Region-Based Classification of Multisensor Optical-SAR Images

Abstract

Multispectral and synthetic aperture radar (SAR) images are known to exhibit complementary properties: unlike optical sensors, SAR provides information about the soil roughness and moisture, and acquires useful data despite clouds and Sun-illumination conditions. However, the analysis of the resulting images turns out to be more dif cult, as compared to the use of optical imagery, due to the noise-like speckle phenomenon. In order to exploit this complementarity for classication purposes, a criticality relies in the de nition of accurate joint optical-SAR statistical models, due to the different physical natures of these two data typologies and to the corresponding differences in the related parametric models.In this paper, a region-based semiparametric classi cation technique is proposed for multisensor optical-SAR images. The method combines the tree-structured Markov random eld approach to segmentation with the dependence tree approach to probability density estimation and with casespeci c bivariate models for the distributions of optical and SAR data. A Bayesian decision rule is formulated at the segment level in order to incorporate spatial-contextual information and to gain robustness against noise.

Marker-Controlled Watershed-Based Segmentation of Multiresolution Remote Sensing Images

“…Traditionally, it is carried out independently on each pixel, but one can instead take advantage of the available object layer assigning a unique class to each object. This approach has a number of advantages: By working on extended objects, one reduces the impact of noise on accuracy, particularly relevant for some imaging modalities, such as SAR [46]; geometrical features can be taken into account to improve classification; interactions among objects can be studied to characterize complex scenes, as done in [2]; last but not least, working on objects rather than pixels reduces the number of processing atoms, thus limiting computational complexity.…”

Sar and Oblique Aerial Optical Image Fusion for Urban Area Image Segmentation

“…A closely related area, classification, has had some success, although the literature mostly concentrates on nadir-view RGB images, e.g. (Voisin et al, 2014) or (Gaetano et al, 2008).…”

A minimum spanning forest based classification method for dedicated breast CT images

“…26 SVMs in conjunction with other methods have been successfully used for remote sensing data classification. [27][28][29][30][31] Minimum spanning forest based methods have been used for segmentation of other images such as hyperspectral images and have been combined with support vector machines to classify various regions. 32 Other works have incorporated the use of a probabilistic support vector machine to determine highly probable markers for minimum spanning forest based classification.…”