The analysis of popular algorithms of the inverse problem dedicated to surface refraction tomography is discussed in the paper. Surface refraction tomography is a method, which is able to estimate a velocity of elastic wave on a layer boundary. This work presents comparison between popular methods of inversion of seismic data: Singular Value Decomposition (SVD), genetic algorithm and ART and SIRT iterative methods. Stability of results and error of estimated solution were analysed. Numerical experiments show, that the most effective method is SVD (it gives acceptable solution in shortest time). Genetic algorithm, which is one of the basic methods for non-linear inversion, gives the best inversion result, but its computation time is the longest one. Significant improvement of convergence was obtained by using some iterations of SIRT in single iteration.