María Magdalena Lucini

Performance of Robust RA Estimator for Bidimensional Autoregressive Models

The additive AR-2D model has been successfully related to the modeling of satelital images both optic and of radar of synthetic opening. Having in mind the errors that are produced in the process of captation and quantification of the image, an interesting subject, is the robust estimation of the parameters in this model. Besides the robust methods in image models are also applied in some important image processing situations such as segmentation by texture and image restoration in the presence of outliers. This paper is concerned with the development and performance of the robust RA estimator proposed by Ojeda (1998) for the estimation of parameters in contaminated AR-2D models. Here, we implement this estimator and we show by simulation study that it has a better performance than the classic least square estimator and the robust M and GM estimators in an additive outlier contaminated image model.

Parameter Estimation in SAR Imagery Using Stochastic Distances and Asymmetric Kernels

In this paper, we analyze several strategies for the estimation of the roughness parameter of the GOI distribution. It has been shown that this distribution is able to characterize a large number of targets in monopolarized synthetic aperture radar (SAR) imagery, deserving the denomination of “Universal Model.” It is indexed by three parameters: 1) the number of looks (which can be estimated in the whole image); 2) a scale parameter; and 3) the roughness or texture parameter. The latter is closely related to the number of elementary backscatters in each pixel, one of the reasons for receiving attention in the literature. Although there are efforts in providing improved and robust estimates for such quantity, its dependable estimation still poses numerical problems in practice. We discuss estimators based on the minimization of stochastic distances between empirical and theoretical densities and argue in favor of using an estimator based on the triangular distance and asymmetric kernels built with inverse Gaussian densities. We also provide new results regarding the heavy-tailedness of this distribution.

Robust Principal Components for Hyperspectral Data Analysis

Remote sensing data present peculiar features and characteristics that may make their statistical processing and analysis a difficult task. Among them, it can be mentioned the volume of data involved, the redundancy, the presence of unexpected values that arise mainly due to noisy pixels and background objects whose responses to the sensor are very different from those of their neighbours. Sometimes, the volume of data and number of variables involved is so large that any statistical analysis becomes unmanageable if data are not condensed in some way. A commonly used method to deal with this situation is Principal Component Analysis (PCA) based on classical statistics: sample mean and covariance matrices. The drawback in using sample covariance or correlation matrices as measures of variability is their high sensitivity to spurious values. In this work we analyse and evaluate the use of some Robust Principal Component techniques and make a comparison of Robust and Classical PCs performances when applied to satellite data provided by the hyperspectral sensor AVIRIS (Airborne Visible/Infrared Imaging Spectrometer). We conclude that some robust approaches are the most reliable and precise when applied as a data reduction technique before performing supervised image classification.

Robust classification of SAR imagery

In this work the G/sub A//sup 0/ distribution is assumed as the universal model for amplitude synthetic aperture radar (SAR) imagery data under the multiplicative model. The observed data, therefore, is assumed to obey a G/sub A//sup 0/ (/spl alpha/, /spl gamma/, n) law, where the parameter n is related to the speckle noise, and (/spl alpha/, /spl gamma/) are related to the ground truth, giving information about the background. Therefore, maps generated by the estimation of (/spl alpha/, /spl gamma/) in each coordinate can be used as the input for classification methods. Maximum likelihood estimators are derived and used to form estimated parameter maps. This estimation can be hampered by the presence of corner reflectors, man-made objects used to calibrate SAR images that produce large return values. In order to alleviate this contamination, robust (M) estimators are also derived for the universal model. Gaussian maximum likelihood classification is used to obtain maps using hard-to-deal-with simulated data, and the superiority of robust estimation is quantitatively assessed.

Simulation of Spatially Correlated Clutter Fields

Correlated 𝒢 distributions can be used to describe the clutter seen in images obtained with coherent illumination, as is the case of B-scan ultrasound, laser, sonar, and synthetic aperture radar (SAR) imagery. These distributions are derived using the square root of the generalized inverse Gaussian distribution for the amplitude backscatter within the multiplicative model. A two-parameter particular case of the amplitude 𝒢 distribution, called , constitutes a modeling improvement with respect to the widespread 𝒦 A distribution when fitting urban, forested, and deforested areas in remote sensing data. This article deals with the modeling and the simulation of correlated -distributed random fields. It is accomplished by means of the Inverse Transform method, applied to Gaussian random fields with spatial correlation. The main feature of this approach is its generality, since it allows the introduction of negative correlation values in the resulting process, necessary for the proper explanation of the shadowing effect in many SAR images.