Original Freeman three-component decomposition was found useful in information extraction from a mix of area: city blocks, forest, ocean and land surfaces, etc.. However, it al- ways suffers from some inconsistencies with real situation such as negative power and scattering mechanism ambiguity. It is probably because of the overestimate of volume scattering power. But intrinsically it is the inconsistency between the assumed models and the PolSAR. The de- orientation method was initially found useful to alleviate this problem to some degree. However, because of the boundedness of deorientation, it is sometimes noneffective. Nonnegative eigen- value decomposition (NNED) can absolutely solve the negative power problem. But the residual matrix except the volume scattering model is assumed equal to two. So in some sense it is under the volume scattering dominance assumption. Therefore, the decomposition should be separated into two individual parts. In this paper, for the pixels that volume scattering dominates, the NNED will be used. For the pixels that volume scattering doesn’t dominate, we will start with the assumed models, and focus on developing a generalized model-based decomposition. Be- cause for model-based scheme, we insist that one should radically start from the models and then a generalized three-component model-based decomposition is proposed, which consist of the surface, double bounce and volume scattering. Among it the generalized scattering mechanism proposed by Cloude is adopted, which provides a more accurate model for surface and dihedral scattering, and the dominant scattering component will be figured out according to the alpha angle parameter to be solved in the generalized model. Through this solution, the dominant scattering mechanism in one pixel will be preserved better. The performance of this approach is demonstrated and evaluated using the airborne AIRSAR and E-SAR data sets. The results show the advantages and improvements especially for alleviating the scattering mechanism ambiguity of oriented buildings from vegetation.