Coastline detection with polynomial transforms and Markovian segmentations

In this paper, the authors present an original method for detecting coastlines on synthetic aperture radar (SAR) images. With this method, the processing of image data is performed in three steps: restoration, segmentation and coastline segment extraction. The problem of image restoration is solved via the polynomial transform. Based on the Markov random field theory, a model for image segmentation is applied. Optimization is achieved by a classical stochastic relaxation technique. The polynomial transform is an image description model which incorporates important properties of visual perception, such as the Gaussian-derivative model of early vision. Based on this, the authors present a technique for directional-sensitive image restoration. The restored image is obtained by means of an inverse polynomial transform which consists of interpolating the transformed coefficients with pattern functions that are products of a polynomial and a window function. They show in this paper how the noise reduction task can be improved by detecting the position and orientation of relevant contours in images degraded by speckle. This method is applied in a coarse-to-fine resolution approach, in which, contour location is not degraded even at the stage of high resolution processing. The presented method performs in a semiautomatic fashion to effectively detect coastlines