The goal of the present paper is on retrieving the electrical and geometrical parameters of a stratified medium with two rough interfaces. The inversion problem is formulated as a cost function optimization problem, and it is solved using the simulated annealing algorithm. The cost function consists in the integrated squared deviation between the co-polarized incoherent intensities obtained from the Small Slope Approximation and those obtained from the Small Perturbation Method. The inversion scheme is applied to the electrical and geometrical parameters involved into the analytical expressions of the incoherent intensities given by the SPM. We study the influence of the shape of the autocorrelation function and the isotropy factor upon the estimation of parameters. We test the sensitivity of the inversion scheme to noisy synthetic data. The study is applied to snow-covered soils in L-band. For the configurations under study, we show that the inverse method is efficient for eight- parameter or ten-parameter predicting problems.