In our study, sea surface characteristics imaged by multi-polarization space-borne synthetic aperture radar (SAR) have been investigated. For the first time, a decomposition of different scattering mechanisms have been performed for ocean satellite SAR imagery to better understand the non-Bragg (Scalar) contribution to the total radar cross section (RCS) and Doppler measurements. Characteristics retrieval and target classification have been established, using polarimetry and Bayesian detection theories. There are generally three types of surface scattering mechanisms occurring when the sea surface is detected by microwave radar, i.e., Bragg, specular, and Rayleigh. Depolarized Bragg contribution corresponds to sea surface capillary wave, while the other two Scalar contributions correspond respectively to the crest of the longer wave before it breaks and foams formed by wave breaking. Different scattering mechanisms induce different polarimetric scattering coefficients and Doppler spectrum. It had been impossible to separate those scattering mechanisms with single polarization radar imageries. On pixel scale, we decomposed radar scattering matrices physically into Bragg and Scalar contributions. The decomposition is an iteration initiated with the radar incidence angle, and controlled by a local incidence angle which is function of co-polarization and cross-polarization. Based on these developments and testing, a strategy has been refined to analyze the signature of different features, to retrieve wind seas and sea swell parameters, as well as slick areas, ships, oil rigs, such polarized targets that may be buried in the Scalar contribution. With polarimetric scattering matrices estimated both for Bragg and Scalar contributions, a sea clutter model describing almost the real sea surface has been improved statistically. From this point, this improved model could be combined with Bayesian detectors to classify man-made metallic targets, such as ships, oil rigs, etc.