DEM (Digital Elevation Model
) is the model that gives elevation of each point of the earth surface in discrete form in a 3-D space. Image registration, implies registration of multi-temporal, multimodal, multi-resolution, images of the same area. Registration of DEMs is now a days gaining a lot of popularity among the research community. DEM registration, in this case, is the registration of multi-temporal DEMs. Popular techniques for feature extraction and matching include wavelet approach, robust SIFT, or are based "super-points". This paper presents the DEM registration scheme based on watershed transformation, followed by two post-processing steps of clustering and morphological operations which is applied on both the DEMs -candidate, as well as, reference DEM. Chain coding based matching is concluding step of the complete process. The system is semiautomatic i.e. expert input is considered before the final registration. Experimental results give good outcomes as shown from the error matrix analysis of RMSE, and PSNR. The system may be extended by using fuzzy classification and contextsensitive learning.