Abstarct: Velocity model building is a crucial step for the construction of a seismic image of the subsurface by any depth imaging method. A wide variety of different velocity model building methods is available. Reflection tomography is one of these methods, which is widely used in industry. One of the drawbacks of that method is, however, that it requires picking of reflection events in the seismic prestack data to provide the traveltime information for the tomographic inversion. This picking is extremely time-consuming, especially in 3D seismic data, and can become difficult or even impossible if the signal-to-noise ratio in the data is low. In this thesis a version of tomography called NIP wave tomography introduced by Duveneck (2004) is used for constriction of imaging velocity model. This technique makes use of traveltime information in the form of kinematic wavefield attributes. These attributes are the coefficients of second-order traveltime approximations in the midpoint and offset coordinates and can be extracted from the seismic prestack data by means of common-reflection-surface (CRS) stack method. The required input data for the tomographic inversion are taken from the CRS stack results at a number of pick locations in the CRS-stacked simulated zero-offset section. The picking process is done on post-stack sections and locations of the points do not need to follow continuous horizons in the stacked section. The problem of estimating the velocity model tomography is addressed in terms of an inverse problem (Tarantola, 1987). This problem is solved in an iterative manner. During each iteration, the difference of observed and modeled data is minimized and the model is updated. This procedure would continue until the misfit falls below a specified value. Modeling the observed data for the first time requires an initial velocity model. Initial velocity model in NIP-tomography contains a constant near surface velocity which increases linearly with depth. In the present thesis, after some introductory discussions, I firstly use some other functions instead of linear function, to produce initial velocity models and check the effect of each of them on NIP-tomography results. In addition to simple mathematical relations, the stacking velocity derived from kinematic wavefield attributes is used in NIP-tomography, as initial velocity model. In the end of the thesis, I evaluate the accuracy and validity of the velocity models derived from various initial velocity models by using them in migration of a real seismic land data and comparing the migrated sections. Finally, the velocity model which leads to best migrated section, would introduce as best velocity model. This velocity model, in turn, obtained from best initial velocity model.