Velocity and direction of transverse wind are important parameters of atmospheric turbulence, largely determining the future wavefront pattern. Real time wind estimation contributes to predictive correction which is used to compensate for the time delay in the control systems of adaptive optics (AO) system. This paper presents two methods to estimate the transverse wind from wavefront reconstructed with Zernike modal using Shack-Hartmann wavefront sensor (SHWS). We compare the performances of block match method and optical flow method, as well as presenting a comprehensive error analysis. It is showed that the transverse wind estimation error will be minimum while optimal mode numbers were chosen. For both methods, the percent deviation of velocity estimation are kept within 30%, and the estimation error of wind direction are kept within 6 degree. The variances of the two methods increase with the slope measurement noise variances increasing. When the slope measurement noise is around 1 pixel, the variance of the transverse wind estimation approaches to theoretical lower bound, which indicates the estimation methods are efficient. According to the minimum estimation deviation, the optical flow method is preferable. Numerical results are used to verify the conclusions.