Aimed at the problem of large random drift of microelectromechanical systems (MEMS) gyroscope and the negative impact of filter due to the outliers in measurement, an outlier rejecting and adaptive filtering method is proposed for MEMS denoising. The Allan variance is used to estimated the parameters of the measurement noise variance, thus the connection between Kalman filter and estimator of measurement noise is avoided and the filter divergence is suppressed effectively. On this basis, the outlier rejecting algorithm of new information is introduced. The negative influence of outliers is removed by revising the new information and the denoising effect is enhanced. The experimental result from the practical gyro measurement show that the mean square error and angle random walk have been reduced significantly by using the proposed method, which prove that the proposed filtering method is effective for denoising of MEMS gyroscope.