Dual-frequency joint phase unwrapping does not need to satisfy the Iton hypothesis, so dualfrequency interferometry can effectively extract elevation information from large terrain undulation areas. This paper presents an effective method for Dual-Frequency Interferometry (DFI) SAR terrain elevation reconstruction, aiming at the low precision of elevation reconstruction in DFI processing. This method improves the key steps in the conventional process. Firstly, Nonlocal Interferogram Estimation (NL-InSAR) technique is used to accurately estimate amplitude map, coherence coefficient, and Interferometric phase before the registration in different bands. After filtering, the filtered amplitude information of each band is used to realize the registration of Interferometric phase in different bands. Then clustering analysis technique is used to mark the effective points and noise points of unwrapping phase, and these effective points are used to filter the mean value of joint phase unwrapping. The airborne measured data used in the experiment include the complex data of C band and X band main and auxiliary SAR images of the same scene. In view of the measured data processing, the proposed method provides better results of elevation reconstruction.