Hierarchical estimation of optical flow using spectral energy method

The estimation of optical flow is an important problem in the dynamic image analysis. There are various methods of estimating the optical flow. The gradient method and matching method are based on the operation in the space-time domain. While spectral energy method is based on the frequency analysis and has robustness against noise. In this paper, we propose a hierarchical estimation of the optical flow using spectral energy method. In the proposed method, we make a set of hierarchical images for each frame. A global motion which is estimated at the coarsest resolution image sequences is corrected by using finer resolution image sequences. Consequently we can estimate both global and local motion accurately with robustness to noise.