Forest above ground biomass (AGB) is an important variable for evaluating ecosystem function and structure across landscape, which is necessary for studying forest productivity, carbon balance and nutrient allocation in forest ecosystem. In this study, a forest biomass estimate technique based on forest backscattering database is developed, and is used to retrieve AGB of Changbai mountain area from ALOS PALSAR dual-polarization data. The forest growth model and the 3D forest radar backscattering model were combined to build a forest multi-polarization radar backscattering database. Then forest AGB was estimated based on this database using statistic regression method and look up table (LUT) method. Two types of LUT searching methods (nearest distance and distance threshold) were used to find the accurate results. The biomass retrieved from forest inventory data was taken as ground truth to evaluate the inversion methods and the precision of the AGB estimation. The inversion results derived from PALSAR FBD data shows that both the statistical regression method and nearest distance LUT method underestimate forest aboveground biomass. The distance threshold LUT method gives the better biomass estimation compared with forest inventory data, the mean absolute error (MAE) of the whole research area is less than 10 Ton/ha.