In order to realize large-scale absolute surface reconstruction, a generalized iterative optimization method for solving the three-flat problem is studied. First, the idea of model-based absolute surface reconstruction is proposed, which considers the problems of absolute surface reconstruction as inverse problems. Then we take the three-flat problems as an example, we introduced two generalized iterative optimization methods for three-flat model. Finally, by both simulation and experiment, it is concluded that the block SOR method with an optimal relaxation factor converges much faster and saves more computational costs and memory space without reducing accuracy. Both simulation and experimental results indicate that the proposed iterative optimization methods are effective for solving the three-flat problem with pixel-level spatial resolution and the measuring precision of two separate measurements is 0.6 nm rms, and the cross-check test result is 0.8 nm rms. It is concluded that the proposed method can correctly reconstruct absolute figures with high efficiency and pixel-level spatial resolution.