Data-Driven Motion Estimation with Spatial Adaptation
Besides being an ill-posed problem, the pel-recursive computation of 2-D optical flow raises a wealth of issues, such as the treatment of outliers, motion discontinuities and occlusion. Our proposed approach deals with these issues within a common framework. It relies on the use of a data-driven technique called Generalized Cross Validation (GCV) to estimate the best regularization scheme for a given moving pixel. In our model, a regularization matrix carries information about different sources of error in its entries and motion vector estimation takes into consideration local image properties following a spatially adaptive. Preliminary experiments indicate that this approach provides robust estimates of the optical flow.