An attitude estimation method for flight vehicles based on statistic of image was improved. The method improved reduces calculation, and improves the real-time performance for attitude estimation. The bank and the pitch angle can be estimated rapidly with a higher robusticity. Its accuracy satisfies the requirement of attitude estimation to control flight vehicles. Several differentmethods were compared with simulation results, and an application was proposed, which resolved the problem of coupling between the bank and the pitch in the attitude estimation. The method improved is applicable for those which are equipped computer vision systems as navigation systems, especially for MAV(micro air vehicles) equipped only minitype computer vision systems. Range of the attitude angle to be estimated is expanded while computer vision system equipped by the method is adopted with a MIMU(miniature inertial measurement unit) assistant, and the accuracy is higher correspondingly.