運動ベクトル場の推定に用いる非線形拡散項の離散化法

概要

論文の詳細を見る
In this paper, estimation of motion vector fields containing boundaries from noisy data is investigated, where the boundaries mean edges or spatial discontinuities. Then, regularization methods which can retain boundaries are required. We adopt evolution equations with non-linear diffusion terms derived from a logarithmic stabilizing functional. Estimates are obtained in the process of time evolution of the equations. Here, discritization method become critical issue. We propose a novel discritization method for these terms, which is a sort of minmax scheme. Results of simulations are displayed, where better estimates and more distinct boundaries of them are obtained with propesed method than with naive method.
三重短期大学の論文
2004-03-20