We discuss the user localization problem assisted by reconfigurable intelligent surfaces (RIS) in millimeter-wave (mmWave) multiple-input multiple-output (MIMO) orthogonal frequency division multiplexing (OFDM) systems. By exploring the inherent sparse scattering characteristics of mmWave channels, we propose a tensor decomposition based method to obtain the channel parameters (i.e. angle, delay) required for user localization. Specifically, the received signal is modeled as a low-rank third-order tensor fitting a typical polyhedral model which contains a factor matrix of channel parameters. Then, we design a structured tensor decomposition method to estimate the channel parameters by using the Vandermonde property of factor matrices. Finally, based on the least squares criterion, the location of user are obtained from the estimated channel parameters. Simulation results show that the proposed estimation scheme can achieve a centimeter-level resolution of user localization.