Slope is one of the key factors which had a close relationship with soil and water losses and land use planning, and it is one of the most important pieces of information in many natural source spatial databases. Slope was usually extracted from digital elevation model (DEM) in regional scale. However, it is difficult to obtain DEMs with grid size less than 10m from continental or country scale topographic maps that are needed to include topographic changes due to terraces because of manpower and financial restrictions. Thus the high resolution DEMs were usually obtained by super-resolution reconstruction or down scaling from low resolution DEMs. As a basic data source, high resolution DEMs created by super-resolution methods do influence the extraction of slope. Dry terraced fields in the hilly loess region around Longquan of Yuzhong County were taken as study area. DEMs and orthophotos obtained in March, 2015 by UAV-based photogrammetry were used. The resolution of the DEMs data was generated in 5m, 10m, 20m and 40m respectively for slope produce. Sparse mixed estimation method (SME), nearest method, bilinear method and cubic method were used for DEMs reconstruction. Finally, the slope was calculated from original DEMs and reconstructed DEMs. The result showed that the SME method was the best method in DEMs super resolution reconstruction for extraction of slope.