Manual seismic horizon picking is the least efficient interpretation technique in terms of time and effort. The loop-tie is a key “element” and the most time-consuming task in manual horizon picking, which ensures the accuracy of horizon picking. Autopicking techniques have been used since the early 1980s. However, there are few studies simulating the procedure of manual seismic horizon picking and quantitatively evaluating the autopicked horizons. In our proposed method, we perform autopicking on inline and crossline seismic vertical slices independently, similar to the manual horizon picking procedure. We then evaluate the picked horizons using a loop-tie step similar to the loop-tie checking in manual horizon picking. To simulate the loop-tie step in manual picking, we define two dip attributes for each time sample of seismic traces, which are the “left” and “right” reflector dips. We only preserve the portions of the tracked horizons that meet the defined loop-tie checking. Next, we merge the tracked horizons centered at the seeded seismic traces. The two-way traveltime of the merged horizons functions as a “hard” control for the final step of autopicking. Finally, we use the seismic dip attribute to track the horizons over the seismic survey under the hard controls. The real data demonstrate that our algorithm can extract accurate horizons near discontinuities such as faults and unconformities.