高速M推定法を用いたロバストガウシアンフィルタの提案

概要

論文の詳細を見る
Gaussian filter to extract the roughness profile from a primary profile suffers strongly from the outliers mingled in the data. To solve this problem, several researches to put this method robust Gaussian filter have been proposed. However, the calculation cost becomes large due to the large number of repetitive applications. In this research, we propose a new robust and fast method by using the fast M-Estimation method. Since the proposed method requires just a few times of repetitive applications of the rectangular smoothing filter, the computation cost could be drastically reduced. The proposed method is just in need of the analytical calculation without any repetitive computation. Moreover it was noted that, at the ultimate time when there are no outliers in the data, the proposed method could provide completely the same results of the original Gaussian filter method.