investigated.

The aim of this paper is to analyze a procedure aimed at human settlement extraction from SAR images based on Local Indicators of Spatial Association (LISA) and textural features, followed by a refined classification scheme using the same features to extract more detailed land use classes according to the CORINE nomenclature. A segmentation technique is exploited in order to identify statistically homogeneous regions and eventually a supervised classification of the selected features allows assigning each region to a class. Tests on the areas of Shanghai show potentials for the use of these techniques in urban area monitoring using moderate resolution SAR.

As an important type of wetlands, fish-pond provides a number of important ecosystem services. It was easily affected by human activities such as the change of cultivation structure. Tracking the fast change of fish-pond can obtain the information that the environment is impacted by human activities to what extent. In this study, we proposed a change detection procedure in order to delineate fast change fish-pond using high resolution, short term time series of SAR imagery. The procedure includes (1) SAR images preprocessing, (2) object extraction, (3) multi-temporal change characteristics analysis of fish-pond, (4) change objects classification, and (5) a final change results analysis. In this study, object-oriented method was adopted to extract and analyze the texture characteristics of fish-pond in the high spatial resolution of SAR data. Objects were generated with suitable scale parameters by means of multi-temporal segmentation approach according to the image features (including backscatter, difference and ratio of three temporal RADARSAT images, shape characters), and the changed objects were extracted through analyzing the shape index and contextual information generated by Definiens software. Six change types of fish-pond were achieved. The results indicate that high resolution SAR data can provide better information for fish-pond change detection in the study area, and the procedure proposed in this study can efficiently obtain the fast change objects with high accuracy between three short term time series SAR images, which are useful for the further classification.

For the speckle consisted in Synthetic Aperture Radar (SAR) image，good result of SAR image segmentation can not be gotten with traditional methods. In this paper，a new method of SAR image segmentation is proposed with the combination of fuzzy neural network and the statistical feature of SAR image. Firstly, texture features of SAR image are extracted with gray level co-occurrence matrix，then SAR image is filtered according to its distribution characteristics in the wavelet domain. Finally, SAR image is segmented according to the vector composed of texture features and gray level of filtered SAR image with fuzzy neural network. The experiment results show that the new method is efficient in SAR image segmentation.

Combination of gray water and land SAR image and wavelet texture information, present a new segmentation method of SAR image surface. Firstly, extracting gray level co-occurrence matrix of the sub-blocks SAR image, then using wavelet transform to extract the norm and the average deviation as the wavelet texture feature information of sub-blocks of sub-image; Accordingly, two types of texture establish a suitable combination of image separation measure multi-dimensional feature space; Finally, using K-means clustering algorithm to segment the SAR water image. The experimental results show that the effect is better than the common segmentation method.

Presents a new method for segmentation of synthetic aperture radar (SAR) images. A Gaussian autoregressive (GAR) model under a multiresolution pairwise Markov framework can be proposed based on texture feature fusion images from in part gray level co-occurrence probability statistics, we examine the texture segmentation of SAR image suing the multi-resolution maximization of the posterior marginal (MPM) estimate with corresponding unsupervised segmentation algorithm on those texture feature fusion images. This method not only use of pixel gray level information, but also the use of pixel space location information, reducing the speckle noise effect for the segmentation. For some SAR images, compared with multiresolution pariwise Markov-GAR model texture segmentation based on gray level images, the results of experimentation show that the segmentation precision can be improved by the method in this paper.

This paper proposes an evolutionary RBF network classifier for polar metric synthetic aperture radar ( SAR) images. The proposed feature extraction process utilizes the full covariance matrix, the gray level co-occurrence matrix (GLCM) based texture features, and the backscattering power (Span) combined with the H/a/A decomposition, which are projected onto a lower dimensional feature space using principal component analysis. An experimental study is performed using the fully polar metric San Francisco Bay data set acquired by the NASA/Jet Propulsion Laboratory Airborne SAR (AIRSAR) at L-band to evaluate the performance of the proposed classifier. Classification results (in terms of confusion matrix, overall accuracy and classification map) compared to the Wish art and a recent NN-based classifiers demonstrate the effectiveness of the proposed algorithm.

This paper discusses methods and parameter settings that help to estimate texture in SAR images. In general, this is a difficult task for SAR images that are characterized by speckle noise and which span a wide range of pixel magnitudes. We applied Gauss Markov Random Field (GMRF) models and Enhanced Model Based Despeckling (EMBD) to 1 meter resolution amplitude images of the German TerraSAR-X mission. The results demonstrate that one can find appropriate parameter combinations that allow robust texture estimation even for different types of target areas.

In this paper, a new method for supervised classification of terrain types in polarimetric Synthetic Aperture Radar (Pol-SAR) images is proposed. This technique is a combination of the texture classification and the maximum likelihood classification based on the complex Wishart distribution for the polarimetric covariance matrix. The texture features are first extracted from the span image based on co-occurrence matrices; and then the classifier combines the texture features with the distance measure based on polarimetric information to obtain the results. Using a NASA/JPL AIRSAR image, the effectiveness of the proposed method is demonstrated.

SAR images are affected by speckle which is a coherent process modelled as a multiplicative noise. It makes the automatic image classification difficult. Thus, many methods have been developed to remove speckle from SAR images while preserving the useful information of the image such as texture. This paper presents an evaluation of texture extraction parameter estimation methods using Cramer-Rao lower bound (CRLB). The first evaluated method is model-based despeckling (MBD) algorithm, which uses Gauss-Markov random fields as prior. The second one is the maximum a posteriori auto-binomial method (MAP-ABM), which rather uses auto-binomial model as prior. The evaluation has been carried out using simulated SAR data. In here, data with increasing number of looks have been used in order to study 1) how the estimated parameters approach the real one, and 2) how their variances get closer to CRLB. The experimental results show the superiority of MBD parameter estimation. Both MBD and MAP-ABM provide the most robust texture parameters when the number of look is between 3 and 4.

This paper focuses on the topic of extracting residential properties information based on texture features in low or moderate resolution SAR image, because the residential area has characters of high bright and regular texture in that scale image. Firstly, filtering the noises of the SAR image by using the FROST algorithm; secondly, obtaining a binary image which contains the residential information by threshold processing; thirdly extracting texture features based on the gray-level co-occurrence matrix and the texture features such as entropy, variance and correlation are selected according to actual situation, then the selected images are combined into a multiband image; Finally, the multi-band image multiplies the binary image and gets a new image, then classifying it by unsupervised classification, connecting the break points by using the morphology algorithm, a boundary of residential areas is obtained. As results, both the theoretical analyses and the experimental results indicate that this method is very efficient in extracting the residential information.

The aim of this paper is to estimate freeze-thaw dates of a small lake using Synthetic Aperture Radar (SAR) images. Here, the estimation of freeze-thaw dates is restated as a classification problem wherein each SAR image in the time series is classified as an ‘ice’ or ‘no ice’ image based on the presence or absence of ice on the lake. We implement and compare different methods consisting of various feature extraction algorithms and nonparametric classification schemes. We obtain the best classification accuracy using histogram information from SAR images applied to a Random Forest classifier. Correct classification rate in the range of 90% is seen in this case and a standard median filter is used to estimate the freeze-thaw dates from the classifier output. We also extend our analysis to the use of texture features and local statistics in solving this problem.

For single-band and single-polarized SAR, its capability to monitor oil spill is limited based on image intensity. Texture features are suggested to improve accuracy of oil spill surveillance. Texture measures, extracted from GLCM (gray-level co-occurrence matrices), are analyzed, which indicates that mean, contrast, variance, entropy, and dissimilarity are effective for oil identification. SAR image is characterized by high resolution and speckle noise, which limits pixel-based approaches. In this paper, object oriented image analysis is used to extract oil slick. This algorithm typically incorporates both spectral and spatial information in the image segmentation phase. Results indicate that texture features extend features of interested objects, and help to improve oil spill surveillance.

A model based on texture feature and fuzzy logic algorithm was constructed to discriminate oil spills from look-alike phenomena in the SAR images. Statistics texture feature of SAR images were extracted and used as the input parameters in the fuzzy logic system. The texture features consisted of entropy second order, angular second moment, contrast and inverse difference moment of dark objects. The system analyzed 38 SAR images with 77 oil spills and 52 look-alikes, and provided the probability of a dark object to be an oil spill. The remaining 26 processed SAR images, which were not included in the training, were used to test the system. The result showed that 80.5% of the oil spills were correctly classified. It seemed that the texture features and fuzzy logic system were effective in identifying oil spills on marine SAR images.

Oil spill pollution is a major environmental threat for many countries in the world, which can cause serious damage to marine environment. Synthetic aperture radar (SAR) has become a valuable tool for marine oil spill monitoring, because of its all-weather and all-day capabilities. However, interpretation of marine SAR imagery is often ambiguous, and some other look-alike features often pose a fundamental challenge to the identification of oil spills and make the discrimination between oil spills and the look-alikes become a necessary procedure. In this paper, co-occurrence matrix method is employed to extract textural features of marine SAR image first, then these features are analyzed and optimized, and then support machine vector (SVM) method is used to identify oil spills in SAR images. Experiments on several SAR images show that method proposed in this paper can improve the detection and identification of oil spill in SAR images.

In this paper a novel method, based on autocorrelation indexes and gray-level co-occurence matrix, for the extraction of urban areas in high resolution SAR images, is presented. It strongly reduces human interpreters' intervention thanks to a high degree of automation within the processing chain and allows a fast and accurate generation of built-up area maps, which can be employed for land mapping and support in relief operations.

In this paper single-band and single-polarization Radarsat-1 SAR image is used to evaluate image classification with textural analysis. Firstly, the statistic information of sample were analyzed using semivariogram to determine the optimum parameters for textural extraction; Then four textures such as Homogeneity, Mean, Angle Second Moment and Entropy had been calculated based on GLCM, and the image data were processed using Support Vector Machine classification. The results show that the water and settlement areas are extracted accurately with accuracy 99.34% and 82.54%, and the SVM method has better extension ability for SAR image classification; Assisting with textural information, the image classification based on SVM has a obvious enhancement to original SAR, especially for some complex objects such as settlement areas(about increasing accuracy 18%).

Contourlet has shown good performance in different aspects of image processing. Because of its charming multi-resolution and multi-direction characteristics, Contourlet is especially efficient in the processing of images with abundant texture. In this paper, a texture classification method is presented for the SAR images based on the Contourlet transform and a minimum distance classifier. The interested image is firstly decomposed with Contourlet transform. Then, the texture feature vector is constructed with only the significant first order statistical features of Contourlet coefficients for the concern of computational complexity. In the classification procedure, the Euclidean distance is used as the measurement of the texture similarity between two feature vectors. Finally, the minimum distance classifier is applied for the classification of the target images. The proposed texture classification method is implemented with the SAR image of the Cairo city. Numerical results show that the proposed method has good performance for the classification of areas with distinctive texture features.

This paper proposes an improved GVF Snake-based method which is applied to Envisat ASAR images to determine the interested object's boundaries. ASAR images have a lot of noise and speckle. But different from other low SNR images, they are also full of texture. While using GVF snake model, the result of the gradient vector field in ASAR images becomes even worse. How to reduce the influence of the unimportant texture and the noise level while to preserve the image features is the key problem. To solve this, log Gabor wavelets that have the ability to preserve the phase data is introduced into the new scheme to correct the gradient vector flow. Results show that the new method can detect the contour of the interested object more accurately and false edges are removed.

Image understanding and retrieval for metric SAR data needs to be tackled by specific algorithms in order to exploit the specific metric scale scene characteristics and the typical sensor phenomenology. A composite approach is developed to this end. Extraction, characterization and simplification/grouping algorithms are employed together with super-resolution for isolated bright scatterers, scatterer alignments and uniform backscatter and texture areas to populate a geospatial database. The generated descriptors can be navigated via different data viewers for scene understanding and data retrieval applications. Examples on real metric SAR data are given.

In this work we address the problem of land cover classification in advanced synthetic aperture radar (ASAR) images. The derivation and assessment of texture features for ASAR image segmentation is investigated using full multidimensional co-occurrence matrices as features. Expansion of local patches in terms of Walsh functions helps identify the optimal distance for the calculation of the co-occurrence matrices. The defined distance agrees with the one chosen by performing exhaustive tests where many distances were tried and the best was chosen from the training data. The well known chi-square test of statistical significance has been used for classification.

This paper is to investigate a new unsupervised approach for the extracted objects based on synthetic aperture radar (SAR) image using improving fuzzy clustering method. The traditional fuzzy c-means clustering (FCM) is very sensitive to the initial value and the number of clusters. The accurate initial value and number of clusters are important parameters to get the accurate result in FCM. SAR image has extensive application in national economy and military field. And a typical characteristic of SAR image is that it is influenced by speckle noise. So the traditional algorithm (Melgani et al., 2000) of FCM applies directly SAR image to get the ideal result difficultly. This paper employs the textural feature in SAR image to extract the transition and propose a new fuzzy unsupervised classification method for SAR images using the transition region to define the initial value and the number of cluster adaptively. The experimental results prove the efficiency and accuracy of this unsupervised method for SAR images

Since the past five years Lockheed Martin Canada is developing an advanced image interpretation system integrating classification tools and target detection operators for multispectral, hyperspectral and polarimetric SAR imagery. Improved classification maps and superior object detection and identification performances have been obtained with a pixel-driven evidential fusion of textural measurements and end-member respectively extracted from polSAR and HIS imagery. As in many applications objects to be extracted occupy several pixels in the image, the objective of current developments is to supplement the pixel-based fusion already performed with radiometric, textural, spectral features by adding specific object topological attributes (shape, size) and contextual description (spatial relationships). This paper presents a description of the new primitives and the results obtained on Ikonos imagery.

As synthesis aperture radar (SAR) represents a powerful Earth observation tool for monitoring geophysical resource globally, SAR images could be used for land description and scene analysis. Actually, various applications specifically require the detection and the analysis of urban areas from original SAR images, which is a difficult task due to the speckle signal of the images and the complexity of these scenes. In this paper, an unsupervised approach to extract urban areas from SAR images has been suggested based on local statistical characteristics and texture information described by Gaussian Markov Random field (MRF) model. First, a probability map of the urban areas is computed based on local statistical characteristics, using an ffmax operator proposed by C. Gouinaud. Then the Gaussian MRF model is adopted to describe the texture of urban zones and the parameters of the model are estimated from the original image together with the probability map. Finally, the urban areas are extracted under Bayesian framework by maximum a posterior (MAP) criterion, with modeling the urban label field by Potts model. The performance of the proposed method is evaluated by experimental results on real SAR images

In this paper a new texture-based change detection approach is proposed to identify the flooded regions in SAR images. The main novelty of our approach is that the most distinctive texture information is automatically learned from the training set. Forty texture features, which are extracted from a pair of bi-temporal SAR images, are used to construct the weak classifier pool. After AdaBoost training, a strong classifier is optimally combined by a small subset of the candidate weak classifiers. The experimental results demonstrate the effectiveness of the proposed approach.

Texture features based on Haralick's co-occurrence matrix were compared for land cover and forestry classification purposes. According to these results, the best texture features were Angular Second Moment, Mean and Entropy, the worst Correlation and Standard Deviation. The best SAR-images were taken during wet snow or ground. Usually, the larger the window used to construct the co-occurrence matrix, the better the results. The suitable length of the spatial step depended on texture feature and classes. The directionless texture features performed usually well. The results of the performed classification experiment were disappointing.

The Canadian Ice Service (CIS) is a government agency responsible for monitoring ice-infested regions in Canada's jurisdiction. Synthetic aperture radar (SAR) is the primary tool used for monitoring such vast, inaccessible regions. Ice maps of different regions are generated each day in support of navigation operations and environmental assessments. Currently, operators digitally segment the SAR data manually using primarily tone and texture visual characteristics. Regions containing multiple ice types are identified, however, it is not feasible to produce a pixel-based segmentation due to time constraints. In this research, advanced methods for performing texture feature extraction, incorporating tonal features, and performing the segmentation are presented. Examples of the segmentation of a SAR image that is difficult to segment manually and that requires the inclusion of both tone and texture features are presented.

A multiscale approach is introduced to classify the JERS-1 mosaic image over the Central African rainforest. A series of texture maps are generated from the 100 m mosaic image at various scales. Using a quadtree model and relating classes at each scale by a Markovian relationship, the multiscale images are classified from course to finer scale. The results are verified at various scales and the evolution of classification is monitored by calculating the error at each stage.

Three techniques for generating texture statistics are examined: the gray-level co-occurrence matrix (GLCM), the gray-level difference vector (GLDV) and the neighboring gray-level dependence matrix (NGLDM). The objective of these statistical approaches is to translate visual texture properties into quantitative descriptors in a manner that they can be used to discriminate relevant land features using additional image processing techniques. These second-order statistical methods are used to generate texture features from C-HH and C-HV airborne synthetic aperture radar (SAR) data collected on July 10, 1990 over an agricultural area in southern Ontario Canada. Texture features generated from the GLCM, GLDV and NGLDM are classified individually using a k-nearest neighbor (k-NN) supervised classifier. The greatest classification improvement (/spl ap/20%) was observed with the mean and correlation texture features derived from the GLCM. However, the selection of a specific second-order statistical technique may not be critical, since similar classification improvements were observed for the GLCM, GLDV and NGLDM statistical techniques. The results reported here highlight the importance of texture processing to methods of classifying agricultural crops using SAR data.

Statistical, textural and Gabor features were extracted from integrated multitemporal multispectral TM data and ERS-1 SAR imagery for urban land use mapping. The computed features are first normalised using the SOM algorithm and then a decision tree algorithm is applied for feature selection. The classification procedure was carried out with a multilayer perceptron, trained with the resilient backpropagation algorithm. The authors' results demonstrate the potential of the proposed methodology.