Abstract:Correct computation of intercept ( A) and gradient (B) has important meaning for AVO anal- ysis. In practice, the existence of different types of seismic noises causes significant difficulties to reliably estimate intercept and gradient. In view of the issue that abnormal noises existed in seismic data made the intercept and gradient, computed by traditional least square method, unreliable, the paper presented the method using robust estimation algorithm to compute the AVO intercept and gradient. The least square method is on the basis of normal distribution of measured data errors,but robust estimation algorithm is suitable for Cauchy distribution that the tail of measured data errors exposed outside. The estimated results of practical model showed that, when the AVO response was seriously distorted by local abnormal noises, using the least square regression method that supposes the noises is Gauss distribution is difficult to estimate the intercept and gradient, while the robust estimation algorithm can give more precious computational results. Therefore, the algorithm presented in the paper has some referent meaning for multi-parameters AVO inversion and other geophysical inversions.