Abstarct: Seismic imaging in geological structures with rapid lateral variation in velocity needs depth migration. This process needs an exact distribution of velocity.For determining velocity model there are two kinds of substructure model: Blocky model and smoothed model. In this thesis smoothed model is used. Smoothing velocity model is unfavorable and the model becomes unreal. Before any depth migration there is velocity smoothing. Kirchhoff migration which is ray baxsed needs the velocity field to be differentiable that means specific amount do smoothing is needed. In this thesis a new method for velocity smoothing is introduced. Velocity smoothing is done with two potential field filters: surface trend filter and upward continuation filter. In the first step the initial raw velocity field, we use two different surface trend filter with various degrees (2, 3, 5, 10, 15, and 16) and upward continuation with different height (20m, 100m) to separate main velocity field (low frequency) and small component of velocity field (high frequency).In the second step we use low pass frequency filter to suppress various wavelengths to gain residual small component. Then we add up this component to the main velocity field to get smoothed velocity fields with different degree of smoothing. After that we apply post stack migration with different velocity models to get the best velocity model. Migration results shows that when we apply surface trend filter method, best velocity model comes out of second degree of surface trend filter and cutting the wavelength lower than 250 meters. We get the best velocity model in upward continuation filter method, while applying upward continuation filter with20 meters high and cutting wavelengths lower than 250 meters.