By utilizing our newly developed multiple arrival (transmitted, reflected, refracted, and mode‐converted) tracking algorithm (multistage irregular shortest‐path method), in combination with the conjugate gradient method to solve the damped minimum norm, constrained least‐squares problem, we present a simultaneous inversion algorithm to recover both the velocity information and the subsurface interface geometry. In the inversion process, we introduce different weighting factors according to the different picking errors for the various seismic arrivals and normalize the two types of model parameters that form the elements of the Jacobian matrix so as to balance the influence on the travel times of the different velocity variations and reflector depths. From numerical tests and by comparison with a subspace inversion algorithm, we show the new approach to be a practical and efficient way to improve the spatial resolution and reduce artifacts in the reconstructed image.