Abstract: Equalization is an effective technique to overcome the multipath fading and improve the performance of radio communications. In visible light communication (VLC), since the propagation medium and the realization approach are different from that of radio communications, conventional equalization algorithms may not be applicable any longer. This paper analyzes the characteristics of visible light channel, summarizes the existing least mean square (LMS) equalization algorithms, and then proposes an improved variable step-size LMS algorithm for VLC. In the proposed algorithm, step-size is adjusted based on the error function instead of the Sigmoid function and Versiera function. Moreover, in order to reduce the complexity and the difficulty of hardware implementation, the error function is approximated by a piecewise elementary function. Simulation results show that, under visible light channel, the proposed algorithm can get faster convergence speed and better tracking performance with low steady-state errors compared with existing algorithms, and thereby are more suitable for VLC systems.