Statistical analysis of ocean surface texture in SAR images

The aim of this paper is to characterize the sea state as it is seen on SAR images. To do so one could perform a spectral analysis in order to extract parameters such as the direction of propagation and the wavelength of the swell. Although relevant, these parameters don´t give information about the surface roughness of the sea. In order to characterize the geometric variation of the sea, a ponctual statistical analysis can be performed. The parameters describing the texture are obtained by considering the shape of the distribution. Using the multiplicative noise model of the radar data, the grey level in the image could be described by the product of a variable which characterizes the organised component in the image with a noise responsible of the speckled texture. The authors propose to model the grey level using the Pearson system of distributions. This system has the interesting property of containing distributions with different shapes. From this model, they obtain two shape parameters describing the texture of the radar images. They focus on the statistical model of the backscattering coefficient. By using a system of distributions for it, they obtain new distributions for the grey level which offers good fits to the histograms of ocean radar images. The distributions in the backscattering coefficient system are also characterized by shape parameters which yield an analysis of the organized component in the images